Appendix N: understanding advertiser demand for digital ..., Study notes of Qualitative research

Download Appendix N: understanding advertiser demand for digital ... and more Study notes Qualitative research in PDF only on Docsity! N1 Appendix N: understanding advertiser demand for digital advertising Introduction This appendix seeks to understand advertiser demand for digital advertising by looking into the behaviour and characteristics of advertisers who purchase digital advertising from Google and Facebook. We first explore the characteristics of advertisers who use Google and Facebook to purchase digital advertising based on information from 2019 data, looking at customer distribution, information on the industries the advertisers are present in, and campaign shares. We then consider how advertisers make the choice of the appropriate form of advertising, looking at the key performance indicators advertisers may have, the choice of digital advertising, and the substitutability between different forms of advertising. Next, we discuss some further choices made by advertisers once they have chosen to purchase digital advertising, and discuss the role that the advertising interface plays, the services media agencies can provide to advertisers, and the thought process behind planning a campaign. We then set out the journey advertisers go through when purchasing search advertising on Google Ads and display advertising on Facebook Ads Manager. We discuss the process of creating a campaign and the options presented to advertisers throughout, bringing in advertiser views on these choices. We then consider the transparency of Google’s and Facebook’s advertiser interfaces, and discuss some default settings in the interface and how these might affect advertiser behaviour. In addition, we discuss advertiser views on the extent of the information provided to the advertiser in the interfaces, the transparency of the auction process, costs and advertiser control over their budget. This appendix combines several pieces of evidence to inform our views, including customer list data from Google and Facebook, views expressed by media agencies and large advertisers through information requests and calls, qualitative research that we commissioned to obtain further views from mostly smaller advertisers, our own desk research of the advertising interfaces, and responses from Google and Facebook to our information requests. N2 Advertiser characteristics To investigate the characteristics of Google’s and Facebook’s advertiser customers, we have undertaken a descriptive analysis of advertiser and industry distribution and campaign shares. We have conducted this analysis on data submitted by Google and Facebook on their customer lists. There are two important factors to note regarding the customer list data that this analysis is based upon: The basis for which customers, and their associated campaigns, impressions and spend on the platforms, were included in the lists supplied by each company were not consistent with one another. The basis for which customers, and their associated campaigns, impressions and spend on the platforms, were included in the lists supplied by each company were not consistent with the way we have defined the UK markets for search and display advertising in our analysis in Chapter 5. This means that the absolute values for total advertising spend on each platform in search and display advertising may not match up with the headline figures set out in our main report. For more detailed discussion of the customer list data, please see the Annex at the end of this document. Google advertiser characteristics The Google dataset contains data on advertisers on Google Ads that were served in the UK in 2019. It is comprised of [200,000-250,000] unique parent entities (ie advertisers),1 spending £[6-8 billion] on an estimated [7-10 million] campaigns and [0.5-2 trillion] ad impressions. Search advertisements made up £[6-8 billion] and display advertisements made up £[400-700 million] of the ad revenues in the dataset. Out of the total unique advertisers, around 20- 35% of advertisers used a media agency, and they accounted for around 70- 90% of total spend, 70-90% of campaigns run and 70-90% of all impressions.2 Figure N.1 shows the breakdown of spend, number of impressions (ie adverts), and number of campaigns by ad product type. 1 Based on Google’s submissions, we note that the data was broadly provided at the parent level, which may not always be inferred to represent a unique advertiser. 2 We note that the variable identifying whether the advertiser uses a media agency may not be fully accurate in accordance with Google’s submissions, and some relationships with agencies may not be fully represented. N5 Figure N.3 Google number of advertisers, split by 2019 Google Ads advertiser spend groups Source: CMA analysis based on Google data. This shows that the majority of Google’s advertisers spent between £100 and £5,000 on Google advertising and in particular between £100 and £1,000. Industry distribution In Figure N.4 we present the total spend by all advertisers and the number of advertisers within each industry.3 Based on Google’s submissions, we note that the data provided on industries may not be an accurate representation of the actual industries in which the advertisers operate, and these are not provided by the advertiser directly. Due to the limitations of the data, we have treated these results with caution and have only made broad generalisations about our findings.4 3 In the data as provided by Google, the column for the vertical (which is used as a proxy for industry) often included two or more categories (as an example, one advertiser account might be classified as one vertical while other accounts might be classified differently). For ease of analysis, the data has been re-coded such that these advertisers are counted as being in all industries listed. While this method is useful for analysis, this also means that the sum of spend and the number of advertisers will be greater than the total spend and number of advertisers due to double counting. We note that it is possible that the advertisers who were listed in multiple verticals might spend the majority of their money on one of these verticals, and little on the others, which would not be captured by this graph. As a sensitivity, similar graphs have been created with the advertisers who were present in several industries belonging to a ‘mixed’ category, and the general distribution across both spend and number of advertisers appeared very similar in both cases. The biggest difference in spend was for [], which prior to including those in several industries was [] and is now []. 4 Despite the limitations of this data, we note that our findings for spend are broadly consistent with the IAB UK & PwC Digital Adspend Study 2019 findings, which show the distribution of spend of five media agency groups in paid search, split by industry. We note that the main difference in our findings is that the automotive sector is not as prevalent. Where our results differ from IAB’s, this may partially be explained either by a difference in classification of industries, or because our figures also include some display advertising. N6 Figure N.4: Total spend and number of advertisers, split by industry Source: CMA analysis based on Google data. This figure shows that there is a wide distribution of industries that spend significant amounts on advertising on Google Ads. Industries that appear to account for relatively high levels of ad spend include retail, finance, travel, and media and entertainment, as well as business and industrial markets and classifieds and local. Demand appears more concentrated in industries such as finance, media and entertainment, and travel, where a considerable amount of spend comes from a relatively smaller number of advertisers. Comparatively, demand is more fragmented in industries such as business and industrial markets, and classifieds and local, where there are a larger number of advertisers. Campaign shares We have also investigated how spend differs between different campaign types. Google has provided data on campaigns and adverts that are paid for on the basis of four different types: cost per click (CPC); cost per impression (CPM); cost per conversion (CPA); and N7 other cost types, including for example cost-per-engagement (CPE), fixed cost-per-day (CPD), cost per view (CPV), cost per lead (CPL). Figure N.5 shows the share within Google of these campaign types by three different weightings according to the data provided, including both search and display.5 This figure shows the breakdown by spend, number of impressions, and by the number of campaigns.6 Figure N.5: Share of campaign types (CPC, CPM, CPA, other) by spend, impressions and campaigns Source: CMA analysis based on Google data. This shows that in all three cases, more than 85% of advertisers or their agencies (in terms of ad spend, number of impressions and number of campaigns) chose to run campaigns and adverts paid for on the basis of CPC, with a small percentage opting for CPM and other means. The least popular method appears to be CPA (at least relative to CPC and CPM). Facebook advertiser characteristics The Facebook dataset contains data on UK-based advertisers who targeted exclusively UK-based users in 2019. It is comprised of [over 1 million] advertisers, who collectively spent £[1-1.5 billion] on [over 9 million] campaigns in 2019.7 Non-video advertisements make up [over £600 million] 5 Search advertising is generally bought on a CPC basis and display advertising is generally bought on a CPM basis. 6 We note that caution should be given when interpreting these results due to the limitations of the dataset provided, as the data came from different sources and information for all campaigns was not matched between the data tables. 7 The total spend in 2019 as obtained from the customer lists is considerably lower than our estimates of Facebook’s revenue in 2019 as reported in Chapter 5. This is likely as the 2019 customer list data only captures spend by UK advertisers who advertise exclusively to UK users. N10 Figure N.8: Total spend and number of advertisers, split by industry10 Source: CMA analysis of Facebook data. This shows that there is a wide distribution of industries that spend significant amounts on Facebook advertising, as with Google. Industries that are prominent across both Google and Facebook include retail, entertainment & media, financial services, and consumer packaged goods. Demand is particularly concentrated in consumer packaged goods, ecommerce and financial services, where a considerable amount of spend comes from a relatively smaller number of advertisers. Campaign shares We have also analysed how spend differs between different campaign objective types. Facebook’s advertiser interface asks advertisers to specify a campaign objective out of the broad categories:11 awareness; consideration; and conversion. 10 Unknown advertisers represent a minority of revenue but a significant proportion of UK-based advertisers who target exclusively UK-based users in 2019. 11 For a full breakdown of the options presented to the advertiser and what these mean, please see the ‘Display advertising using Facebook Ads Manager’ section. N11 Each of these objectives represents a campaign focused on customers at a different stage of the campaign funnel shown in Figure N.10. Advertisers often run multiple campaigns and each of these will have its own objective, so the more complex an advertiser’s overall strategy is, the more campaigns will likely need to be run by the advertiser. Figure N.9 shows the breakdown of spend and the number of campaigns by campaign objective. Figure N.9: Share of campaign objectives Source: CMA analysis of Facebook data. This shows that the most popular campaign objective is consideration, followed by conversion, and awareness is the least popular of the three. We note that many media agencies and advertisers have indicated to us that display advertising is particularly effective for increasing brand awareness.12 However, it is not clear how this aligns with Facebook’s campaign objectives. It may be that consideration and conversion campaigns are also effective in increasing brand awareness, as even customers who do not interact with an ad will see it, which may help explain why these objectives are frequently chosen. Choice of advertising When advertisers decide to purchase advertising, there are several decisions they must make in the process, including considerations of what their key performance indicators (KPIs) should be and what format is best suited to their needs. This section discusses what advertisers and media agencies have told us about how they use KPIs and choose digital advertising, examining the substitutability of different forms of advertising. 12 See the ‘Search vs display’ sub-section. N12 KPIs used in digital advertising Business objectives are typically defined more precisely according to key performance indicators (KPIs) relating to the outcomes the advertising campaign is hoping to achieve. Typically, these will relate to either brand awareness at one extreme or to specific actions (eg sales, enquiries) at the other. Several advertisers and media agencies we contacted referred to the ‘marketing funnel’ or ‘purchase funnel’, which illustrates the mental steps along the consumer journey to making a purchase of a product or service – from awareness to consideration to purchase. This is illustrated in Figure N.10. Figure N.10: The purchase funnel Source: CMA. At the top of the funnel, KPIs relate to improving the awareness of consumers that are ‘out-of-market’ and are not currently aware of the product or service. At the bottom of the funnel, KPIs relate to selling to those customers that may have expressed some preference for the product or service but have not yet made a purchase. The funnel will be specific to each business and can evolve over time with the business. Sometimes KPIs are expressed in absolute terms eg a target increase in sales, reach or brand recognition. More commonly they are expressed as a return on cost eg return on investment (ROI) or return on advertising spend (ROAS), the aim being to maximise the benefit to the advertising from allocating expenditure effectively across multiple advertising channels.13 13 Or equivalently to minimise the reciprocal eg cost per acquisition (CPA), cost per reach. N15 products (eg gambling, pharmaceuticals); tech stacks and the capability to integrate data (ie DMPs). Of the advertisers contacted during the qualitative research who chose to use either Google or Facebook or both for digital advertising, there were many similar reasons for their choice. Six key reasons were given, these were: awareness and reputation – Google and Facebook are both well-known big brands which are seen as a ‘safe’ choice; coverage and targeting – linked to their size, both Google and Facebook are recognised as covering a very high percentage of the population, and are therefore well suited for both ‘average’ audiences and niche audiences, with the ability to target by ‘need’ based on their key-words in Google; meets company objectives – Google is able to allow the advertiser to access new customers and specifically target sales, while Facebook is useful for brand building and awareness raising, reach, and increasing sales by driving traffic; easy to use – perception that both Google and Facebook are relatively easy to use, with minimal knowledge and/or cost needed to get started; value for money – both Google and Facebook are seen as better value for money than offline advertising, driven primarily by the ability to target specific audiences; ability to retarget and use sub-brands – a minority of advertisers viewed this as an initial reason for choosing Google and Facebook. Advertisers also identified some differences in their initial reasons for choosing to advertise on Google and Facebook, with the key difference relating to Google’s search engine functionality. When deciding whether or not to use Google, advertisers were also deciding whether to deploy a ‘search’ or ‘display’ strategy (or a combination of both). Advertising type substitutability To consider substitutability, we have collected views from advertisers and media agencies on how they decide to allocate budget between different media types and the comparative advantages and disadvantages of each. As discussed in the ‘Advertiser characteristics’ section, large advertisers make up the vast majority of revenue Google and Facebook receive from advertisers. We have collected views from 23 large advertisers and four major media N16 agencies that collectively represent a large proportion of customer demand for Google and Facebook. In addition, we commissioned qualitative research to explore the experiences of 24 mostly smaller advertisers using Google and Facebook. For further discussion of the advertisers and media agencies contacted and the qualitative research, please see the Annex. Multi-homing We have collected views from both large and small advertisers on the extent to which advertisers use platforms other than Google and/or Facebook (ie advertisers who are multi-homing). As large and small advertisers differ in terms of their preferences and behaviour, we first look at views expressed by large advertisers and media agencies, and then smaller advertisers. All of the large advertisers that responded to our questionnaire used multiple platforms to buy digital advertising. Most of these advertisers use media agencies and are generally quite knowledgeable about digital advertising and may be more sophisticated in its usage. The media agencies and large advertisers we contacted told us that the benefits to using multiple advertising platforms was that this provided advertisers with greater variety in several dimensions, allowing them to better achieve their KPIs. Using multiple platforms may allow advertisers to improve audience reach, access audiences in a variety of mindsets,16 use multiple formats and have more flexibility in optimising targeting capability and cost effectiveness. One media agency noted that the cost of only using one platform would usually relate to missing out on inventory or data that are exclusive to certain platforms (for example YouTube inventory in DV360 or Amazon data in Amazon DSP). Some advertisers mentioned that a multi- platform approach also avoided them being overly reliant on a single platform as this could create inefficient dependencies. Media agencies and advertisers told us that it is not possible fully to compare and contrast media input metrics consistently from digital platform to digital platform. Different platforms can have different approaches to measurement and custom metrics that require manual effort to combine reports and assess impact. There are often issues moving data in and out of platforms due to concerns from large scale players around value of owned data, as well as the need to consider applicable privacy laws (eg if moving data from a platform involves an international transfer of that data). Cross-channel attribution is a challenge because platforms guard their data for both monetary and data 16 For example, an audience actively searching for a solution (eg on Google, Bing or Amazon), has a different mindset to an audience consuming content on a news site. N17 protection purposes. Cookie based attribution neither allows for de-duping device impressions nor accurately represents the conversion path, and further also requires user consent. However, most respondents suggested that while there were issues of comparability across platforms, these were not sufficient to stop them using multiple platforms. One media agency noted that these issues are also true of offline advertising channels and that the role of any agency is to understand the differences between the platforms and the relative merits of each. Smaller advertisers may be more likely to only use Google or Facebook rather than smaller platforms, as Google and Facebook have sufficient scale and reach for the advertisers’ needs and offer greater convenience due to the ease of their interface than using multiple platforms. Of the 24 advertisers contacted during the qualitative research that we commissioned, most were smaller advertisers. Roughly half of the respondents were single-homing (ie using only one platform). Amongst multi- homers, most were multi-homing solely using Facebook and Google. Only a minority were multi-homing using other providers in addition to Facebook and/or Google. Of those advertisers who were multi-homing, the majority spent much more on one provider than any others, indicating that they behaved similarly to those advertisers using only one platform. Of the advertisers who were only using one platform (ie Google or Facebook), these were primarily the smallest advertisers, namely micro or small SMEs who are less sophisticated and less tech savvy, who either spend less on advertising or who have a simpler product mix. The decision to use only one platform was rarely reached after a detailed review of all possible approaches and providers available. In fact, the decision was often driven more by a lack of knowledge or time to engage with other opportunities and could be more of a default non-choice than an active decision. Some respondents had simply not even considered any other providers as they were happy with the results they currently got and had not contemplated the need to do more. However, some advertisers had given this more thought, and there were a number of key reasons specified for why these advertisers were only using one platform. Some respondents stated that they had no perceived need to advertise elsewhere, as their advertising was performing at a level that it was unnecessary to advertise elsewhere. However, some advertisers acknowledged that this may have been driven by a lack of knowledge of effective alternatives, or due to their product mix not being conductive to other forms of advertising. Some respondents stated that they viewed their current situation of using only one platform as ‘good enough’, with acceptable ROI N20 easier to judge ROI; easier to ‘do it yourself’ compared to traditional advertising, requires minimal investment and knowledge, easy to get started; faster to implement and more flexible than offline advertising; and perception that it is what everyone else is doing. Some of these reasons indicate that there is limited substitutability between online and offline, and that advertisers may consider that there are greater barriers to entry to offline sources due to them requiring greater investment or knowledge. However, we note that the advertisers contacted during the qualitative research all advertised online on either Google or Facebook at least to some extent, and these views may not reflect the views of all advertisers. One SME advertiser21 in the golf industry said that digital advertising is considerably more effective than offline traditional media. They use Facebook to reach interested golfers by using the keyword search and target users within an 80-mile radius, with £100 getting them around 35,000 interested golfers. They say this is a more effective strategy than buying quarter of a page in a national golf magazine, which will cost considerably more and will not be viewed by as many people. In this case, online advertising has a better ROI for the advertiser. Some large advertisers thought that video-on-demand is the next best alternative to Facebook for display. For example, one advertiser pointed out that they consider ITV, Channel4 and Sky as alternatives to Facebook when deciding where to advertise, and that the advantages of advertising on TV were the high-quality inventory, high viewability, big size of screen and audio on and brand safety. Another advertiser noted that 4OD has good content and high video completion rates and they rank 4OD and Sky higher than Facebook for achieving relevant KPIs in video advertising. Advertisers may then treat offline advertising as a complimentary channel within their campaign to achieve certain KPIs. Those who consider video-on-demand a viable route mentioned that this method does not allow for viewability and audience measurement and some do not offer the option of buying inventory programmatically. Moreover, one advertiser mentioned that video-on-demand is more expensive than TV and it has some brand safety and appropriate content concerns. Hence, this 21 The advertiser had between 10 and 49 employees. N21 advertiser said they need to use third parties to verify inventory is human- viewed and in brand safe environments. Of the advertisers contacted during the qualitative research who did use offline media, some examples of media format include using local newspapers, mailshots, leaflets, posters and national newspapers and magazines. The reasons advertisers gave for using traditional media included: a matter of habit – they have always used traditional media; wanting to support local businesses (eg magazines, events etc), which is unrelated to advertising; building presence locally (eg ads in local papers, posters); and targeting specific postcodes or streets with direct mail or leaflets. We note that these reasons mostly differ from the reasons stated by advertisers for why they purchase search or display advertising, which is discussed below. This may indicate that advertisers may use offline advertising as a complimentary channel within their campaigns to achieve certain objectives. Search vs display All media agencies and most advertisers we contacted told us that search and display advertising are not substitutable, mainly because they perform different roles within the customer purchase journey. Search is intent-based advertising designated to provide immediate answers to those consumers that have already shown interest in buying the product and are at the end of the purchase funnel (‘in-market consumers’), whereas display is suitable for raising brand awareness and reaching new audiences that might not yet have shown interest (‘out-of-market consumers’). This distinction is reflected in the advertising KPIs search and display are typically used to target. Almost all respondents suggested each had relative advantages in targeting different KPIs: display being more versatile and typically cheaper at driving reach and brand awareness objectives, while search being closer to the point of purchase and so better for targeting actions, conversions or sales. For example, some advertisers pointed out that display advertising largely drives demand generation and traffic volume whereas search largely drives demand conversion. In the same vein, another advertiser suggested that search has more action and sales KPIs than display. Another advertiser said N22 that search can consistently deliver on its core KPIs of impression share and click-through-rates, notwithstanding that display has improved in viewability and on-target accuracy over the years. Another advertiser observed that search is appealing because it drives lower cost per sale, given that customers are in-market and ready to buy. They further said that this type of advertising provides economies of scale with the sheer number of users who access the search platforms. Another advertiser noted that search advertising is the strongest channel for positioning the brand and products for ‘in-market’ customers. Some respondents noted that the distinction was not absolute in all cases, in that search can also sometimes be used as an upper funnel to build brands, and display can sometimes be an effective sales driver. One advertiser told us that as budgets for search advertising continue to increase this allows the advertiser to target a higher level of generic keywords and therefore move further up the purchase funnel. Another advertiser told us that search and display can sometimes be substitutable depending on campaign objectives, with another advertiser stating that the benefits of search and display are most similar for when there is a conversion goal as a key objective. Most respondents also noted that search is text-based and so has limited space for the client’s creative message, while display can be used for imagery and video. This means that display can use advertising creative to persuade undecided customers with no intention to purchase to move further in market. One advertiser pointed out that the capability to include creative content in display advertising enables them to showcase brand values and product range in an exciting way, and another advertiser considers that display offers a richer variety of formats such as static, video, in-feed and stories. Another advertiser noted that display is disruptive and aimed at reach and frequency across passive audiences. A few respondents mentioned that search and display advertising work well in tandem and one advertiser noted that they can complement each other through synergy (eg a customer sees a display ad, then searches for more information). One media agency said that display and search are most powerful when they are planned and optimised holistically with the objective of driving advertiser outcomes against a measurable KPI. Most advertisers contacted during the qualitative research tended to use either search or display as their main approach, with the other approach used to support it (if at all). This indicates that advertisers do not view search and display campaign as substitutable, but rather potentially as complimentary. As an illustration, a Scottish whisky shop may use search advertising to attempt to target conversions by selling individual bottles of whisky, and use display advertising to promote a local whisky tasting. N25 of time. One advertiser thought that Google will, in general, bring new products to the market quicker than Bing and have higher volumes of searches to test hypothesis. If these bid tests are successful, then they can be replicated on Bing. Some large advertisers identified benefits of using Bing. Several thought that Bing had lower CPCs than Google which made it attractive. Several advertisers were drawn to Bing due to the ability to target an older audience, with one advertiser saying Bing was better suited to advertising higher price products due to this audience. Some advertisers thought that an advantage of Bing was that it owns LinkedIn and is working on incorporating that data into its platform, which Google may not be able to replicate. One advertiser thought that Bing had a stronger control over the search partner network than on Google, and users tended to have a stronger conversion rate. Similarly, another advertiser thought Bing had less bidding competition and alternative audience opportunities. One advertiser liked that Bing could be used to advertise across both Bing and Yahoo!. Some advertisers contacted felt they had little option but to use Google due to its prominence in order to stay visible in their own highly competitive segments, indicating a lack of comparable alternatives for search advertising, despite the availability of Bing and other platforms. This was echoed by the advertisers contacted during the qualitative research, who thought that Google had almost complete coverage, such that any other search engine was rarely seen as adding any incremental benefit. One advertiser stated, ‘Google is around 90% of all searches. It’s the market!’. Within display Advertisers can purchase display advertising through owned and operated platforms, the largest of which is Facebook which also owns Instagram, with other platforms such as YouTube (owned by Google), Snapchat, Pinterest, TikTok, LinkedIn, Verizon Media and Amazon. Of the 17 large advertisers contacted who specified what platform(s) they used to purchase display advertising, 14 indicated that they used Facebook to some degree. Advertisers identified several advantages of using Facebook. Several advertisers thought that the main advantage of Facebook was its mass reach, and some advertisers thought that Facebook had good targeting capabilities linked to its mass reach, based on both user behaviour and geographical targeting. Some advertisers thought that Facebook had good ROI. One advertiser thought Facebook drove strong engagement and was useful for reaching a younger audience. One advertiser thought that the advertising formats on Facebook (ie ‘in feed’ adverts) produced better results N26 because they were less intrusive to users (ie more sympathetic to the user experience). One advertiser thought that some (but not all) newsfeed placements can have less of a brand safety concern as they are based on users’ preferences, though opinions on brand safety overall were mixed. Advertisers also used several other platforms, with some of the most popular ones named being Google (eg DV360 to purchase advertising on YouTube), Twitter and Amazon amongst others. There were several advantages of using Google highlighted by advertisers. Some advertisers thought Google had mass reach, and some thought that Google had good ROI. Some advertisers thought that DV360 had strong integration with Google’s Marketing Platform Products, which lead to easier, more efficient activity management. One advertiser thought that DV360 had access to unique data which was beneficial. Some advertisers also highlighted some advantages of other platforms, for example, one advertiser thought Snapchat was beneficial for a young audience, and some advertisers thought Twitter was good for an engaged audience or for tapping into big cultural moments. Another advertiser used several platforms in combination (eg Facebook, Twitter, YouTube, Pinterest, Instagram) for their mass reach. This may indicate some degree on complementarity for display advertising. Video vs non-video Within display, advertisers also must decide whether they want to create a video or non-video advert. Some media agencies we contacted told us that the decision to use video or non-video within display advertising was typically made early on in the planning process and may be driven by the decision of how best to convey the advertiser’s message. Some advertisers stated that choice of format for display (video vs non-video) will affect the KPIs of the campaign, which would limit the substitutability between video and non-video advertising, though one advertiser told us that both video and non-video display advertising can be used for both performance and brand advertising, indicating some limited degree of substitutability depending on the campaign KPIs. Some advertisers contacted during the qualitative research felt that the use of video was actively promoted by Facebook to ‘increase engagement’. Video was seen as more eye-catching than ‘static’ ads such as banner ads or wrap- arounds, but some advertisers felt that it wasn’t suitable for their products or they lacked the time, knowledge or money to produce video. On the other hand, we note that the decision of whether to run a video or non- video display advert is made very late on in the Facebook Ads Manager interface, namely during the final ‘Ad’ stage which deals with the final ad N27 visualisation and configuration.22 This may indicate that the interface treats video and non-video adverts similarly, as all choices related to campaign objectives, budget and targeting amongst others are made before the decision on whether a video or non-video advert will be run. Programmatic vs direct purchase Most display advertising is bought via open auctions using programmatic technology rather than direct from publishers. Media agencies told us that this had several advantages. One media agency mentioned that there can be increased wastage when buying directly from publishers, as direct buys come in bulk on a CPM basis. Within those bulk buys can be impressions against consumers not desired by the advertiser. In contrast, using open auction allows the advertisers to ‘cherry pick’ the right impression delivered to the right audience and only buy and pay for these impressions therefore reducing waste. In the same vein, some media agencies consider that targeting is better with programmatic buying, as the buyer is layering on audience sets that fit within the brand’s target audience and setting frequency rules across multiple publishers (for example it is possible to identify users that have already been exposed to the ad and avoid duplication and un-targeted media delivery). One media agency noted that using programmatic buying provides access to data assets, for example platforms like Google and Amazon give unique access to targeting based on audience data. They also said that it is possible to use clients’ first-party data (like cookies and CRM) to segment and target audiences. Moreover, the volume of data available for buyers in open auctions allows advertisers to adapt creative in real-time to better suit the time and place that a user is seeing the ad. Some media agencies also said that using programmatic buying can be more cost-efficient. For instance, one media agency noted that the buyer is able to bid for an impression based on a value ascribed through the algorithm within the DSP, which can be optimised against an outcome (eg engagement, site visitation or sale). This may mean that the individual cost of an impression may be higher, but typically the cost per outcome is lower, given the reduction in media wastage flagged above. Another media agency said that buying in open auctions offers access to large volumes of inventory and this scale allows media agencies to optimise against client goals more effectively, as the algorithms which power the campaigns have more opportunity to optimise 22 For further discussion on this, see the ‘Display advertising using Facebook Ads Manager’ section. N30 best practice ‘strategic’ advice and external expertise in designing campaigns and allocating budget across advertising channels; greater economies of scale and scope in resource requirements (manpower, technology and expertise); and preferred trading arrangements eg volume discounts, negotiated by the agency. Smaller advertisers more commonly go directly to the platforms to use their self-service interfaces eg Google Ads or Facebook Ad Manager. As discussed in the ‘Advertiser characteristics’ section, around 20-35% of Google’s advertising customers on Google Ads in 2019 in this dataset used a media agency, and these customers accounted for around 70-90% of spend, indicating that many of the advertisers who used a media agency were primarily larger advertisers who spent considerable amounts on digital advertising with Google. As the vast majority of advertisers on Google Ads in 2019 were smaller advertisers (as seen in Figure N.2 and Figure N.3), these figures indicate that some smaller advertisers do use media agencies, but many do not. Of the smaller advertisers contacted during the qualitative research, many stated that they had not considered using a media agency as it slowed down their decision making and there was a perceived loss of control to a third party. Of the smaller advertisers contacted who did use a media agency, many saw this as outsourcing an element of their business that they themselves weren’t skilled in, as they had decided to focus on other elements of their business. Very large advertisers, whose business is heavily reliant on advertising (for example online travel agencies (OTAs) and digital comparison tools (DCTs)) may also go directly to the platforms, effectively performing the media agency functions in-house. However, in this case, rather than using the off-the-shelf interface they build their own bespoke buying tools and connect directly to the platform’s API. Several of the large advertisers we contacted bought advertising using an in-house team, noting that this ensured they developed greater in-house expertise and allowed them to purchase more efficiently (by eliminating media agency costs). Media agencies are fairly concentrated, with five major media agency groups – WPP, Omnicom, Publicis Media, Interpublic and Dentsu Aegis. Of the large advertisers we spoke to who specified which media agency they use, only one out of 17 did not use a media agency in one of these five media agency groups. N31 Media agencies perform a range of functions including communications strategy, media planning and buying, digital execution, content creation, entertainment marketing and sponsorship, data and analytics to provide data- driven marketing solutions for clients, programmatic services, search, social and customer relationship management (CRM). Media agencies are typically principals at law when buying space from media owners (though in a minority of cases may act as agents). Acting as a principal implies that they effectively resell the advertising inventory they purchase to their clients, taking on some of the risk that the advertising fails to achieve agreed service levels. In most cases media agencies are remunerated on a commission basis (though sometimes on a fixed fee or FTE basis and sometimes with performance fees). In general, media agencies exploit their scale to negotiate volume rebates with media owners based on aggregate spend across the media agency. They typically then share these rebates proportionally with their advertising clients. However, media agencies have told us that their ability to negotiate discounts from the large digital advertising platforms (eg Google, Facebook and Amazon) is limited, as the vast majority of inventory is sold by auction and the large platforms tend not to offer discounts. There are some opportunities to negotiate volume discounts from smaller platforms or media owners, particularly those that work on a fixed-price basis when selling to advertisers. Some of the advertisers that responded to our questionnaire had recently moved certain functions of their media buying in-house. Most of the media agencies we contacted noted that that ‘in-housing’ was becoming increasingly popular and that as a result direct competition from the platforms had increased over the last five years. They noted that this was sometimes motivated by a desire to reduce costs and sometime by a desire to shorten the gap between data generation and data application in a controlled and secure fashion. However, one media agency did not see this trend as competition per se but rather a natural evolution of relationships with its clients, where agencies are more valued for their strategic consultation around media execution than simply their ability to have a team available to drive campaign delivery. Planning a campaign Media agencies and large advertisers told us that, typically, advertising campaigns are planned by first defining the business objectives and available budget to achieve these objectives, though some smaller advertisers told us that they sometimes only have an approximate idea of the budget they will N32 need. Advertisers use digital advertising to reach a range of objectives, though these can be broadly categorized into raising brand awareness or driving sales and conversions. For example, a business objective might be to maximise an increase in sales or to improvement in brand recognition, given a certain budget. The advertiser will also set the budget for the campaign, the target audience, the campaign duration and the creative concept and formats available (as assigned by its creative agency). In general, the level of sophistication in planning process of the campaign will depend on the experience and confidence of the advertisers. Advertisers, or commonly the media agencies advising them, will then determine a series of metrics that are used for campaign optimisation when deciding how to allocate budget across different media. These metrics will vary dependent on the channel and objectives. They may be simple actions that can be tracked online in real time (eg test drive requests submitted for an automotive client) or proxy metrics for a more considered purchase (eg key mortgage related pages visited in a particular order for a financial client). One media agency stated that if they are using YouTube, they will make a recommendation on whether they should optimise towards reach or narrower audience focus based on the client’s business objectives and the role that the channel needs to play in the plan. The interfaces Facebook Ads Manager and Google Ads correspond with the views expressed by advertisers, with these interfaces asking the advertiser to first specify some kind of marketing objective or goal, as further discussed in the following section. Further choices relating to budget, the target audience, bidding strategy, ad schedule, and ad visualisation amongst others are all made subsequently after this initial campaign objective or goal choice. Google Ads and Facebook Ads Manager This section goes into detail on the advertiser journey of purchasing search and display advertising on Google Ads and Facebook Ads Manager respectively, showing screenshots of the interface and bringing in views from advertisers on the specific choices made and how they approach them. We are interested in this as it helps inform our views on advertiser demand and their behaviour, as well as the choices they must make when purchasing digital advertising. In the ‘Transparency and choice architecture’ section following this we will discuss why some of the options presented as defaults may not always be in the best interests of some advertisers and advertiser views on the extent of the information provided by the interface, as well as specific transparency issues mentioned. N35 Figure N.14: Google Ads selecting a campaign type Source: Screenshot taken 08/04/2020, Google Ads. If a campaign is set up with a goal that tracks conversions, such as sales, the advertiser can now set up a conversion action to modify their website to track these actions.32 Figure N.15: Google Ads setting up a conversion action Source: Screenshot taken 08/04/2020, Google Ads. After this, the initial stage of the campaign creation has been completed, and the advertiser can move onto the Campaign Settings. 32 A conversion action is a specific customer activity which is valuable to the advertiser. When a user interacts with an ad (by either clicking an ad or viewing it), Google Ads stores cookies that contain information about the interaction once the advertiser’s website has been modified to allow Google Ads access to track the action performed by a user on the website, the conversion tracking tag installed reads the cookie and sends it to Google Ads with the conversion information. N36 Campaign settings In the next stage of the interface, the advertiser makes the majority of the choices regarding the campaign settings. Under ‘Networks’, the advertiser can choose to ‘Include Google search partners’ and ‘Include Google Display Network’, both of which are automatically preselected.33 Under ‘More settings’, the advertiser can choose to specify start and end dates for the campaign. The default setting is for the ads to run continuously (ie no end date specified). Audience targeting The next option presented to the advertiser in the interface relates to audience targeting. Audience targeting allows advertisers to only bid for search queries from users within the targeted audience – for example a certain demographic. Almost all the large advertisers we contacted reported using audience targeting to some extent as part of their overall search advertising strategy. In the Google Ads interface, the advertiser can choose which audiences to target in the ‘Targeting and audiences’ section of the second stage of the interface. Options include choosing the location34 of the audience to target, some more specific targeting options35 and exclusion options, as seen in Figure N.16. 33 Search and Display Networks are discussed in more detail in the ‘Defaults and choice architecture’ section below. 34 This has automatically set to the United States in this case. 35 This has been set to include ‘people in, or who show interest in, your target location’. Other options include ‘people who are regularly in the target location’, or ‘who are searching for the target location’. N37 Figure N.16: Google Ads targeting and audiences selection Source: Screenshot taken 08/04/2020, Google Ads. The advertiser can also choose to create an audience under the ‘Audience’ section of the interface, and choose the audience targeting setting. This audience can then be saved and re-used for other campaigns. Figure N.17: Google Ads audience selection Source: Screenshot taken 08/04/2020, Google Ads. N40 and Adobe) to automate bidding optimisation using rules relating to the advertising client’s KPIs. Choosing keywords The advertiser would then move onto the third stage of the campaign set-up process, which allows the advertiser to choose the keywords to bid for. In the Google Ads interface, the third stage of the campaign creation option allows the advertiser to set up an ad group and select the keywords to bid for. The figure below shows the screen presented to the advertiser. Figure N.19: Setting up an ad group and selecting keywords in Google Ads Source: Screenshot taken 08/04/2020, Google Ads. If no alternative option is specified, the match type will be a Broad Match. The advertiser can specify a Phrase Match or Exact Match using quotation marks or square brackets respectively. If the advertiser has linked their related website or entered the product or service, the interface will provide some keyword suggestions, which if added will automatically be placed in a Broad Match format. Once this has been finalised, the third stage of the campaign set-up process has been completed. Most of the advertisers contacted during the qualitative research saw the keyword search as a major benefit of Google, enabling advertisers to target people based on need, as they are targeting ‘in-market’ customers who want to buy their product. Large advertisers and media agencies we spoke to mentioned various dimensions that they consider when deciding on which keywords to bid for. These include: N41 Generic vs targeted: One media agency told us that keywords are selected based on the different roles they play in the marketing funnel, as shown in Figure N.10. Specifically, targeted keywords are most useful at the bottom of the funnel ie to target consumers who are closer to purchase, whereas more generically targeted keywords are often used to drive awareness at the top of the funnel. Some respondents told us that it was preferable to bid on more targeted keywords where possible as the relevance and intent will be higher from the customer for their product or service. Different areas of search intent: One media agency told us that it considered four different areas of search intent, each requiring a bespoke collection of keywords to be created, and each requiring a tailored experience: — Needs-based keywords, eg How do I care for damaged hair? How do I clean red wine from a carpet? — Brand/Product Information based keywords, eg product name or product-specific information and reassurance eg allergy advice, ingredients, environmental info and efficacy. — Passions/Purpose based keywords, eg natural beauty or men’s parental leave. — Purchase, eg by product name or brand name baby gift set. Branded keywords: One advertiser told us that it aimed to maintain number one position on its branded terms regardless of cost, to a point. This strategy is deployed to convert all the demand generated by its ‘above-the-line’ (brand) advertising. Search volume: how large the pool of searches for the keyword is. Cost per click: the historic price of the keyword. Respondents noted various sources used to research potential keywords, including the advertiser’s website and any other marketing material, the media agency’s expertise, Google and Bing’s keyword planners (for information on search volumes and costs) and GAP analysis showing keywords that would be relevant for the advertiser’s business. Several advertisers and media agencies noted that it was crucial that keywords directly reflect the content of the client’s website in order to maximise relevance (a relevant factor in the search auction). Therefore, it is N42 important to have high quality and updated content available on the landing page website. One advertiser noted that it would get cheaper CPCs due to a higher ‘Quality Score’, as it is able to tailor the ad copy to the specific keywords that customers search for. Another advertiser told us that it did not have sufficient content on site to justify significant investment in search advertising. In addition to the keywords themselves, advertisers are also able to choose matching functions, which are available from both Google and Bing: Exact Match means that an ad is shown only when a user types in an exact term in a very specific order (eg ‘black satin formal dress’ would match only when a user types ‘black satin formal dress’ and not ‘formal black satin dress’). This ‘match type’ tends to have the lowest cost per click (CPC) as it provides the most relevant results for the search term, so the advert and experience should be highly relevant for the consumer. Broad Match means that ads would be shown against semantically related searches (eg ‘formal dress’ would also match ‘Little Black Dress’, ‘ball gown’ or ‘cocktail dress’). Google Ads uses Broad Match by default, unless Exact Match or Phrase Match is specified. Phrase Matching, where we specify a phrase and that phrase has to appear anywhere in the search term typed by the consumer (eg ‘formal dress’ would match against ‘black silk strapless formal dress’). One media agency told us that both Broad Match and Phrase Match typically have a higher CPC than Exact Match. This allows the media agency to discover new keywords that consumers are using, which the media agency can then add to their Exact Match list. To protect reputation and avoid budget wastage, the media agency uses negative keywords within Broad Match and Phrase Matched campaigns. A negative keyword is an exact phrase that the media agency does not want to match against, either because the client does not want to be associated with a term or because they want a cheaper and more relevant Exact Match campaign to be shown (in the above example ‘black satin formal dress’ would be a negative in Broad Match and Phrase Match campaigns). Finalising the ad The fourth and final stage of the campaign set-up process allows the advertiser to customise the appearance of their advert, the text to display and website to direct to. A preview is included on the right-hand side of what the ad will look like. N45 Figure N.21: Facebook Ads Manager initial choice Source: Screenshot taken 07/04/2020, Facebook Ads Manager Conversion campaigns will require advertisers to, for example, install Pixels on their website in order to track successful actions. Consideration may be the most commonly chosen campaign objective (as discussed in the ‘Advertiser characteristics’ section) due to the fact that conversion campaigns require further integration with Facebook which some advertisers may not be willing to do, and awareness campaigns may not frequently be chosen by advertisers due to the difficulty in measuring the effectiveness of these advertisements, as it does not involve the users undertaking some observable action such as clicking a link or making a purchase. The advertiser can then manage their campaign objectives and choose to run an A/B split test,42 as well as optimise their campaign budget, as can be seen in the figure below.43 42 An A/B test is a randomised experiment using a two-sample hypothesis test. For example, an advertiser may want to compare the effectiveness of two different advertising strategies such as targeting the advertisement at two distinct audiences. If this option is chosen, Facebook will divide the advertiser’s budget to equally and randomly split exposure between each version of the advertisement’s target audience. The results of the A/B test can then be used the measure the effectiveness of one strategy over another. 43 Neither of these options have been turned on by default. N46 Figure N.22: Facebook Ads Manager campaign objective choices Source: Screenshot taken 07/04/2020, Facebook Ads Manager. Based on our experience of setting up an ad campaign, if the advertiser chooses to turn on ‘Campaign budget optimisation’, the advertiser can specify the Campaign Budget and Campaign Bid Strategy. The default setting for Campaign Budget is Daily Budget of £100. If Lifetime Budget is selected, the default amount is £700. The advertiser can also make some choices regarding the campaign bid strategy, ad schedule and delivery type, as can be seen in the figure below. The default setting for Campaign Bid Strategy is ‘Lowest cost’ (the most conversions within your budget).44 In advanced settings, the ad schedule is set to run all the time. The delivery type is set to ‘standard’.45 44 Other options include ‘Bid cap’ (control bid in each auction) and ‘Target cost’ (aim for a consistent cost per result). 45 The delivery type relates to pacing, namely how often the advertiser’s ads will be shown and how quickly the budget will be used. The alternative option is ‘accelerated’ delivery. N47 Figure N.23: Facebook Ads Manager campaign budget optimisation Source: Screenshot taken 07/04/2020, Facebook Ads Manager. Once this has been completed, the advertiser moves onto the second stage of the campaign creation process, which is the Ad Set configuration. Throughout this stage of the ad set-up process, there is a panel on the right-hand side which displays a dial indicating the audience size in terms of potential reach and gives an indication of whether the audience selection is broad or specific. There is also a panel which shows the estimated daily results in terms of reach and link clicks in this case. During this stage, the advertiser has the option to use Dynamic creative46 or Offer47 for their campaign. Audience targeting The next option presented to the advertiser in the Facebook Ads Manager interface relates to setting up audience targeting. While search advertising is driven primarily by the objective of generating sales (or other actions) from consumers for which there is evidence of intent, display advertising is 46 This can be used to upload assets (like images and headlines) which Facebook uses to automatically generate ‘optimised creative combinations’ for the websites audience, and is turned off by default. 47 This drives more conversion by creating an offer people can save and get reminders about, is also turned off by default. N50 Contextual Spaces, and More Apps and Sites. The preselected places to advertise listed are all affiliated to Facebook.49 Figure N.26: Facebook Ads Manager manual placement options Source: Screenshots taken 07/04/2020, Facebook Ads Manager. Bidding optimisation, budget and schedule The advertiser can then move onto choosing their ad budget and schedule.50 In this section, the advertiser makes three broad decisions, namely they choose what ad set budget to set, whether they want to include a cost control or not, and whether they want to use an automated or manual bidding strategy. 49 Some placements will not be available depending on the objective that the advertiser chooses during the initial set-up process. For example, the Facebook Audience Network (FAN) is only available for certain advertising objectives, such as increasing traffic to the advertiser’s website, which has been chosen in this case. For other objectives, such as driving users to view an advertiser’s Facebook page, FAN is not available as it would not be an optimal way to achieve the objective. 50 As seen previously, Facebook offers the option to set a budget at both the ad set and campaign level, including the ability to set a daily budget or a lifetime budget. N51 Within the interface, the optimisation for ad delivery is set to the suggested option predetermined by prior choices, which is ‘Link clicks’ in this case.51 No cost control has been set. The daily budget has been set to £5 if ads are run continuously, and £20 if an end date (1 month from the present) is selected. The lifetime budget is set to £350 by default. These numbers appear to be default settings and do not change regardless of the options chosen previously. The schedule is automatically set to run continuously starting on the present day. If an end date for the campaign is specified, it is automatically set to a month from the current day, and ads are set to run all the time, as discussed previously. Figure N.27: Facebook Ads Manager budget and schedule options Source: Screenshot taken 07/04/2020, Facebook Ads Manager. Under these settings, the options for how the advertiser is charged are ‘Impressions’ and ‘Link click (CPC)’, of which ‘Impressions’ has been selected by default with these options. Under ‘more settings’, the advertiser can view a graph of the estimated reach and results of the chosen settings. The ad delivery type, which is Facebook’s ‘pacing’ system, has been set to ‘standard’ by default.52 The campaign bid strategy has been set to ‘Lowest cost’ by default which is the ‘auto bid’ strategy if no cost control is set. If an advertiser chooses to 51 Other options include ‘Landing page views’, ‘Impressions’, and ‘Daily unique reach’. 52 The alternative pacing method is ‘accelerated’. The default standard setting will try to maximise the value an advertiser receives from each ad by bidding the minimum amount that will win enough auctions to spend all of the advertiser’s budget over the duration of its ad campaign. N52 specify a cost control strategy, they can choose from ‘Bid cap’, ‘Target cost’, ‘Cost cap’, ‘Minimum return on ad spend (ROAS)’ and ‘Highest value’. Facebook have submitted that [90-100]% of UK advertisers in 2019 used the auto bidding feature (ie no cost control). For further discussion of Facebook auction mechanisms and bid prices, see Appendix Q. Figure N.28: Facebook Ads budget and schedule options continued Source: Screenshot taken 07/04/2020, Facebook Ads Manager. Once this has been completed, the advertiser has finished this stage of the ad set-up process. Finalising the ad The advertiser can now move on to the final stage, which involves creating the ad itself. In ‘Identity’, the Facebook page or Instagram account to be advertised can be linked to the advertiser’s business. ‘Format’ allows customisation of the format of the ad. ‘Single image or video’ has been selected by default.53 ‘Media’ allows the advertiser to upload and set up their assets. ‘Text & Links’ allows the advertiser to add links, text, headlines and more. Once this has been set up, the campaign is ready to be run. 53 The other options are ‘Carousel’ and ‘Collection’. N55 If the ‘Include Google search partners’ check box is unselected, text below the check box states ‘Most advertisers include their ads on Google search partner site’. If the ‘Include Google Display Network’ check box is unselected, text below the check box states ‘Don’t miss the opportunity to reach more people across 3 million sites and apps’. Given this text, it seems that Google may be nudging advertisers into keeping these options selected. By opting users into the Search and Display networks as well as the main search platform, Google may be trying to leverage its strong market position from search into display. In particular, Google may be able to increase the ad spend of smaller advertisers who are more likely to directly use the Google Ads interface. Figure N.29: Google Ads networks default settings Source: Screenshot taken 08/04/2020, Google Ads. Google has submitted that it has presented the Search Network as a preselected check box to reflect the fact that the vast majority of advertisers choose to keep this option selected, in order to simplify the set-up process. According to Google data on UK advertisers in 2019, [40-50]% of ad campaigns by revenue include the Search Network. If the vast majority of advertisers keep the Search Network option selected but only [40-50]% of ad campaigns by revenue include the Search Network, this could indicate that large advertisers who spend more, are more knowledgeable about digital advertising and are more sophisticated in its usage, are more likely to unselect this option. Smaller advertisers are likely to be less sophisticated, so by presenting this option as selected by default, this could be consistent with Google exploiting behavioural biases to influence advertisers to keep this option selected even when not in the best interests of the advertiser. It is also unclear how presenting this option as preselected rather than unselected simplifies the set-up process for advertisers. We also note the circularity of this argument, as it is possible the vast majority of advertisers keep this selected because it’s a default and it may not always be beneficial for the advertisers who keep it selected. It may be the case that it would be more efficient for the advertiser to have their ads only appear on the Google search page. However, we note that large advertisers may have N56 specific reasons for unselecting this box, and similarly small advertisers may have reasons for keeping it selected. If the advertiser clicks ‘Learn more’ for Display Network, a panel on the right explains towards the end that if a Search Network campaign is expanded to Display, this will increase conversions and thus daily spend by an average of 15%. Placing this relevant information at the end of the information panel may prompt advertisers to leave the box ticked. Google has submitted that it has presented the Display Network as a preselected option to simplify set-up for (typically new) advertisers. We note that there is an inconsistency between Google’s stated rationale for why it has preselected this option compared to the Search Networks option, which it presented as preselected as the vast majority of users kept the Search Networks check box selected. As with the Search Networks option, it is also unclear how presenting this option as preselected rather than unselected simplifies the set-up process for advertisers. Google has also submitted that [80-90]% of advertisers uncheck this option during set-up, making it unclear why this option is preselected if the vast majority of users unselect it. Given the fact that [80-90]% of advertisers unselect this option, this indicates that this default is not in the best interests of most advertisers. Those advertisers who choose to keep the Display Network selected are potentially being disadvantaged, though we note some advertisers may have genuine reasons to use this option. Most advertisers contacted during the qualitative research rejected the inclusion of Google’s Search Network and the Display Network in particular, as most claimed that they wanted to advertise only on Google as this was seen as the best brand. Another setting that has a similar effect to a default relates to the keyword selection, where the default keyword match type is a Broad Match unless something else is specified. If a website is linked and the advertiser chooses to select one of the keyword suggestions, this will copy the keyword as a Broad Match by default. Some advertisers said that the default matching option for keywords was not the best for them, eg having generic names in a competitive sector. Some advertisers indicated that Broad Match will typically have a higher CPC than Exact Match, for example. This is consistent with our own analysis, as discussed in Appendix Q. Another default option is under ‘Budget’, where Google Ads allows advertisers to choose a bid strategy. As discussed above, there are three decisions the advertiser is making at this stage, namely the choice of what budget to set, N57 whether they want to set a maximum CPC bid limit or not, and whether they want to use an automated or manual bidding strategy. Under the choices chosen during the interface walkthrough, the default choice is to focus on ‘Clicks’ and use the related automated bidding strategy ‘Maximise clicks’, as shown in Figure N.18. Google recommends that advertisers do not select a bid strategy directly. This may lead to the advertiser not clicking this option, particularly if they have not realised that there is a range of automated and manual strategies they can choose from.55 Consequently, this may lead advertisers to miss out on better bidding strategies available to them. In particular, less confident advertisers who are unsure about which option is best may simply accept the default automated bidding strategy. As shown in Figure N.18, there is no maximum CPC bid limit specified by default. Instead, the default is for advertisers to allow Google to set their bids through one of its automated bidding strategies. It is possible that advertisers who do not specify a maximum CPC limit may in certain circumstances pay for advertising where the cost is higher than the amount they are prepared to pay and may in fact lead to negative ROI, and may be better off by specifying a maximum CPC bid limit. Google has submitted that for UK advertisers in 2019, [50-60]% of ad campaigns by revenue set a maximum CPC bid limit, and [40-50]% used automated bidding strategies. This submission implies that while it is possible to use an automated bidding strategy and set a maximum CPC bid limit, in practice this is not done by advertisers. When Manual CPC is selected, the option ‘Help increase conversion with Enhanced CPC’ is pre-ticked. Google have submitted that this feature typically helps advertisers optimise their campaign goals by achieving a higher number of conversions, and reflects the choice that Google observes that most advertisers make since their goal is to maximise conversions. Default settings on Facebook Ads Manager As with Google, throughout the Facebook Ads interface, there are several default options presented to the advertiser, and the view of the advertisers contacted during the qualitative research was mixed on defaults. Based on our experience of setting up an ad campaign, the first default presented to the advertiser is under ‘Campaign Budget Optimisation’, as 55 If the advertiser chooses to click on the ‘?’ icon and chooses to learn more, the help panel on the right-hand side provides information on which bid strategy may be best to the advertiser considering their goals, and discusses some of the automated bid strategies and the manual one. The list of automated bid strategies include ‘Target CPA’, ‘Target ROAS’, ‘Maximise Clicks’, ‘Maximise Conversions’, ‘Maximise Conversion Value’ and ‘Target Impression Share’. The manual bid strategy is ‘Manual CPC’. N60 appropriate options for their campaigns.59 Overall, we believe there is a balance to be struck between ease of use and transparency. It is unclear if Google and Facebook currently have sufficient incentive to improve the transparency of their interface considering their market power, though we also note that it is unclear how feasible it would be for Google and Facebook to provide further information within the interfaces, and what information this should be. This section sets out some advertiser views on concerns about transparency issues related to the information provided within Google and Facebook’s advertising interfaces, and the specific areas advertisers have highlighted that they would like further guidance on. Google Ads Most advertisers contacted during the qualitative research indicated that they were generally happy with the interface at a high level, and that it was relatively easy to use. However, several advertisers indicated that they found the Google interface to be more confusing or complicated to use compared to the Facebook interface. Most advertisers criticised the time and money they had to spend to learn how to advertise effectively, particularly in regard to search terms and how to use them. Some advertisers said that while it may be easy to tweak a campaign, they were sometimes unsure about what they could do to improve it, with some advertisers potentially arbitrarily changing options to see what works best. There was a general perception that advertisers had to use ‘trial and error’ to learn how Google worked best. Some advertisers said that they had to use articles and online videos to learn how to effectively use the interface, indicating that they felt that Google didn’t provide sufficient information on these options. Several advertisers said that they wanted Google to do more to help at certain stages of the ad buying process, for example by providing basic advice on what does and does not work when selecting audiences or providing more guidance on budgeting. For example, some advertisers contacted during the qualitative research said that they only had an approximate idea of the level of budget they would need to meet their objectives and advertise effectively. There was a general perception from qualitative survey respondents that Google’s ‘advice’ is primarily focussed on getting people to spend money on 59 On the other hand, we note that it is also possible that additional information may have the opposite effect, and may potentially over-load the advertiser with information which may reduce interaction or demand. N61 campaigns, rather than optimising the effectiveness of any campaign. Some advertisers also believed that they could be advertising more effectively, but didn’t know how to go about it, and had not received any advice or guidance from Google on this, with a general perception that Google wasn’t doing enough to help advertisers. In our analysis of Google’s advertiser interface outlined above, we found that Google does provide some information about the options available to advertisers, and how they differ. For example, throughout the interface, there are information icons at certain choice stages which provide more information about a specific option. There is sometimes a ‘Learn more’ option which, if clicked, opens a side panel on the right-hand side of the interface with additional information on the option, with occasional examples and links to direct the advertiser to where they can find further information on the option to help inform their decision. However, we note that this information does not discuss the possible effects of choosing one option over another on the advertiser’s budget or the eventual effectiveness of the advert measured against the advertiser’s KPIs. We also note that advertisers have indicated that there are other sources of publicly available information that can be used to inform their judgement. Facebook Ads Manager Most advertisers contacted during the qualitative research indicated that they were generally happy with the interface at a high level, and that it was relatively easy to use. As with Google, advertisers were initially attracted to Facebook as it was seen as relatively easy to get started, but then criticised the amount of time they had to spend learning how to effectively advertise on it, having to use ‘trial and error’ to learn how it worked best. Most smaller advertisers said they were likely to underspend initially while ‘testing the waters’, though this money may be wasted as it does not reach enough people to generate a decent or identifiable ROI. Advertisers using Facebook raised similar concerns on the ability to access relevant information as for Google. In our analysis of Facebook’s advertiser interface, we found that throughout the interface, there are information icons which provide further information about the given option, as seen with Google. Occasionally clicking ‘Learn more’ opens a new tab with further information on the choice the advertiser is asked to make. During the Ad Set set-up, there is a panel on the right with an indication of the estimated audience size and estimated daily results the advertiser can expect, N62 which change to reflect the options chosen by the advertiser, as can be seen in Figure N.30. Figure N.30: Facebook Ads Manager audience size and estimated daily results panel Source: Screenshot taken 07/04/2020, Facebook Ads Manager. This information allows the advertiser access to some information that will potentially be very relevant to them on the effect of various choices on audience size, expected daily reach and link clicks. The panel on the right-hand side displays a dial indicating the audience size in terms of potential reach, which allows the advertiser to assess if the audience selection is broad or specific. The dial is green if it is somewhere between specific and broad, red if it is specific, and yellow if it is broad. This colour- coding indicates to the advertiser that Facebook considers too narrow or too broad audience sizes to not be optimal, with too specific potentially being worse than too broad.60 This may nudge advertisers into targeting certain audiences, though there is no discuss of this in the interface, making it more difficult for the advertiser to properly assess their appropriate audience size. 60 Similarly, during the ‘Audience’ section, if the advertiser chooses to specify the location, a text box on the screen states: ‘Keep your audience broad. A large potential audience lets Facebook look for the lowest cost result. Defining a very specific audience can make your campaign less effective. The best opportunity to connect with people may be outside the audience you expect.’. N65 One large advertiser said that there was limited transparency from Google on why costs had risen, and whether this inflation was driven by Google (based on adjustments to its pricing model) or whether it was due to competitor bidding activity. This advertiser claimed that the system has been designed such that Google can use its market strength to exploit and monetise the wider advertising efforts of those who have little option but to invest in Google’s branded search advertising proposition in order to stay visible in their own highly competitive segments. Similarly, some large advertisers said that there has been a consistent CPC inflation year over year with Google. Some respondents from the qualitative research had begun to question their ability to take their business elsewhere if costs for Google continued to increase at the current rate, however there was a tacit admission amongst these advertisers that they had to be on Google, as there was no other search engine with the same reach and the ability to target specific groups. Control over budgets Some advertisers contacted during the qualitative research thought that Google will take the decision to go over-budget if it thinks it is making sales, and that they did not have sufficient control over budgeting within Google. Advertisers were not clear on why this happens or why they are being charged more if they didn’t sign off on the increase. Some claimed that they believed the defaults within the interface were set to do this and believed this to be a deliberate aspect of Google’s business model and did not like the lack of control with regards to their budget. We note that these views are opinions and we have not verified these claims. Some of the smaller advertisers who were less confident and less tech savvy may have kept the default settings, including the relevant automated bidding strategy. Considering that several advertisers complained about Google overspending their budget, this may have been linked to the default bid strategy which the advertiser agreed to without fully understanding. As with Google, some advertisers claimed that Facebook had gone over their budget and did not like the perceived lack of control over this, though this was more pronounced for Google than Facebook. As shown in ‘Display advertising using Facebook Ads Manager’ section, the interface allows the advertiser to set daily and lifetime budgets across both the ad set and campaign level within the Facebook Ads Manager interface. However, Facebook has submitted that there are circumstances where this daily budget can be flexed to maximise results, although the total amount will average out over the course of seven days to ensure that the lifetime budget is not exceeded. N66 Conclusions Based on Google and Facebook’s data, there is a minority of very large customers who account for a very high proportion of spend on Google and Facebook, and there are many smaller advertisers who spend comparatively little on advertising and do not account for a significant proportion of spend. Advertisers and media agencies have told us that there are limited effective substitutes for search and display advertising, as advertisers have largely told us that offline resources are not a good substitute, and search and display advertising largely perform a different role in the customer purchase journey. We note there are other alternatives to Google in search advertising such as Bing and other smaller search engines, and alternative platforms and offline advertising for Facebook in display advertising. However, given the high degree of market power Google and Facebook have, it is important to ensure that competition in search and display works effectively. Advertisers choosing to run a search or display campaign must make several choices regarding their campaign objectives, the audience they wish to target, where they wish to advertise, the bid strategy they want to employ, amongst others. We have considered the options presented to advertisers during the search and display campaign set-up process with Google Ads and Facebook Ads Manager respectively, bringing in views from advertisers and media agencies on how they approach this process. Advertisers have told us that they have some concerns about transparency issues with both Google and Facebook. We have also considered the choice architecture of Google Ads and Facebook Ads Manager and some default options within these, as both interfaces have some default settings which may have the potential to influence less experienced, typically smaller, advertisers into selecting certain options. There is a risk that these options may not be in some advertisers’ best interests. We have looked at the information provided via Google and Facebook’s advertising interfaces. We have found that while Google and Facebook do provide some relevant information to advertisers on the choices they must make, some advertisers felt there was not enough relevant information provided to them, with many advertisers having to use trial and error to learn how to advertise effectively. We also set out some specific concerns advertisers have raised about transparency over the auction process, costs and control over budgets. For Google in particular, we note that advertisers expressed concerns about persistent cost inflation and a perceived lack of control over their budgets. N67 A number of these concerns, particularly relating to choice architecture and the use of defaults, could be investigated under the ‘trust and transparency’ objective of the code of conduct which is discussed in more detail in Appendix U. N70 Facebook data We have collected data on all UK based advertisers who advertised exclusively to UK based users through Facebook in 2019, including data on advertisements on Facebook subsidiaries purchased through the Facebook advertising interface.67 The data records spend (in USD) and the number of campaigns run. Both variables can be broken down into several distinctions, including the campaign objective (ie awareness, consideration, conversion) and the format of the advert (ie video and non-video). Facebook also provided self-reported industry data. We must exercise some caution in interpreting how similar advertisers in the same industry may be, however there appears to be the right level of aggregation of the categories68 to allow for relatively reliable breakdown of key variables by industry type. This dataset only allows for a limited depth of analysis because we do not hold detailed information on the auction environment, or the level of targeting done by Facebook. Caveats For a number of accounts69 the total spend does not equal the sum of the campaign objectives columns (Awareness, Consideration, Conversion and Other). Similarly, there were instances70 where the total spend did not equal the sum of the ad formats (video and non-video). Facebook submitted that a small number of discrepancies like this are normal at the individual account level. These problems can occur for various reasons related to data logging or the specific nature of certain ads. For example, some ads are shown in both video and non-video form (eg an ad could appear as a video on Facebook, but in a non-video format as part of an Instagram story). In such cases, adding together figures across video and non-video would give a greater figure compared to the total column. Facebook submitted data in US dollars, however we have performed an exchange rate conversion to the data to present results in GBP.71 This could 67 Such as Instagram. 68 The only notable exception to this is that ‘government’, ‘politics’ and ‘government and politics’ are three separate albeit minor categories in the data. The data provided has been re-coded such that all observations with these industries listed have been changed to ‘government and politics’. 69 Less than [0-5]% of the total, after allowing for some discrepancies due to rounding. 70 Less than [0-5]% of the total, after allowing for some discrepancies due to rounding. 71 The same conversion has been done for Facebook as with Google. N71 cause minor differences from the observation’s recorded spend and actual spend. Large advertiser and media agency views As part of our investigation into the behaviour and characteristics of advertisers, we contacted 40 of the largest advertisers by spend on digital advertising, and five major media agencies. We received views from 23 large advertisers and four media agencies and summarised these responses. We have collected views from a range of corporations, retailers, chains, financial services and institutions, supermarket chains, as well as other assorted companies. Considering the concentration of the advertiser base on Google and Facebook, with a minority of large advertisers representing the majority of spend on digital advertising,72 the large advertisers contacted reflect a large proportion of the customer demand for Google and Facebook. For both Google and Facebook, the top [5-10]% of advertisers by spend73 account for more than 85% of the total spend on Google and Facebook advertising in 2019, as detailed in the ‘Advertiser characteristics’ section. Many of the large advertisers we have collected views from use a media agency to purchase digital advertising. Media agencies are independent brokerages hired by advertisers for a range of services, including advising on and purchasing digital advertising.74 We have collected views from four major media agency groups (in text, these are referred to as ‘media agencies’). We have also collected views from large advertisers who choose to purchase digital advertising using an in-house buying team rather than an external media agency. Of the 22 large advertisers who indicated to us whether they use a media agency or not, seven stated that they perform some, if not all, digital advertising purchasing in-house. We have collected views from large advertisers as these advertisers make up the majority of the revenue received by both Google and Facebook from advertisers. For smaller advertisers, who generally make up a very small proportion of total spend, some qualitative research has been undertaken instead to find their views. 72 For more discussion on this, please see the ‘Advertiser characteristics’ section. 73 Where advertisers have been sorted by spend on either Google or Facebook advertising in 2019 from smallest to largest. 74 In addition to advertising, media agencies may also advise the advertiser on public relations and other relevant forms of media management. N72 Qualitative research As part of our investigation into the behaviour and characteristics of advertisers, we commissioned an independent research agency, Jigsaw, to collect views from a range of advertisers and summarised these responses. For further detail, please see the Qualitative Research Report.75 The overall objective of the qualitative research was to capture information from advertising customers of Google and Facebook to inform the assessment of the nature of competition and consumer behaviour in the digital advertising market in the UK, in particular smaller advertisers, as these advertisers make up the majority of the advertisers on Google and Facebook. The qualitative research involved telephone-based interviews with 24 advertisers, lasting a duration of 45-60 minutes. The advertisers chosen were supplied by Google and Facebook and supplemented by free-found respondents due to the quality of the supplied sample of advertisers. This fieldwork was conducted in March and April of 2020.76 Of the 24 advertisers contacted, 14 used Google and 19 used Facebook, meaning nine advertisers used both Google and Facebook. The advertisers contacted ranged from small advertisers whose companies employ less than 10 people (ie micro advertisers), to larger advertisers with over 250 employees, with most being either micros or smaller SMEs with either under 10 or 10 to 49 employees. Advertisers with between 10 and 249 employees follow the general characteristics of small and medium sized enterprises (SMEs) and do not have large budgets for advertising, and the level of technological savviness, sophistication and confidence varies across the advertisers. Two of the contacted advertisers were large companies with over 250 employees, who may be more sophisticated in their digital advertising usage.77 The figure below illustrates the spread of spend by the advertisers contacted who advertise on Google and Facebook, showing that the majority of the advertisers contacted who used Google spent between £1,001 and £100,000, while the majority of advertisers who used Facebook was between £50 and £100,000.78 We note that this graph should not be considered representative of the overall advertiser base. 75 The report is published under the CMA Online Platforms and Digital Advertising Market Study case page. 76 As this fieldwork was conducted during the COVID-19 lockdown, the efficacy of the supplied sample of customers from Google and Facebook may have been affected. 77 Of the 24 advertisers, three were media agencies with all other advertisers belonging to other sectors, and two advertisers used third parties to manage their advertising. 78 Quotas were set on the qualitative sample to ensure a spread of advertisers, as the choice of advertisers was not intended to be representative, but to rather cover different perspectives.