The deprecation of third-party cookies affects all aspects of programmatic ad buying as we have known  it – personalisation, audience targeting, tracking and measurement, frequency capping, and more. As such,  the advertising world has had to rapidly adapt. Whilst some have developed alternative IDs, two post-cookie  alternatives have emerged as the most popular amongst much of the industry – first-party data-based and  contextual targeting.  

Fifty Technology has produced this Deep Dive in partnership with ExchangeWire. Fifty uses AI, technology and data to help brands understand their customers and power advertising to best engage them. At our core sits the Fifty platform, which makes sense of complex audience datasets and develops them into tribes, an impactful and insightful customer segmentation interface with applications across all marketing functions. FiftyAurora is our brand new, ID-free, privacy-first solution that provides effective audience targeting by projecting a map of where an audience will be, rather than needing to track them everywhere.

First-party data has become more valuable than ever, with some in the industry dubbing it “the new oil”, and  it’s easy to see why – it’s collected directly from a webpage in real-time, meaning that it can deliver far more  accurate insights than those provided by second- or third-party data. And since it’s collected by the site owner  based on the consent preferences of each user, first-party data solutions offer comparatively more privacy than  the third-party cookie did. 

However, first-party data is not a perfect cure-all. At present, relatively few companies have access to enough  first-party data to make effective use of a first-party data solution, meaning that those who benefit most from  its rising value are big publishers, brands, and walled gardens, with smaller and more recently-established  players missing out. Then there’s the uncertainty of just how privacy-centric first-party data solutions really  are – questions remain over the efficacy of cookie consent banners (or pop-ups) in gaining meaningful consent  from website users, with many sites accused of using ‘dark patterns’ to mislead users into handing over more  information than they really want to. And whilst first-party data promises more accurate insights, this accuracy  is dependent on users being addressable – that is, remaining logged-in to a website, a condition which is seldom met at any scale today. 

Contextual, meanwhile, has seen a resurgence since the demise of the cookie was announced, with some touting  this period as the “rebirth” of the targeting method. Recent research found that 52% of UK and US marketers  plan to up their investment in contextual targeting over the next two years and 86% of media owners expect  the method to become more prevalent. The merits of contextual precede the death of the cookie – it has long  been known that contextual works across all formats and channels (unlike cookies, which were exclusively  digital browser-based), can outperform cookies across CTRs, VTRs, CPL, CPC, and viewability benchmarks, and  offers built-in brand safety. Its immense scalability is a key benefit, but what really puts contextual ahead of  competing post-cookie offerings is that it is inherently privacy-preserving, relying on identifying the content  on a page rather than the user behind the screen. With regulation only expected to increase, this is a vital  distinction that has, and still will, draw the attention and investment of advertisers. And because it doesn’t rely  on any information about the user, contextual doesn’t require swathes of first-party data or user consent to  work, making it a viable option for brands and publishers with little or no first-party data (as well as a parallel  tactic alongside first-party targeting for larger agencies, brands, and networks). 

Considering its scalability, universality, and the fact that it has privacy at its core, it is of little surprise that  contextual has become the first-choice post-cookie solution for many in our industry. However, despite its many  benefits, it would be remiss to say that traditional contextual targeting is perfect. Because it is designed to deliver  ads based on an analysis of the content of a webpage, the solution depends far too heavily on assumptions about  the users visiting. As a result, contextual targeting is naturally less precise than first-party and other solutions.  

Neither solution is perfect: first-party data lacks scale and treads a fine line with privacy, whilst contextual is  short of precision. Yet, despite their individual shortcomings, first-party data is still incredibly valuable, and  contextual still holds a lot of promise for the post-cookie landscape. We believe that combining the two where  possible (or using an effective proxy for first-party data) can enhance both and help brands and publishers reach  more potential customers with greater accuracy by addressing the following challenges. 

UNDERSTANDING THE AUDIENCE

In order to effectively reach prospective customers, advertisers have to be able to understand who they’re  actually looking for. Traditional contextual falls short on providing audience understanding for advertisers  because it is predicated on assumptions about a consumer’s interests and preferences based on the content of the  page they’re visiting rather than on measurable insights.  

Let’s take the example of a person visiting the website of a film publisher and being presented with a banner ad for  a company that offers screenwriting workshops. Whilst this sounds like a logical pairing — it’s not unreasonable  to assume that many budding screenwriters are interested in reading up on the current film scene — this process  does not guarantee that the company will reach their target audience. This is because there are a number of  possible reasons why this reader has decided to visit this page: it could be that they are a casual cinema-goer who  just wants to browse reviews before deciding what to watch; or perhaps they are interested in filmmaking, but  from a visual or sound design perspective, and aren’t as concerned about the ins-and-outs of a script. 

By assuming that everyone who visits a particular website does so because they share the same interest (i.e.  reading a film publication because they’re interested in screenwriting), traditional contextual essentially targets  based on a one-size-fits all approach. This totally disregards the divergences and nuances that create the various  subsets that exist within a wider audience group, resulting both in wasted impressions for the brand whose ad is  displayed and missed opportunities to put more suitable companies in front of their potential customers.  

This problem can be resolved by incorporating more audience insights into traditional contextual targeting. Doing so requires a greater quantity of more granular data, but advertisers shouldn’t be put off if they don’t have  access to first-party data, as social data (information taken from the public domain, not private accounts) can  be used to the same effect. By analysing more consumer information, advertisers will be able to identify and  define audience subsets, enabling them to tailor separate messaging so that they can build different campaigns  specifically for different groups. By understanding that their audiences consist of clusters of interconnected but  nonetheless distinct cohorts, advertisers will be able to deliver bespoke campaign messaging that drives higher  engagement on the right pages.

Looking beyond the initial target audience

At present, the broad nature of many traditional contextual targeting solutions serves to entrench the perception  that an audience is a homogenous group united by one or two identical interests. As we’ve explained above, this  impedes advertisers’ ability to capture the attention of prospective customers through personalised campaigns. But it also prevents them from exploring further to find people who may be interested in a brand but who don’t fit  within the criteria assigned to the target audience. 

Let’s take our example of a target audience for a screenwriting workshop a step further. By examining the  affinities and interactions of the wider pool of web users who display an interest in film, the company will be able  to narrow down who would likely be most interested in the workshop. But they will also come across other web  users who share other, related interests with the target audience. In this case, the initial target users may also  display an interest in things like theatre, musicals, or literature by following or interacting with the social pages of  entities within these spheres (such as those of a production company, a publishing house, or an actor, director, or  novelist). As these people have interests and affinities in common with the initial target audience, they may also be  interested in screenwriting, and thus could be ideal prospects for the workshop operator.  

By incorporating social data, advertisers can refine their contextual strategy to reach beyond the target audience  to potential customers who they would not otherwise have been aware of. Brands will be able to identify audiences  that sit outside their standard consumer base and truly prospect. 

Reaching the audience in a privacy-compliant way

Being able to tell where your desired audience will be found online is a fundamental need for brands and  publishers, and the fear of losing this capability is part of the reason why some in the industry are reluctant to give  up third-party cookies and alternative IDs. Traditional contextual predicts where part of an audience will be, but as  we’ve established above, these predictions are based on broad assumptions, and too often result in targeting that is  imprecise and sub-optimal. 

The industry can enhance traditional contextual solutions by combining the natural language processing (NLP),  semantic analysis, and other techniques many of them already employ with the deeper insights garnered by  gathering and analysing audience data. Doing so will equip the solutions to not just assess the details of a webpage,  but to determine whether the right web users are actually visiting it, and therefore where an ad will be most  effectively placed. And all of this can be done in a completely privacy-compliant way, requiring neither a cookie nor  any other kind of ID to work. 

With the addition of audience insight, traditional contextual can become more effective at putting the right ad in  front of the right user without compromising on the vital privacy-preserving quality that makes contextual so  important as we approach the post-cookie era.

KEY LONG-TERM TRENDS

The deprecation of third-party cookies and implementation of landmark legislation, such as Europe’s GDPR, the  California Consumer Privacy Act (CCPA), and China’s Personal Information Protection Law (PIPL), have already  forced significant changes to the mechanics of online advertising. With cookies set to go for good once Chrome  stops supporting them at the end of next year, and more regulation on the way, further shifts are inevitable.  Here, we give our perspective on three key long-term trends as the industry prepares for targeting in a post cookie landscape. 

A false dawn: Alternative IDs are on borrowed time

Perhaps to insulate themselves as much as possible from the fallout of the cookiepocalypse, some in the industry have developed their own IDs to take the place of the tracking technology and continue targeting in a similar way. 

Whilst these may fill the gap in the short term, the move is ultimately misguided, because  IDs are destined for the same fate as the third-party cookie: it is only a matter of time until  regulators from across the globe clamp down on them. This is something which Google  clearly realised when they announced that Chrome will not support universal IDs after the  cookie deadline, saying they “aren’t a sustainable long-term investment” in the face of  tightening regulation.  

Moreover, alternative IDs can be completely undermined by technology providers – last year,  Apple (who have made privacy a key brand pillar) introduced Hide My Email, a feature which  allows users to generate fake email addresses (linked back to the original) to prevent the  website owners and advertisers they give them to from being able to track them. Although  not widely adopted (yet), the feature is a clear example of Big Tech’s ability to render ID based targeting redundant, and it’s very possible that we’ll see similar initiatives in the  future. As Fifty’s Head of Productisation Alex Hawkesworth has pointed out, technology  providers have the power to fight back against the advertising industry’s failure to respect  the privacy of users in the pursuit of profit.

Return to sender: Authentication levels will be slow to increase

We’re already seeing more publishers implement logins. Some worry that this  trend will pave the way to the abandonment of advertising in favour of a paywall  or subscription service, whilst others tout logins as the foundation upon which the  advertising ecosystem of the future will be built.  

The fact is, by 2020, only a median of 2% of web traffic was authenticated. Even in a world  where the majority of websites have implemented some form of login, the likelihood of  people choosing to log in or stay logged in to every website they visit is low. Those who can  enjoy the benefits of logins are currently limited to larger publishers and ‘walled gardens’  like GAMA (Google, Apple, Meta, and Amazon; formerly GAFA) in particular. Whilst this is  both the product and perpetuation of a power imbalance within digital advertising, it is  also a result of there being a recognisable value to users being logged into these companies’  properties, a value exchange which very few other sites can claim to offer. 

The failure to convey that personalised advertising can benefit web users (evidenced by the finding that 63% of consumers believe that sharing their data creates more value  for marketers than it does for them) continues to dog the industry; this, combined with  consumer distrust, has fuelled high opt-out levels and created a landscape where people are  reluctant to share information. Unless publishers and brands successfully communicate the  value that creating an account will provide, efforts to authenticate users via logins will be  unlikely to bear fruit.  

Creativity, context, and commerce: back to the future

Decades of targeted advertising using the third-party cookie has led to a dearth of  creative planning across web and mobile compared to the brand-led approach used for TV and OOH.  

However, the shift to contextually-powered advertising will force change here, fuelled  further by the development and implementation of tools that can design and direct creative to the people it will resonate with most. FiftyAurora, for instance, combines audience insights  with the technology used in traditional contextual to enable marketers to target at scale to reach their desired audiences with greater precision and uncover subsets of their consumer  bases that require bespoke messaging. 

Moreover, the new privacy-focused age will give renewed attention to the full sales funnel, rather than hounding consumers already interested in a given item, or (worse still) who have already left the sales funnel. ID-free technology solutions able to leverage semantic  analysis and customer CRM data, such as FiftyAurora, alongside timely media interaction, are better equipped for this new age than simple cookie mimics in terms of scale and range of channels. 

Gone will be the days of the bland ad unit following you around the web, earnestly extolling the virtues of a product you have already purchased. The paint-by-numbers approach will be replaced by a complex masterpiece, and creative advertising powered by rich contextual data will lead to a redefined relationship between brand and user. We will see an evolution from personalised advertising to humanised advertising. 

FIFTYAURORA: ENHANCED CONTEXTUAL TARGETING FOR THE POST-COOKIE AGE

As we’ve established, we believe that combining traditional contextual targeting methods with first-party data  (or an effective substitute like declared social data) can enhance the power of both and create a strong solution for the cookieless era. Here, Fifty Head of Productisation Alex Hawkesworth elaborates on how FiftyAurora uses audience insights to power next-level contextual targeting in a completely privacy-compliant manner.  

How does Aurora resolve the current issues with traditional contextual targeting?

Current contextual targeting focuses only on the content of a webpage or media item, which can be a limited proxy  for understanding the audience you hope to reach. And understanding audiences is the key of any advertising –  something we have lost sight of with constant tracking of consumers. FiftyAurora overcomes this downside as it  is powered by our human insight platform, allowing Fifty to deeply understand the audience and its subgroups as  well as analyse their content preferences, interests, and other real world signals like location or gender.  

By understanding an audience first we can use this analysis to model the relevant topics, categories and webpages  that a campaign should be targeting for a specific audience, rather than being limited by an unclassified keyword list. 

Our deep human understanding allows FiftyAurora to extend desired audience targeting beyond the central topic  or interest of the campaign to include adjacent or unlinked topics that also engage the audience, thereby increasing  audience reach without diluting audience relevance.  

Another disadvantage of traditional contextual solutions is that they rely upon simple signals like the presence  of a keyword match regardless of the overarching context of an article. This means that there is always a level  of wastage – pages that include the relevant keywords but are not likely to be read by the audience. In contrast,  FiftyAurora is able to use complex topical classification of audience media consumption patterns to exclude any  inventory not relevant or read by a target audience. 

Overall, FiftyAurora provides brands and publishers with an evolution of Contextual audience targeting – a solution  that uses essential audience insights to improve performance and scale while avoiding wasted spend. 

What kinds of data does Aurora use, and how does Fifty ensure the safety of this information?

FiftyAurora audiences contain no personal data or IDs. We use large-scale aggregated web and social data to build  our human insights and audience understanding, and take significant steps to ensure that any intelligence we  derive from this is fully separated or air gapped from the targeting that we provide to clients. FiftyAurora matches  the topic relevance of a page against the topics derived from insight aggregated tribe data rather than ever directly  targeting the members of that insight audience.  

In order to help brands to more effectively use their first party data outside of ID limited environments while maintaining compliance and protection, we have developed the Secure Data Exchange (SDX). SDX incorporates the highest levels of digital security and completely removes any concerns about data leakage or inference as it models CRM into the Fifty insight platform without preserving the original seed. 

What is Fifty’s SDX, and how does it work alongside Aurora to help advertisers and publishers  adjust to a privacy-first world?

SDX is a Strongroom technology (building further on the data privacy of clean rooms like Infosum or Snowflake) and Fifty strongrooms are virtual machines whose processes and data are not accessible to anyone (including Fifty)  outside of them. They maintain data privacy and security by employing strong encryption techniques at every  stage of the process and remain totally isolated through confidential computing. 

So the SDX instance is isolated (in fact, in the case of any changes or updates, Fifty have to create an entirely new  virtual machine) but programmed to only receive files in an encrypted format, process them, and extrude a single  output – this means it is able to securely receive and handle hashed CRM.  

SDX matches the hashed emails into the Fifty data set and processes it to produce a Fifty study derived from the  customer data and its taxonomy in isolation. The output from SDX is unique in that the seed (the CRM and the  matched data) is destroyed through a process called crypto shredding (and so is not restorable by any means) before the strongroom exports the remaining study. This ensures that we provide both complete security of client data  and the deep insight and human understanding that our clients have come to expect. 

This methodology preserves the insights aligned with the original taxonomy of the data and allows the client to safely  translate their consumer records into new learnings and a completely scalable non-ID dependent prospecting tool. 

The combination of FiftyAurora and SDX enables Fifty to provide value to all advertisers and publishers in the  privacy-centric era – Aurora allows us to act as the proxy for consumer data where companies lack it. And those  that have strong first-party data can rely on Fifty to supercharge it.