The Cookie-Free World and Its Implications for Marketers

Everyone in marketing is aware of the imminent decline of third-party cookies.

Google has already made 1% of Chrome users available as test cohorts for their technologies designed to replace cookies: Topics and the ‘Protected Audience’ API.

They are also moving Android OS over to the same tools as part of their Privacy Sandbox push.

These tools being offered to advertisers have received heavy criticism.

Initially, many different approaches were proposed by open working groups collaborating with Google, which have been whittled down by privacy watchdogs (such as the W3C’s Technical Advisory Group (TAG)) and industry bodies to the two main approaches left for the activity of placing ads:

• Topics for prospecting
• Protected Audience API for remarketing

Because of continuing criticism from those same groups, the TAG advised that Topics was functionally no different from cookies as regards privacy as recently as January this year – as it stands, both of Google’s new tools for advertisers are now opt-in.

The Privacy Sandbox rollout is underway, Origin Trials are live, with up to 7% of trial traffic allocated to A/B testing of these technologies today.

As marketers, what we should expect from these technologies is limited.

They are opt-in, meaning less than 2% or so of all Chrome traffic will ultimately be eligible.

Topics – despite having a slightly expanded taxonomy to choose from to associate ads with traffic behaviour – is almost comically limited, with 470 options.

Meaning it is still the case that, for the vast majority of advertisers, you will be buying ads at a higher cost due to auction volumes increasing, providing traffic that is less relevant than before, with higher bounce rates and subsequently lower conversion rates.

Costs will go up; quality will go down.

When Privacy Sandbox completes its origin trials and rolls out in June 2024, that will affect 62% of the existing browser marketplace.

For any marketer responsible for display ads – as well as Paid Social and Search – which rely on retargeting for large chunks of targeting and higher return media spend allocation, this is a catastrophic change that appears mainly calibrated to increase profits for Google’s parent, Alphabet, rather than improve user privacy.

Cookies are only the beginning: Mobile OS opt-in coming to Android

It has been little reported that Android is following in iOS’s footsteps and will essentially remove advertising data from Social at an even greater scale than Apple’s change.

Those familiar with the iOS 14.5 update in April 2021 – which required opt-in for personal data to be shared with advertisers, and which cost Meta somewhere in the region of $10bn in the year after its release – will know that the opt-in rate for Androids’ coming Privacy Sandbox is likely to be similar: 2% or so.

Which means, effectively, once Privacy Sandbox is completed by this time next year, advertisers will be left with a vanishingly small proportion of today’s targetable audience.

For context: Chrome has 62% browser market share and Android has 70% Mobile market share.

So, that is a devastating outcome for marketers.

We will have less targeted auctions to participate in, and those that do occur will drive much less relevant traffic at a higher price than today.

All of this is why, over the last year or so, as the cookie-free deadline has loomed ever larger, I have been fielding more and more requests from customers and CMOs for how advertising can be measured in the cookieless future – and how on earth marketers can spend efficiently to grow with such a major collapse of the audience ecosystem.

The good news is there is an alternative technology to cookies: AI.

Cookies and their place in your tech stack

Cookies were first used for web analytics in 1994, to:

• keep track of ecommerce basket contents
• allow more than one purchase at a time to be added to the cart during a session

So, it should be surprising that we are still using them in modern analytics stacks today.

And yet the vast majority of our analytics continues to be powered by the combination of a first-party pixel and cookie – including the latest and greatest version of Google Analytics: GA4.

In fact, all marketing analytics tools today, with the sole exception of Gartner quadrant leader Corvidae.ai, use cookies.

This is no surprise to me: Corvidae was spun out of my performance marketing agency QueryClick to launch as a standalone SaaS product last year.

Corvidae was built in response to the inaccuracy of our existing analytics platforms.

Its stitching process requires no cookies, instead using proprietary AI techniques to build cross-device conversion paths compliantly and securely for which Corvidae holds patents globally.

AI as a technology is now mature: witness the impact of the likes of ChatGPT, Bard and other LLM AI.

In marketing, we are blessed with two things that allow for highly effective AI application:

• very large data volumes
• clearly definable outcomes (conversions)

These allow for training, verifiable testing of accuracy and superb results in stitching anonymised conversion paths – replacing the role of the cookie in measurement and ad placement.

Today, Corvidae stitches paths that are 3-4 times as long a cookie-based paths, and which can be plugged into AdTech like Google ads to cut CPAs by 40-50%, and grow revenue by between 50-78%.

The return of contextual advertising and a new way to measure performance

Corvidae recently sponsored a Censuswide survey of 150 highest spending retail businesses within the UK.

From the results with ‘Marketing in the Cookieless Future’, we can see that today, 97% of respondents are concerned about the coming loss of cookies having a material impact on their ability to either understand which marketing activities are effective or even serve ads effectively.

I would like to explore an assumption about cookies that is causing this deep concern, which was also surveyed: 79.3% believe cookies are a technology that provides accurate data.

I have found, in the eight years of development and research into cookie-based data, that the idea cookies have ever offered accuracy is false.

Whether that be for measurement, or ad placement targeting.

When there was a less fragmented online advertising world – like the world when I first began my research – marketers mainly had only two sources of truth for their digital marketing activity:

• their own analytics platforms
• their Search Advertising spend

Both of these were completely dominated by Google Ads and Google Analytics in most markets back then – and to some extent, in analytics, that remains true today.

Google has always expended a lot of effort to make sure that there was no discrepancy between the reporting offered by its Ad platform and its Analytics platform.

However these sources of truth both depend on cookies.

With Facebook’s rise over the last decade, we have slowly seen a second major silo for our marketing spend emerge, which has caused question marks about the accuracy of Meta’s reporting.

This is often due to the fact that Meta revenue does not reconcile back to total cash in bank – or to revenue in analytics platforms like GA or Adobe, which are click-based.

We now of course operate in a multi-siloed world, with the recent explosive growth of Amazon as a key third channel for brands, Apple and Android for mobile app advertising and tracking, and the (perhaps brief, bans permitting!) current rise of TikTok as a meaningful alternative to Meta.

The blame for that discrepancy between self-reporting and cash in bank has been laid at the door of lookback windows for Display and Social, seeking to attribute value to an impression and a generous self-interest in claiming credit for sales where no click has directly occurred by Social, Mobile and Display ad platforms.

This is simply a variation on that age-old problem of an ad impression being given credit for a sale even though the ad may or may not have been seen.

And may even have been seen after the conversion occurred in many instances – as is rife in retargeting.

However, the cause is more pernicious than lookback windows and generous self-reporting.

We can find evidence partly in the failure to launch of Programmatic advertising as a meaningful part of marketers’ ad spend.

If third-party cookies are so accurate then why, after a decade, is Programmatic Display spend not at least a meaningful part of our marketing budgets along with Search and Social?

For most businesses, Display is a single digit percentage of budgets, and those are dominated by retargeting – explicitly not outreach or new customer acquisition – to try to capture customers who didn’t convert first time round.

Indeed, there are large, scientifically robust, studies that show cookies in fact do not do a good job in providing data for targeting ads.

For example, in 2018 – a much less fragmented online world than 2023:

In this study by Nico Neumann, Catherine E. Tucker & Timothy Whitfield, we see that to get any incrementality at all in cookie-based ad placement, advertisers must spend 2.5 times the cost of the ad for a normal display ad, and 1.6 times the cost for video.

In short, this study is telling us that cookies do not work as advertised.

It is only premium inventory online that actually adds value for ads – meaning the contextual targeting data intended to be provided by cookies is essentially ineffective.

Despite that same data being very effective when it was available on social platforms, like Facebook, where server-level data (including impression serve and viewability data) is available to Facebook’s ad optimisation engine.

The evidence is telling us very clearly: cookie-based targeting data is – and always has been – broken.

When looking at client analytics data eight years ago, we found that the same failure of the cookie to do its job was present in first-party cookies – those cookies used by your web analytics platform – GA4 and Adobe.

Even back then, we were able to show that first-party cookies are only 20% accurate.

This is compared to the 98% accuracy of customer conversion paths stitched together with AI.

The cause of cookie inaccuracy is simple: cookies do not stitch the vast majority of journeys together.

Instead, they only provide fragments of journeys.

Today’s online journeys are fragmented across:

• multiple devices
• apps
• browsers
• media types

The cookie is expected to persistently join all those interactions together, when it has no technical means to do so.

What was useful in 1996 is not fit for purpose in 2023.

This is why moving to AI-based measurement within platforms, like Corvidae, are key to claiming back control of measurement and allowing a movement towards ad optimisation that is based on the probability an ad is going to be relevant to your audience: the use of contextual advertising strategies, which I’ll get onto in this blog in a little bit.

First, lets cover the only alternative technology in the conversation: second-party data (AKA third-party data in disguise…)

Can second-party data matching save us?

75.3% of respondents within our recent survey responded that they need to:

• integrate analytics with ad buying
• keep new customer acquisition costs down

This has motivated attempts to replicate the cookie approach in other ways – for example the IAB Tech Lab’s Project REARC and associated media identifier frameworks.

There has been a major issue with these approaches, that cannot be ignored by serious brands.

However, there are significant legal issues with such an approach that ultimately snowballed into a finding that today rules the IAB to be in breach of GDPR.

Given businesses will be reluctant to incur significant fines by using direct cookie replacements, embracing an approach to marketing that does not have any personally identifiable information (PII) in its operation is essential.

Probabilistically generating conversion paths using AI is the only solution that currently offers any ability to measure and optimise media spend and report back on marketing performance with any degree of accuracy today.

And yes, I am considering MMM approaches when I make that statement, not least because of the endemic lack of ability to action any kind of tactical CPA optimisation with MMM (Media Mix Modelling / Econometrics).

So, what alternative options are businesses looking at?

A variety of approaches are being investigated by businesses.

42.7% will remove third-party data altogether from their marketing mix which, given the lack of incrementality discussed earlier, is an excellent trend to see emerging, and which will hopefully encourage greatly simplified tech stacks.

Around 48% of the cost of placing an ad is lost in the gap between the advertiser and the publisher serving the ad in ISBA’s landmark study from 2020: a huge gap that has driven a surge in poor quality inventory over recent years.

Chrome’s options (Topics and ‘Protected Audience’ API) are part of the picture (43.3%) – which is expected as they will be the only way to serve targeted display ads to 60% of browsers online.

However, as already covered, both Topics and the ‘Protected Audience’ API have been dogged by privacy concerns over the years, and today they both require opt-in by the user.

Contextual advertising is the frontrunner: placing ads where you think your important customers will be based on the content around the ad rather than dynamically targeting to cookies is a robust, simple approach that is being embraced by 40% of respondents.

Although, of course, it also brings a lack of optimisation control and finding favourable deals on high quality inventory is dependant on your ability to leverage good ad pricing either directly or with your media agency.

This is a transparent and controllable approach however, that finally removes waste, eliminates cost from your tech stack, and – when combined with accurate AI measurement – allows for CPA reduction even when growing.

For example, from our work with clients, we’ve seen how Corvidae conversion paths, when fed back into Google Ads, allowed a 71% increase in revenue, at a CPA 21% lower than before.

This is a huge, net new growth in revenue that demonstrates buying ads earlier in customer conversion paths makes you a clear winner in the digital auction market we all operate in.

Measuring the performance of contextual strategy activity will be key in the cookie-free world.

Hence, the interest in Unified Attribution platforms like Corvidae in the Censuswide report (38.7%) – and should those systems also be free of first-party cookies, as Corvidae is, they can offer extremely valuable optimisation control and cheaper new customer acquisition in the modern marketing landscape.

In summary: an industry sleepwalking into a compliance and targeting nightmare

Working with Censuswide on the report has been fascinating, but also worrying.

Our industry – marketers generally – are unprepared for the real impact of removal of cookies from the mix, which is already underway.

The majority still believe they can rely on tools like the IAB’s Unified ID – already ruled to be uncompliant with GDPR by the EU – or Chrome’s tools, despite their likely 2% opt-in rate.

84% still intend to use GA4 in the future despite the 202 EU finding ruling it uncompliant under GDPR and local DPAs in France, Italy, Austria and other EU countries declaring it illegal, creating hundreds of legal cases already.

On the same basis, Facebook was recently fined by the EU $1.2bn – this is a live issue.

To everyone reading this, I hope you can see there are other options out there that are:

• compliant
• don’t rely on cookies
• and allow for dramatic CPA reduction and net new revenue growth

It’s crucial that you don’t spend any more time delaying action on cookies: marketers need to start to adopt forward thinking tech stacks and embrace modern marketing strategies that also respect our key markets’ data compliance regulation today.

If you’d like to find out more about Corvidae, you can get in touch here.

And please do download a copy of our recent report: Marketing in the Cookieless Future below.