What you need to know:

• WPP preparing for internet to be “95 per cent untrackable” in short order, and alt-IDs based on hashed emails to be regulated out, with Australia first to move.
• GroupM Nexus CEO Ryan Menezes suggests one-to-one targeting high risk, one-to-many prospecting for new customers much safer via cleanrooms where data never transferred.
• GroupM “never” wants to own data, given incoming risks.
• Very different model to previous approach, and those pursued by rival holding companies.
• But Menezes predicts ID “loopholes” will rapidly close.

Performance problems

WPP is preparing for 95 per cent of the internet to be untrackable “in the very near future” due to privacy overhauls – with Australia at the vanguard, according to Ryan Menezes, Australian CEO of the holding company's performance unit, GroupM Nexus.

Menezes warns the local market has not yet fully grasped the implications of the first tranche of Privacy Act legislation laid before parliament last September – and what it means for data matching of tagged, ‘de-identified’ user IDs. These digital footprint trackers are based on the hashed email addresses now used across swathes of the media and marketing industry to underpin customer acquisition through user ad targeting.

Loopholes closing

Menezes said the specific language used within the Privacy Act overhaul is in effect “closing the industry loopholes that technology companies have used to be able to create what they call de-identified IDs”. He reckons the industry is now at a crossroads – publishers, brands, agencies and the broader digital supply chain.

But some are more immediately facing existential challenges than others.

Amid a looming supply chain shakeout – with third party data brokers “heavily at risk” – Menezes thinks data may become more expensive, purely because there is less of it.

But he reckons what is left legally legitimate will ultimately be better quality.

“It’s just a prediction,” per Menezes. “But we are at an inflection point.”

Hashes smoked

Tech firms argue that because the individual is not explicitly identified, hashed email addresses don’t count as personal information. So matching those hashed identifiers in order to target people around the web is ‘privacy compliant’.

Consumer advocates and privacy specialists have long warned that is not the case. Australia’s regulators appear to agree, recommending that lawmakers treat de-identified data as personal information.

GroupM is preparing for that recommendation to be implemented – and the firm globally is pulling away from the identity and data ownership-based approaches being pursued aggressively by rival holdcos. Hence this month acquiring clean room provider InfoSum.

95% untrackable

GroupM thinks closing the hashed email loophole, “is probably one of the [recommendations] they will enforce – and if they do, it will be the first time globally that has ever happened”, per Menezes.

“Our position is we believe that de-identified IDs will become redundant in the near future,” he told Mi3. “Today I'd say around 35 per cent of the open Internet is trackable. We believe that in the very near future, almost 95 per cent of the open internet will become untrackable.”

Apple’s hobbling of tracking already means circa half of the Australian market is essentially untrackable, hence targeted ads skewing almost entirely to the other 50 per cent of the population, “because the programmatic systems only prioritise where they can see the signals,” per Menezes. “We see that continuing to shrink.”

Prospecting hobbled

Closing the hashed ID “loophole” has major implications for brands, publishers, agencies and the digital ad supply chain, said Menezes.

“Forget about how clients target their existing customers – because in that instance, a lot of the time, they've got the consent management and the privacy controls in place.

“But if you think about how we prospect customers [i.e. new customer acquisition] – we're heavily reliant on the different platforms and walled gardens.

“Every single platform has their own identifier, The Trade Desk has its UID [Universal ID], Google has its own approach … To be able to connect with customers, and to work within the current ecosystem, they've all built their own identifiers – and that's the space where we feel there will be considerable impact, particularly to how we prospect customers.”

Leave it out

Hence GroupM moving to a model where data “never leaves the bunker”, per Menezes, and instead remains with its original owners, but with the insights from that data connected and essentially turned into ‘lookalike’ prospects via AI-powered federated learning models.

“The data actually never moves, so the ownership and the control of that data is always in the hands of the data controller – and that could be the client, it could be the publisher or the partner,” he said.

“They always continue to own and continue to maintain that ownership and control of their data. It never, ever, ever leaves its location and you will never, ever lose control of that. So that's the way we foresee it.

“The way federated AI models work is the data stays where it is, the model continues to get updated, and that connects straight in with the piping – or the media activation – to be able to make sure that we can target and connect with prospective customers in a more meaningful way, not on a one-to-one level, but on a one-to-many level.

“But all of it sits with its own bunker. So we're not in the data ownership business – and we never want to be.”

Holdco zag

Ditching identity graph accumulation is a distinct shift – and counter to the strategies being pursued by rivals. Of the ad holding company majors, Publicis Group has made the largest data and identity investments in recent years, spending billions on Epsilon and more lately, Lotame, while IPG spent US$2.3bn in Acxiom in 2018.

“If you look at what else exists in the industry, with some of the other big holdcos and the data structures that they've built, it's all around the ownership of the data and how they use that to be able to connect with customers,” per Menezes.

“Our piece is, we own the technology where the data lives, and the technology helps you collaborate in using the different data sets.”

GroupM had previously attempted to build an identity-based approach via MPlatform, “which was probably very similar to what currently exists with some of the other holdcos – it primarily relied on cookies and de-identified IDs”, Menezes acknowledged.

“To keep that platform functional, you're constantly having to drop data agreements for the data that gets pulled into that, which you're basically then using for activation.

“That is high risk territory, because number one, there's an immense amount of liability involved in drawing up these data agreements, and you're not really sure if you've got the consent to be able to actually use this data; and number two, it's completely grounded in IDs and cookies, which means that it'll be redundant in the very, very near future,” he said.

“Hence the InfoSum acquisition – to be able to actually facilitate [prospecting] using federated AI models … It will take the learnings out of the individual data, yes, but the application is never on an individualistic level,” added Menezes.

“The model uses hundreds of different variables of the information that it gathers and applies it on almost a cohort level – and go and find those [lookalike] cohorts. It's a move away from precision targeting into predictive targeting – that’s probably the best way of explaining it.”

He predicts the broader ad industry will ultimately take a similar approach.

“I think any of the big tech companies will move to these broader, federated, cohorts-based models – and they'll be able to shift the fastest because they've got the scale of data that's accessible.”

Postcode gold

As new privacy rules take shape, Menezes thinks the ad industry will need to reorientate around a different set of signals, with contextual and geospatial data coming to the fore.

“Geospatial data, for us, is super interesting. For example, if you're in the FMCG category, you could model the purchase propensity to SA2, or SA4 geolocation levels [the geographical classification defined by the Australian Bureau of Statistics] or even postcode location levels across the whole country,” he said.

“That gives you a really good understanding of the geospatial areas where you have a higher propensity of people buying your product, and where there's greatest opportunity for growth.

“Then you can start to model that based on, for example, your competition – you've got a multitude of different data assets that you can use to your advantage. And you’re not using an individual there [to target], you're using different types of location, which is where we think there's opportunity.”