Independent ad tech companies including Quantcast, Tapad, Drawbridge and AdBrain recently unveiled 'open data' initiatives to help brands scale their automated advertising efforts with methods that run counter to those proposed by 'closed ecosystem' players such as Facebook and Google.
Quantcast is poised to host its Supernova event in London today (22 October), just weeks after it launched its Audience Grid product that lets advertisers target audiences across different web environments using a probablistic data-targeting method.
This means advertisers can target desired audience segments across multiple devices by matching their own data sets with third-party data profiles - effectively giving them a 'single view' of the customer across a number of screens and in different web environments.
Similarly, Exchange Lab also recently announced that it has partnered with independent ad tech companies including AdBrain, Drawbridge, Tapad in a tie-up that lets advertisers buy media on its exchange using their respective cross-screen targeting capabilities using similar probablistic methodologies.
The accuracy of these probablistic targeting methods ranges depending on the quality of the data, with estimates over accuracy rates ranging from 60-to-90 per cent. This runs counter to the deterministic targeting methods offered by the internet's biggest advertising companies such as Facebook, and Google, which use their logged-in user data to target audiences across screens (this is commonly known as deterministic targeting).
Deterministic players are keen to point out their first party data has accuracy rates superior to those offered by their rivals, but such a method means advertisers limit third-party verification of their results, and restricts advertisers to their ecosystems (hence the term 'walled garden').
The result is that advertisers must target audiences in each individual walled garden, giving them limited capability to synchronise campaigns across screens, and the different web environments, plus there are questions over how much data advertisers 'donate' to said players.
The case for open
Commenting on the debate: Konrad Feldman, Quantcast CEO, said probabilistic modelling doesn't require any Personally Identifiable Information (PII) thus it can help assuage concerns over whether or not it will incur a consumer backlash.
"We believe that an open platform that makes available data from a broad range of sources and provides flexibility for advertisers and publishers to mix and match as they deem appropriate offers the greatest utility for an industry desperately trying to make better, more consistent, scaled and accurate decisions from data," he added.
Responding to the claims that deterministic data sets were 'better' compared to the probablistic modelling, he said: "There are trade-offs to both approaches. Ultimately ‘better’ can only be gauged in the context of the outcome. Deterministic data can be accurate, but is often lacking in scale – you only know about attribute X for those that have declared it."
Commenting on the same issue, Gareth Davies, CEO of AdBrain, echoed Feldman's point on the probablistic modelling being less likely to infringe targeting regulations around PII.
He added: "Deterministic matching is precise, but does it scale? Facebook has been particularly successful at building rich behavioural profiles around their users, but ironically the more precise your targeting becomes, the less you're able to actually scale campaigns."
Is deterministic 100% reliable?
Critics of the probablistic approach often claim that these methods do have gaps, despite the message such walled garden players espouse in the market. AdBrain's Davies told The Drum the primary place where this model falls short of the 100 per cent accuracy mark in on mobile web traffic where internet users don't normally use their logins.
He added: "One size doesn't fit all, it turns out Google and Facebook, despite their massive success, are bound to encounter blind spots. With 968 million daily active users Facebook is capable of reaching 13 per cent of the human race on an average day.
"That's impressive, but the reality of consumer behaviour is a lot more nuanced than big numbers. Take for example the different environments we use to access the internet; mobile browsers, desktop browsers, apps, connected devices. Making sense of data across these disparate environments is a lot more complicated than just relying on users to login."
Mobile device usage accounts for 60 per cent of all internet traffic, and a large portion of mobile browser usage is the equivalent of "going dark", according to Davies.
Although, he did add: "Google is a unique case, between Android, Google Search, and the DoubleClick stack, they reach all corners of the world. But without unique identifiers like a Gmail login, it's difficult to precisely connect all of this data together."