Origin, a cross industry initiative led by advertiser trade group ISBA which wants to establish a cross media measurement service for campaign planning, reporting and evaluation, says a crucial series of proof of concept tests have yielded very positive results.
The tests show that Virtual ID, a core component of a proposed global cross-media measurement solution, is effective at integrating TV panel-based data with digital, impression-level data. Origin says the results take the industry a step closer to advertisers’ desired ‘North Star’ for cross-media measurement.
Introducing ‘Virtual ID’
Origin was set up in 2019 following a new initiative from the World Federation of Advertisers (WFA) to establish the basic principles for effective cross-media measurement. The WFA, alongside national trade groups like ISBA and the ANA, individual advertisers, and many of the tech giants including Facebook, Google, Snapchat, Amazon, and TikTok, has been developing technical standards and the core tech components for a global cross-media measurement solution.
This global effort, dubbed ‘Halo’, has now developed a full framework for a privacy-safe solution, as well as some of the core tech components.
Origin’s role has been both to feed into the global discussions based on its conversations with local players, and to take the tech developed at a global level and translate it into a workable, privacy-safe solution.
A key component of Halo’s technical framework is something called the ‘Virtual ID’ (VID). VID sits at the heart of Halo’s solution, and its role is essentially to label impression data consistently across all data providers, allowing different data sources to be effectively and consistently combined.
Joe Lewis, Origin research lead, described the Virtual ID as “the central deduplication model that will sit at the heart of the solution.”
Lewis outlined how VID works. “In short it’s basically a micro model approach where every single impression from any publisher is assigned to a virtual person, and that’s done based on a probabilistic mechanism. But the beauty of it is that as deterministic data is collected, ingested, validated and verified, the model will start to use that deterministic data, instead of probabilities. And so this will extend to things like identity networks.”
The WFA lays out the VID model as follows:
-
A ‘respondent’ ID listing is created equal to the Census population.
-
Gender and Age Group classifications are assigned to these based on universes sizes.
-
A homogeneous group indicator is assigned to these indicating groups of similar behaviour for channels in the media campaign.
-
Relative rates of advertising exposure to each channel are assigned to these homogeneous groups.
-
Given Census campaign impressions, these are assigned in a probabilistic way using an allocation algorithm.
-
Cookie information collected for the Census impressions may also be used in this allocation process.
Lewis added that “one of the beauties of this approach is that it will allow us to effectively use first party data, and hopefully in time provide use cases beyond reach measurement – such as frequency optimisation, and linking to outcomes via a.probabilistic mechanism.”
Halo proved that this model works in theory, but ISBA’s job has been to demonstrate that it works with real data, and TV panel data in particular. Matching up impression level data with panel data, where one individual on a panel represents a big section of the population which fluctuates daily, is challenging. It’s important that any cross-media measurement tool is able to accurately preserve TV’s specific reach and strengths.
The exact nature of the tests is quite complex, and those who wish to dissect the data in depth can access the results here. But in essence, ISBA worked with research group RSMB and census data company IPSOS Mori to test different variations of its model in their ability to measure total reach, and found that while each of the variations have their own strengths and weaknesses, the concept as a whole holds up.
The final report from RSMB notes that “results were encouraging in all methods, broadly conserving systematic variations in TV campaign reach across channels within the Virtual ID model. The preservation of reach suggests potential for the model to work well across TV and website campaigns, as well as website campaigns themselves.”
Origin says this means a vital threshold has been crossed in developing a working cross-media measurement tool. There’s still more work to be done, particularly in testing the model at full scale in the real world. But this proof of concept test, alongside Halo’s own tests, show that the model and the maths hold up.
Holding the door open for broadcasters
While the results are very good news, and generally exceeded expectations according to Origin, there’s still plenty more work to do.
After further testing, it will be up to each individual market to translate the WFA’s principles and tech components into a full, working solution for the market. ISBA, through Origin, will facilitate the creation of such a tool in the UK, while Origin’s director Richard Halton says he’s seen enthusiasm towards the WFA’s work from a number of other European markets.
And as Origin’s work continues, Halton says he is “holding the door open” for broadcasters to get more involved in the initiative.
While the tech companies have been very cooperative and keen to get involved, broadcasters have been much less so. In the UK and US, broadcasters tend so far to have prioritised NBCUniversal’s CFlight standard for cross-media measurement.
But there’s no reason for CFlight and Halo not to work in tandem, and the more broadcasters get involved, the stronger Halo and the eventual tools developed by Origin and other local initiatives will be.
One common concern from broadcasters is that cross-media measurement solutions will undervalue their inventory. A premium broadcaster is unlikely to approve of any system which equates an ad on their prime time content to an ad shown to a user as they scroll through Facebook.
But Richard Halton reiterated that what Origin is building at the moment is just trying to accurately count how many users have been reached, without placing value on any given impression. It will always be up to the advertiser to decide the value of each media source.
Halton said that he hopes these test results will encourage more broadcaster participation. “Quite a lot of the work over the last six months has been to satisfy the concerns we’ve had and broadcasters have had about whether the solution we’re developing works as well for panel-based TV data as it does for first-party platform data,” he said.
Origin has also recently completed an audit from UK TV measurement body BARB, which gives BARB’s stamp of approval to Origin’s methods and allows Origin to license BARB’s data.
And Halton added that he feels it’s very much in broadcasters’ own interests to establish working cross-media measurement.
“We’ve had some excellent roundtable discussions with very senior CMOs who say they think they are underinvesting in certain media because they don’t have comparable data, and we’ve certainly heard cases where it’s TV which is being under-bought,” said Halton.
