What you need to know:

• The World Federation of Advertisers is developing code that can be repurposed in each market to build a cross media measurement solution looking at reach and frequency across television and digital.
• It is a hybrid approach, using census log data and a panel to be accurate and independent.
• Big tech players, including Google, Amazon, Meta and Tiktok parent company ByteDance, are on board to share their data.
• The code involves virtual IDs – synthetic audiences – and a panel that teaches a model how users in a market behave. “Panel only solutions don't have a future,” per WFA’s Matt Green.
• The AANA insists it will “absolutely” lead an Australian cross media solution, but it is waiting for more data from pilot markets in UK and US before doing more. Canada and Germany have set up steering groups to be “fast followers”, a stance former AANA boss John Broome said Australia would adopt.
• There’s no guarantee Big Tech will benefit from a cross media solution, said Green. In fact, he reckons "there's a good chance that reach from a TV side might look quite good”.

Global advertisers think they are on the verge of solving one of the biggest challenges in media – cross channel campaign measurement. They're backing synthetic audience panels and a learning modelling algorithm to crack the code.

Major advertisers like Unilever and P&G have for years sought a way to measure de-duplicated audiences across television and digital, which could cut waste by more than 10 per cent.

The World Federation of Advertisers (WFA) is working on the raw code to make it happen. Already Google, Amazon, Meta and TikTok’s parent company ByteDance have signed on to share their data.

The WFA has been quietly building the code underpinning a white-labelled cross media measurement solution under the name Halo. Written by 15 to 20 engineers from the likes of Nielsen, Kantar and Comscore, the code is being used in pilots by ISBA in the UK, under what is known as Project Origin, and the ANA in the US. Other countries – Canada and Germany, for example – have steering groups set up to take the code and build a local solution.

“This is the future of measurement. You use data to train models. That's the way this industry is going,” per WFA’s Director of Global Media Services, Matt Green.

“It's a virtual ID that's being trained on a very accurate, reliable, gold-standard panel. The panel is still there. That's not gone away. In fact, that needs to be as reliable and accurate as ever. But it doesn't need to try and measure everything because that's not realistic. It just needs to do what it does.”

Local industry not moving yet

The WFA says the peak body for advertisers in each market should push forward cross media measurement plans. The Australian Association of National Advertisers (AANA) said Australia would be a “fast follower” market under previous incumbent John Broome. Current CEO Julie Flynn insists that stance has not changed.

“We are following the pilots closely and looking to provide feedback on those pilots to our members and the industry,” said Flynn. “If the development of the open source code and the pilot proves successful, the AANA will look to engage with our cross-industry partners to develop a solution for Australia.”

Seven’s CEO, James Warburton, told a media briefing last month he was open to cross media measurement, but it would be a difficult discussion.

“I don’t think we’re opposed to any of that,” he said. “But we’ve got our work cut out just getting our industry aligned and moving in the same direction, let alone joining with the rest. I’m not sure the conversation has really been had.”

Big Tech and TV

With Meta, Google, Amazon and ByteDance on board, there may be some hesitation among television networks to join a cross media solution that directly compares a TV impression to a Facebook ad. “If you leave it to the TV guys, it won't happen because they've got an interest in it not happening… which I can understand,” Green said. “Broadcasters are more reticent to have their inventory compared to other parts, to other actors in the industry.”

But that’s not warranted, he added. There’s a risk for the Big Tech companies as well.

“They (Big Tech) do want to be better measured, and they will be hoping that they won't look terrible and that they can continue to capture more ad spend through it,” Green said.

“But I think it's going to be actually really interesting to see. I don't think it's a given that they'll do well out of it at all… TV companies have been reticent and concerned about this, but actually I think they might do really well out of it. Because we all know how we behave on digital, don't we?

“I think there's a good chance that reach from a TV side might look quite good in comparison to some of these platforms. The impression count is high from these platforms, obviously, but as a fully viewed impression, fully viewed reach metric that's not going to stand up as well.”

The Halo solution uses data from existing measurement solutions, like TV’s panel-based currency in the UK, BARB, for example.

“That's the currency also upon which the whole multibillion-pound UK TV market is traded. And what Origin is doing does not seek to replace that. That continues to be a foundational part of the way that TV media is traded. The solutions that we are building are solutions which would sit across the top,” Green said.

“So you can log-in today to see what yesterday's unduplicated reach and frequency was between TV and digital. So that's like a reporting layer. But actually you could become a bit of an optimisation product as well, because you go, okay, 'great, as it transpires, I'm building up too much frequency on YouTube here. So I'm going to move some money into Channel Seven because there's an opportunity'.”

The privacy problem

There are big privacy hurdles to overcome to develop a cross media measurement solution.

As an example, imagine a person called Joe. Joe consumes linear TV, BVOD, Disney+, Netflix, Mi3, the Sydney Morning Herald, BBC, and has an Uber, Facebook and YouTube account. He has three emails, different social IDs, and different phone numbers. Matching all of those together, measuring his activity and determining reach and frequency of ads would be a fraught issue.

A cross media measurement tool might use one of those IDs – an email address, for example – but would miss platforms where he logs in with a different email.

There were other suggestions – creating a new common ID currency, for example, or using a single identity or data provider. Those were seen as not future-proof and likely have low match rates. The “Virtual People ID model”, which maps different IDs using observations from a panel, is what the WFA settled on.

“We're building a synthetic society here, to use a different vernacular. It sends that virtual ID back to the measurement solution, it says 'Here's a virtual impression, which is a male aged XYZ', based on how the model has examined that impression,” Green said.

“It's the best way of doing it, I think, because it just avoids any of the difficult privacy conversations. This is a very privacy safe approach which everyone can be comfortable with. It gives us what we need.”

How does it work?

The Halo virtual ID model is fairly straightforward. There’s a virtual ID or VID for every person in a target population – circa 25 million VIDs in Australia, for example – that match the demographics of the market.

“If there are 60 million people in the UK, there should be no more than 60 million Virtual IDs, and if there are 30 million males, there should be 30 million male Virtual IDs, and so on and so forth,” Green says. The model uses government census data to generate the virtual ID population.

“And then basically, you can get to a place where you can just count up your Virtual IDs, which should give you a decent representation of reach and frequency of your campaign.”

Publishers and data providers assign these VIDs to their impressions independently of their own internal systems, using a neutral third party. That way, a news website can keep their impression data secure inside their own systems.

The model uses existing identifiers, profiles and other impression data to get the VIDs right – if a news website knows a view is a female user between 18 and 25 years-old, they’re likely to be assigned a similar VID.

Importantly, the model is underpinned by a panel. Each market that selects, controls and audits its panel group. If a publisher doesn’t have any data on a user, the best guess comes from the panel.

“The more data that's available to the model, the better. The better it can label those impressions. But the default mechanism is to use the characteristics in the panel. Therefore, it would be no more or less accurate than the panel,” Green said. But panels alone, as they have been used in the past, are effectively useless when it comes to cross media measurement.

“Panel only solutions don't have a future. You could say that with complete objectivity. There's no future for panel only solutions,” he said.

“They can barely measure TV with all the channels that exist there. So you're never, ever going to have a panel which is big enough to measure TV and digital.”