Fundamentally, Users Aren't Cookies: Q&A with Ezra Pierce, CEO, Avocet & David de Jong, CEO, Screen6

With people now almost exclusively cross-device, effective measurement and management of their consumption is crucial. ExchangeWire spoke with Ezra Pierce, co-founder and CEO, Avocet and David de Jong, co-founder and CEO, Screen6 about their partnership, which allows cross-device conversion management within a DSP and how the industry can tackle the cross-device challenge. 

ExchangeWire: Screen6, a cross-device platform that operates relatively under the radar partnering with Avocet, a relatively new DSP in the market – how did the partnership come about?

David de Jong: Both leadership teams have known each other for a couple of years now, during which time we have picked each other’s brains to gain new insights on various technologies. Since about a year now we have been trialling some of our newest platform innovations with Avocet, which have been a steady supporter of our company. Avocet adapted their platform to use our cross-device identities with ease, and were able to launch recently.

What are the capabilities of managing cross-device conversions within Avocet?

Ezra Pierce: There's a long list – attribution, frequency, audience targeting, reporting, etc., but the shorthand is that everything that relies on the concept of a 'user' in Avocet is cross-device enabled. Fundamentally, users aren't cookies and the technology has to reflect that or everything is compromised – conversion counts, reach estimates, targeting capabilities, everything. It's not a bolt-on; it's a fundamental shift in how you consider the user.

How are clients utilising the cross-device functionality?

Ezra Pierce, Avocet

Ezra Pierce, Avocet

Ezra Pierce: The biggest win has been in properly measuring the impact of mobile advertising. Some conversions just don't happen on mobile devices, but that doesn't mean that mobile advertising doesn't drive conversions, it just means that the value is only visible within a cross-device paradigm. That's a great information-asymmetry to have in hand, seeing the value of something that everyone else is missing.

The other major change is in how data is used. Rather than data being locked to its origin (a browser or mobile device) Avocet's clients are now able to deploy user data from any origin across all the user's devices. That has a number of knock-on effects. With more impressions available against a particular segment we can be more selective about where and when we reach a user. It's a cliché, but it gets us closer to 'right message, right person, right time'.

Why are DSPs in general not focused on cross-device?

Ezra Pierce: I don't think it's a lack of trying. More than one of the major DSPs have purchased a cross-device company with a view to acquiring cross-device functionalities. Weirdly those acquisition haven't yielded functionality, so you have to ask "why?". My guess is that it's just hard for a legacy DSP to go beyond a trivial implementation of cross-device. Moving from the deterministic world of cookies and device IDs to a more robust user-concept is like performing open-heart surgery.

Avocet had the advantage of beginning with the position that 1) impressions aren't views and 2) users aren't cookies; so we architected a system where an impression was seen as a possible view and a user was a network of probabilistically connected cookies, statistical IDs, and devices. No one was planning for that in 2009, but in 2014 it would have been mad not to.

How do you deal with concerns over cross-device match-rate accuracy?

David de Jong: I think there are a few misconceptions here. Match rates are typically referred to as the ability to connect data between two (or more) sources. What's essential in cross-device matching is that the percentage of known cross-device users is as large as possible. This can be referred to as reach or, more technically (if verified), as recall.

Because Screen6 creates a private device graph for each of its clients, these clients can verify Screen6’s data against their own deterministic data easily and continuously, unlike competitors who don’t offer full access to the device graph. Our clients consistently verify, with high average precision (90%+), across the whole dataset.

Ezra Pierce: We have a simple approach to the match-rate accuracy, it's just a measure of uncertainty and when you know the uncertainty around a prediction you can control for it. So we attenuate cross-device conversions, reach, etc. based on the match-rate accuracy. That's the mathematically honest way of handling match-rate accuracy.

The amusing contradiction is that people are very focused on match rates, the false-positive error rate, but are blithely unconcerned about the false-negative error rate – the users who aren't matched to their devices. In other words, if you don't identify users cross-device you will report erroneous reach, totally miss users that you should be targeting, and have erroneous conversion counts... 100% of the time. The decision is whether to make a good prediction of matches with well-understood error rates, or make no prediction and know that you're always wrong since you're implicitly assuming that users all only have one browser or device.

What are your views on probabilistic versus deterministic data?

David de Jong, Screen6

David de Jong, Screen6

David de Jong: A popular discussion point in all of our meetings with clients, partners, and prospects highlights the fallacies of deterministic data. While many in the industry are still under the notion that deterministic data guarantees 100% precise matches for users cross-device, some of our counterparts have finally admitted publicly that relying on log-in data from multiple devices has its shortcomings.

The fact is that even the publishers with the largest number of cross-device, logged-in users have challenges with scale and require the implementation of a probabilistic method to understand their entire ecosystem's activities.

Also, and this is more evident in Europe, privacy regulations are more in favour of probabilistic solutions, which don't require any PII.

With the partnership making Screen6 and Avocet trailblazers, do you think other cross device platforms and DSPs will follow suit?

Ezra Pierce: For the overall health of the ecosystem, they have to. DSPs have a central position in the ecosystem and a responsibility that comes from that. We're the point at which data and media is targeted, predictions are made and the bids are issued. That means that the progress of the industry is limited by the speed at which DSPs enable innovative companies like Screen6. Looking ahead, it's hard to see how a DSP without cross-device can be relevant in an increasingly fragmented media space.