David Murphy talks to Jason Cooper, general manager – mobile, at Integral Ad Sciences, about viewability, brand safety and fraud.
MM: So Jason the area you work in, around integrity in mobile advertising if that works as a catch-all term, is very hot right now. How did you end up here?
JC: Yes we deliver media quality for brands and agencies around ad fraud – is your ad being seen by a human? Viewability – is it in view in the mobile browser or app? And brand safety – is the place where it is appearing appropriate for the brand?
We do this for display, video and mobile, and mobile is the most recent addition to our portfolio. It came about through Integral Ad Science’s purchase of Simplytics 16 months ago. Simplytics was an agency-side ad server for mobile that was built to do better job of ad serving and analytics, for brands and agencies. Most agencies have a Google DFA-type (DoubleClick For Advertisers) product, but the challenge is that it’s built around a URL/cookie. Then apps turn up and it falls over, so we thought we could do a better job.
The issues on mobile started out three years ago with viewability initially. Ads go through a lifecycle. In display, you had good bandwidth, Flash worked, ads loaded quickly, so there was no problem. But then people started putting banners into apps over 3G, so now you have latency - the ads were bigger because there was no Apple support for Flash on mobile, so brands and agencies had to work out how to do HTML5 on mobile, ending up with an ad unit that is 10x the file size, and deliver that in a snappy way over 3G.
With the work we were doing at Simplytics, for every impression served, we had four impression-related metrics that agencies had never seen before. First we had Delivered Impression – did the ad tag get to the browser? Then Verified Impression – this is the ad tag saying: ‘Am I in app or mobile web because I have a different set of instructions for apps’ which causes trafficking issues. Thirdly, was it rendered/viewable? To serve the ad, you are downloading assets from a third-party ad server, so if they don’t load in time, people would not see the ad. And then finally, for any mobile web traffic, we also reported on the domain, so did the ad run on the Guardian or a porn website for example. So all of this was in our reports and agencies and advertisers were not getting all that from the ad networks.
We were doing this for our own interests initially, to try and debug what happens when you send an ad down the pipe, but as we saw the results, we then went to the agencies and told them that their ads looked great, but in many cases they weren’t being seen.
MM: That must have rubbed some of the ad networks up the wrong way I imagine, if you are exposing how poorly a network is performing in actually serving the ads the advertisers are paying for?
JC: Anyone with a long term vision should be endeavouring to make sure they have a clean, high-performing network and should not be afraid to be transparent, because transparency leads to trust and that leads to increased revenues.
MM: So that was then, where are you now?
JC: So at Simplytics, we were looking at in-app and mobile web viewability, and we had deals in place with Starcom and Omnicom so we have spent the last year unpicking the functionality out of the ad server and putting it into the core product.
The main product is around viewability in-app and on mobile web. We started in-app nine months ago and on mobile web four months ago. We are tapping into the MRAID standard. MRAID is designed to solve a creative problem, to make sure the ad just works, but as part of that there is a hook to give us an insight into the viewable state. Most mobile ad networks and SSPs (Supply Side Platforms) use it, though some outliers such as Facebook, Tumblr and Twitter do their own thing.
The Media Ratings Council (MRC) issued new guidance on viewability a few months ago which follows the display standard, where 50 per cent of the pixels have to be in view for greater than one second. But how do you calculate the number of pixels in view in an app as opposed to a browser. With the browser you have tracking pixels, but not in-app, so we have proposed a Verification API (Application Programming Interface). We believe it will solve the problems we see in apps round viewability, and also help with brand safety.
MM: So how does it work?
JC: We have developed our own SDK (Software Development Kit) that needs to be bundled into the ad vendor’s SDK, to allow us to query the viewability of the ad independently, which the MRC and the IAB (Internet Advertising Bureau) both want. So we have written the API, we’ve reached out to agencies and said: ‘These are the problems that need solving, will you support it?’
MM: And what’s the response been like?
JC: We have had public support from Starcom and M&C and also some private support. And we have asked the IAB if they think it has a home in the way that MRAID has a home. We’ve been very open about it, presented it to the IAB and MRC’s working groups to solicit feedback, and we are about to build a test SDK which we’ll get working with MoPub and Millennial and we have an open dialogue with a lot of others. So we are hoping that it will sit within the IAB and have working group support, but we can still build out the solution whether that happens or not.
It will also help with brand safety by enabling us to identify the app where an ad is being served. In display we have a brand safety proposition that uses the URL to decide whether to serve the ad, but apps don’t have URLs, so we need to find a way to disclose the app name. The verification proposal solves this, but again, it depends on the SSPs supporting it.
MM: One of the other issues with viewability is just the discrepancy in how it’s measured. I heard one newspaper exec recently bemoaning the fact that his publication has scored between 50 per cent and 80 per cent viewability for its online ads. Is this issue anywhere close to being resolved?
JC: I guess the first question I would ask is, is the company giving you the viewability score MRC-accredited? But I think it’s also fair to say that the way we consume media on the phone is different to online. We’re out and about looking for information quickly, using our thumbs to scroll up and down and in and out so the guidelines for mobile probably do need to be reviewed and the MRC has acknowledged this.
MM: OK so we’ve covered viewability and touched on brand safety; what about fraud?
JC: It’s a big issue in mobile. What gets most of the press is bot fraud, where advertisers are paying for impressions served to bots, rather than humans, but location fraud is also a big issue in mobile. This is where the developer or publisher sends a fake lat/long coordinate in the ad call in order to increase the CPM they are getting, because knowing the user’s location generates a higher CPM. Ford, for example, would pay more to serve an ad to me if they knew I was close to a Ford dealership, so you get these lat/long coordinates in ad calls for ads served in apps where the user’s location is not required and is completely made up. The verification API makes the true lat/long coordinates available to us so we can identify these fraudulent location calls.
MM: And at the risk of asking a stupid question, how important is all this stuff?
JC: It’s huge. With fraud, we need to develop new models and classification scores to understand the severity of the issue so being exposed through lat/long will help in this respect. With viewability, there is a level of comfort because some apps have vector placeholders for ads, so there has been this assumption that the ads are always in view within apps, but in fact, our benchmark says only 81 per cent of in-app ads are viewable, which is almost a fifth of in-app ads not being seen.
And of course brand safety is hugely important - if there is no way of independently knowing the environment you are going into, such as the name of the app and its content, that’s a huge risk that no brand should want to take.
Jason Cooper is general manager – mobile, at Integral Ad Science