The MRC Is Cracking Down On Invalid Traffic, But Discrepancy Issue Lives On

MRCfraudguidelinesDespite ongoing efforts by the Media Rating Council to close the reconciliation gap between vendors, discrepancies continue to haunt viewable impression counts.

And there are some who say the MRC is needlessly adding complexity to the process with its reporting requirements.

“There’s a lot of slicing and dicing of the data going on,” said David Hahn, SVP of product management at Integral Ad Science.

Hahn was referring to invalid traffic guidelines put out by the MRC in October, following a comment period over the summer, that expand the definition of what invalid traffic is by splitting it into two categories: general invalid traffic (aka traffic detected using routine means) and sophisticated invalid traffic (the tough stuff).

“The general cases are the black and white issues – either it’s invalid or it isn’t,” said David Gunzerath, SVP and associate director at the MRC. “With sophisticated, you get into the gray areas where vendors might be using different probabilistic techniques and making a judgment, and that’s where we see less consistency between vendor A and vendor B, because they might make those judgments somewhat differently.”

General Vs. Sophisticated

General invalid traffic detection is a matter of rooting out known problems – data center IPs, for example, or known bots, spiders and other crawlers. Vendors check them against a list of offenders and chuck them out of the overall viewable impression count. Done. Sophisticated invalid traffic is far more difficult to detect, including thorny issues like adware, malware, hijacked devices and tags, falsified location data, stacked ads and cookie stuffing.

Having made the distinction between general and sophisticated invalid traffic in its guidelines, the MRC is also adding more nuance to the accreditation requirements by turning general invalid traffic detection into a prerequisite for accreditation. If vendors want to get accredited for sophisticated invalid traffic detection, they’ll have to go through a separate accreditation process.

The MRC also wants accredited vendors to get more granular in their reporting – and that’s where things get a little hairy.

Whereas in the past vendors would have provided clients with one metric to quantify total viewable impressions (overall impressions minus overall invalid traffic), the MRC is now asking vendors to share two numbers (viewable impressions net general invalid traffic and viewable impressions net sophisticated invalid traffic) – if they’re able to do sophisticated invalid traffic detection, that is.

General invalid traffic detection is not all that difficult to do, so, in a sense, removing it from the equation separates the men from the boys because any vendor worth its salt should be able to filter general invalid traffic without a problem.

From the MRC’s perspective, that should help brands, agencies and publishers select the right viewability vendor to work with, said George Ivie, CEO and executive director of the MRC.

“No matter who someone picks for any measurement metric in digital, they’ll be getting a vendor that uses general techniques – that’ll just be a given – so now that’s not something they have to worry about,” Ivie said. “If you want to find a vendor that uses sophisticated techniques, and we recommend you do, the field will be narrowed. There simply won’t be as many of those.”

In other words, buyers and sellers can be confident that there won’t be a notable difference in the general invalid traffic metrics reported by different vendors on the same traffic, and if they want someone who can do something more sophisticated, they can seek it out.

It’s easy to see how this will be a boon for vendors who get accredited for sophisticated detection. Why go with someone that only sells vanilla ice cream when there’s a Baskin Robbins around the corner?

GroupM, for one, certainly sees it that way.

“Our position is that we don’t work with a vendor that can only do general invalid traffic detection,” said Joe Barone, managing partner of digital ad ops at GroupM and a member of the MRC’s Digital Committee. “We require vendors to scrub the sophisticated invalid traffic from the viewability count, and for our publishers to accept payment on viewability, they understand that we’ve scrubbed all fraud from those numbers.”

appleorangesValid Concerns?

But none of that does away with the discrepancy problem.

General invalid traffic isn’t a significant source of discrepancies, said Hahn, so separating it out isn’t actually going to do much to help clients reconcile the differences between sophisticated vendors.

That’s because each vendor uses its own proprietary methods to filter sophisticated invalid traffic, and rather than evaluating them against each other – which the MRC can’t do because they really are very different beasts – the MRC judges the sophisticated guys on whether they’re capable of doing what they promise to be able to do, not on how they do it.

It’s a bit of a Catch-22: In order for there to be no discrepancies between vendors, each would have to use the same methodology for sophisticated invalid traffic detection. But if all vendors used the same methodology for sophisticated invalid traffic detection, they wouldn’t have a business model – and it would be easier for the bad actors to get ahead of the game.

As it stands, two different vendors can both be classified as sophisticated and have very different ways of doing things.

“Our concern is that we exist in a very complicated industry as it is,” Hahn said. “If you ask any advertiser out there if they care whether an ad was detected by a general or a sophisticated system or if they care if the fraud count is separated from the viewability count, the answer is probably going to be, ‘No.’ They just care about the opportunity for their ad to be seen by a human and we can make them happy by giving them one number.”

GroupM’s Barone can relate to both sides of the debate.

“Vendors are not pleased because they feel like this introduces complexity, while on the other hand, the MRC is actually helping us evaluate vendors – both of those things are absolutely true,” he said. “And the complexity will always be a concern here because there’s so much data available.”

That said, Barone is in favor of having separate reporting for general and sophisticated invalid traffic because it gives GroupM “the ability to compare vendors and inventory sources more deeply and to understand the nature of the issues we’re seeing.”

GroupM also doesn’t shy away from data. The ecosystem is complex, but data scientists can handle it.

“Have a look at the face of an analytics person who opens up a massive spreadsheet packed with data and columns – to them it’s like Christmas morning,” Barone joked. “But, really, the more data they have available to run performance against, the more insight they have to see what’s working and what’s not.”

But Barone acknowledged that discrepancies generate confusion, and confusion makes billing into a bit of a nightmare.

“It’s not an easy thing to look at the numbers and figure out who’s the best – you can’t just assume that whoever reports the most fraud did the best job,” he said. “That’s why GroupM does extensive in-market testing of vendors.”

That isn’t, however, much help to the average media planner just trying to do her job.

“We’re already inundated with data and different reports and overwhelmed with all of the various parties in the digital transaction, so the thing we should be trying to do as an industry is to simplify things,” Hahn said. “If it’s difficult for a Ph.D. to deconstruct a spreadsheet to see what it really means, what about everybody else?”

OK, but you have to start somewhere, said Sherrill Mane, SVP of research, analytics and measurement at the Interactive Advertising Bureau.

“The detection and filtering out of fraudulent activity is of paramount importance to all measurement of consumer activity in digital media – it doesn’t matter if it’s viewability or another metric,” Mane said. “If it were up to me, everybody will ultimately be able to do sophisticated, but you have to begin with the basics that every measurement system can live up to.”


Add a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>