How we can start to tackle ad fraud: more value and new algorithms

Tackling ad fraud: more value and new algorithms

We must place more value on the signals of buying and move away from targeting algorithms that are no longer fit for purpose.



It was almost 30 years ago the first online transactions were made, although there are different claimants for that particular honour including pizza, weed, a CD or computer parts.

From the mid-90s, brands began to put up websites on the new medium.  Ads soon followed; the first generally being credited to AT&T in 1994.

While online audiences were small in those early days, few doubted the Internet was destined for great things. The early evangelists and advocates went on to build sites for socialising, buying, reading, viewing, listening, and some of them built the first agency services teams to help brands go digital.

But within a very short space of time, less reputable characters began to show up at the party. They weren’t interested in such matters as delivering a good consumer experience or building a brand’s reputation. They were interested in ripping off advertisers for as much cash as they could get their hands on. And within a very short time after that first ad appeared, there was a lot of cash to grab.

Today, estimates put ad fraud of various kinds at approximately $68bn globally this year, although just last week the ANA in the US put the number at almost double that, at $120bn. Whichever, this represents an enormous drag on the sector’s performance. Fraud steals money from legitimate publishers and weakens the brands who pay the bills.

As the size and scope of the fraud problem and all of its manifestations became increasingly apparent, the industry began to build counter-strategies to detect and block, or deter, fraud.

All of these were, and are, important. But let’s be frank. Fraud is extremely lucrative to the bad guys; they will always have a strong incentive to stay ahead of the industry’s attempts to defend itself, and the industry’s attempts to act cohesively have been (as is usually the way of these things) underwhelming.

We need new algorithms

Two factors led those responsible for spending advertising money to unwittingly feed the fraud beast: planning based on reach and chasing cheap CPMs. When a planner aims for high efficiency, as many impressions as their budget can buy, netting the lowest CPM, the internet is happy to oblige. Compared to offline media, online offers massive audiences and, some cases, prices that seem too good to be true.

They seem that way because they aren’t true. Some CPMs are that low because the audience isn’t human: the audience is software and devices that consume media and click without end.

So a perfect storm is created: the planning approach combined with algorithms relentlessly chasing “efficiency” delivers ad budgets in part to audiences of bots.

As long as we chase those signals, the situation will not get better. We feed the beast we want to defeat.

We must change course.  We must place more value on the signals of buying, not just reading/viewing/clicking. This means shifting to more emphasis on premium inventory with higher CPMs. It also means being smarter about measuring impact, and not impact in terms of clicks, but impact in terms of things only humans do, like buying products and services, both on and offline.

That may seem obvious, but the fact is it’s a lot easier said than done.

We need targeting algorithms that can estimate real business lift across web properties. These must be integrated with offline media effects, and other drivers of business, in order to get attribution right. We have to get away from what is in many cases the false promise of last-click attribution and focus on what combination of commercial activities works best.

Be prepared to pay more, but get more in return

Ultimately this means getting into the best mix of advertising, sponsorship, point of sale, events and the rest.

For now, and as a start, it means recognising that action online is impacted by activities elsewhere. And ‘elsewhere’ includes offline media channels as well as online.

On the web, planning for reach and impression efficiency is not sufficient to deliver real world results for brands.  In fact, they play right into the hands of the bad guys.

But will planners and brand managers be willing to pay higher CPMs to avoid bots?  In our view, yes, if we can assure targeting that compensates for higher CPMs with better results. Win-win.

In our view, repurposing what many call marketing mix models is a path of great promise. Such models, if well designed, can do the job of estimating real lift opportunity, both on and offline; but they must be adapted for use in planning and buying digital media.

The marketing community has the capability to build these models and adapt them for this purpose. To do so, the support of advertisers is required, which will come with a greater understanding of the issues and how to address them. It is strongly in their interest to make this shift: better business results for brands and a setback for fraudsters.

Our aim should be to take money out of the hands of the crooks who would debase the promise of digital marketing and put it to work building brands and sales and value for real people.

Let’s move away from planning approaches and targeting algorithms that are no longer fit for purpose. And let’s reinvigorate the potential the web offers brands and legitimate publishers to grow and prosper.

Brian Jacobs is founding partner of Crater Lake & Co. He has spent over 35 years in advertising, media and research agencies including spells at Leo Burnett (UK, EMEA, international media director), Carat International (managing director), Universal McCann (EMEA director) and Millward Brown (EVP, global media). In 2006 he started Brian Jacobs and Associates and has co-authored two books on the media industry. 

Media Jobs