Who Is interested in Fraud Detection? Officially – Everybody, but…
According to different studies, done both by independent analysts as well as such respectful organizations like IAB amount of fraudulent impressions, clicks or attribution can get up to 20-30% of all online ad inventory.
Final figure depends on many factors – country, industry, ad-server, site structure and dynamics etc. But still fraud remains one of the biggest problems in the current advertising environment.
And this problem isn’t going down in the recent years. How come? Are technologies not powerful enough to deal with fraud? Is it really a problem of smart data scientists cracking this nut?
Well, let’s ask one very old question – Cui bono? Who particularly is interested in finding frauds?
Ad-servers? From one hand – yes, as it should provide better value for their customers. At the same time if it’s indeed 20% of all impressions – it means that ad-servers would essentially get 1/5 less money for giving better service. And in our competing economy it’s doubtful that they can address it by increasing costs.
Publishers? Same story – giving “clear” inventory is important for their reputation so they’d never approve fraud. At the same time if publisher size and power is evaluated based on amount of annual impressions it’s not in their interests to actively fight it.
Advertisers? Well, now we’re talking. They’re interested to pay for actual value, be it clicks or audience. But what means they have to control fraud? Shall they buy their own data mining solution and analyze historical data in hope to avoid bots, click farms, iframe stuffing and all other fraud-related issues? Or hire independent company and insist on putting their own pixels to be able to independently track and compare results, not relying on ad-servers data? Is it really what they want to do with their business or is it just enforced activity to manage costs and efficiency of their advertising budget?
Obviously they’re doing just that even if it’s not their primary area of expertise. And they have contradicting data to publishers and because of that publishers have to reserve some amount of inventory. And publishers are wasting this inventory because of contradictions in data, which in its turn is not necessary because of fraud – but because of different ways how ad-servers track users. So we’re not removing fraud from the equation but now publishers are getting less from their inventory – fraud is plaguing trust between publishers and advertisers.
There are some products which are claiming to solve this issue, both post-fact analysis as well as real-time recommendations. None so far has proved to win in this war with fraud.
Is there a solution? Well, we don’t know the answer. But what is clear is that it cannot be solved without participation of all parties involved, including ad-servers and publishers. They should be eager to share data which is important for analysis. And it’s also associated with costs – as they should be rewarded for such transparency.