Four Things You Can Do to Ensure Attribution Data Integrity and Prevent Ad Fraud

By Michael Kennedy, Director of Product Management, Veritone &
Brian Kim, Manager of Analytics, Veritone One

Enter “ad fraud” in your favorite search engine, and you’re hit with an almost endless list of results. That’s because ad fraud — the act of falsely portraying the number of online clicks, impressions, conversions, or other data to boost results — is a costly problem for digital advertisers, and by extension, you (as the outlet that runs the ads).

Digital ad coordinator Pocketmath put it this way in a 2017 article:

“Ad fraud scams advertisers by creating fake traffic, leads and ad placement. Advertisers are made to pay for exposure that never happened, or if it did, it failed to reach the intended audience. In short, advertisers lose money by paying for ghost ads.”

Unfortunately, there are bad actors out there who create fabricated, artificial traffic to make advertising outcomes look better than they are (and this ad fraud can happen within non-digital attribution models as well as digital).

To quote Bob McCurdy, VP of Corporate Sales, Beasley Media Group, in his recent post: ‘There is No Going Back,’ “It’s not like digital metrics are flawless. In mobile marketing, ad channels are free to claim a click on an ad without informing attribution partners of a prior impression. This creates a real issue where many cases of fraudulent clicks are delivered which could easily be validated with impression data attached. Bots/IVT, etc. are not perfect but digital still growing 20%-30% annually. Looks pretty attractive.

There’s more digital ad fraud in this country than there is total radio ad spend. The MRC approved counting a display ad as “delivered” if 50% of it is visible for 1 second. For video it is 50% of it visible for 2 second. And we have an issue with radio analytic attribution?”

This corruption and manipulation of data by falsely portraying the number of online clicks, impressions, conversions, or other data is a problem for you because it compromises the purity of the information you supply to your advertisers, which can degrade their trust. What’s more, the ability to refine a campaign depends on knowing what’s working and what isn’t. Bad data masks problems and makes it harder to develop long-term, successful campaigns. Those factors threaten both your reputation and your revenue.

That means you have to protect the integrity of your ad data.

To quote Bob McCurdy, one more time, It’s long past time to do and try something different with the understanding that very few if any sets of data are perfect, but in the hands of professionals, valuable insights can still be extracted from them.”  

Here are four things to ask yourself when it comes to maintaining data integrity, especially if you’re in the market for a broadcast attribution solution.


1. Does the attribution solution integrate with playout solutions, or does it require you to upload log files for each campaign?
This is important because integration means ad metadata gets ingested automatically instead of being logged manually and pushed through CSV files that can be altered. An integrated system not only eliminates manual processes, but importantly, it means the data can’t be manipulated.

2. Do you keep an eye on your results to ensure analytics aren’t skewed?
Looking for the telltale signs of fraud can help you spot when something is amiss and alert you to the actions of unscrupulous vendors or even internal bad actors like, say, overzealous account executives. It’s a good idea to monitor results and make sure they’re in line with historic benchmarks and pass the “reasonableness test”. Take a look at two areas:

First, evaluate your web analytics program for:

  • Conversion rates — If there is fraudulent activity going on, then you’ll typically see a huge dip in conversion rate. That’s because, even though there might be more leads, the quality of the leads is poor hence the lack of conversion
  • Length of visits — Another stat to evaluate is time on site to be sure visitors are spending at least the average amount of time you would expect. An influx in shorter visits should be a red flag.

Second, evaluate your broadcast attribution analytics program for:

  • Traffic metrics — Does the number of unique visitors or sessions reported by the broadcast attribution program exceed those same metrics in your website analytics program? That is an impossibility and a telltale sign of reporting fraud and/or a gravely flawed sampling model.

3. Do you have a system in place for blacklisting certain IP addresses?
Visits to an advertiser’s site from internal or related sources skew results. It could be a completely innocent interaction, such as when one of your salespeople or the advertiser’s webmaster checks a link, but it could also be that someone on the inside is trying to increase response to make the results look better than they are. (Remember that overzealous account executive we mentioned earlier?) Regardless, the result is the same: Those visits get credit when they shouldn’t, and advertisers end up paying for it, or receiving misrepresentative data.Fortunately, it’s easy for advertisers to take preemptive steps within Google Analytics, and potentially other tools, to keep unwanted traffic from internal parties and others — such as ad agencies or the broadcasters they work with (you) — from convoluting or confusing the data. It’s a good idea to instruct your advertisers to block related IP addresses to prevent those users from trying to manipulate campaigns.

4. Do you and your salespeople know what to do when a campaign isn’t working?
Suppose you’ve done all of the above, and after eliminating ad fraud as the cause, the data indicates that the campaign still isn’t working. Then what? Hint: The answer is NOT to go rogue and manipulate results! Quick-stacking the ads or otherwise trying to generate clicks or site visits just to artificially inflate the data logs is short-sighted and a bad idea. Instead, use the data to strengthen rapport with the advertiser and start a conversation about how to make things work.

We (the authors of this blog, Michael Kennedy and Brian Kim) were talking to our President of Veritone, Ryan Steelberg, the other day and took note of something he said: “The criticality of data attribution will only intensify as the media industry continues to evolve  — and this includes radio, tv, podcasts and other mediums sharing content — for both digital and non-digital advertising campaigns,”

Ryan continued, “The consumer of those analytics may not always be the same. As Bob McCurdy noted, not all advertisers care to know the analytics. Broadcasters on the other hand, have a vested interest in understanding the results of an advertising campaign and should ultimately want to share that information whether the results are good or bad with their advertisers so that they can work together to determine how to yield optimal performance.”  

Optimization is a standard part of any campaign. Understanding where things are underperforming gives you a chance to recalibrate, replace copy, and tailor the campaign based on those insights.

With this deliberate approach, it’s truly a win-win for all interested parties: advertisers, those promoting the ads, and consumers.