If your agency is recommending programmatic ad buying, change agencies

  • According to research conducted by a professor at MIT programmatic ad buys may not just be unproductive, it may be counterproductive.
  • 3rd party data may be useless at the most elementary of targeting tasks.
  • Using data acquired from 3rd party brokers improved targeting performance by 184%,but creates extra costs of about 238% on average in comparison to random placements.
  • Adobe inspected traffic across thousands of its client sites and found that 28% of the traffic showed “non-human signals” indicating that it was fraudulent.

As pharma marketers move more money to online, hopefully, they are going to be running into major a major problem: ad fraud and poor targeting.

According to Forbes “fraudsters attacking the ad tech ecosystem are skilled cybercriminals who, after they attack, are continuously monitoring to see where they have succeeded and where they have failed. When fraudulent traffic is blocked, following the same rules consistently, the attacker can identify which parameters trigger an anti-fraud system and automatically avoid them so that the fake traffic can sail in undetected”.  This means they are one step ahead of measurement people and they are succeeding.

Then there is the MIT study.

“Study One finds that using buying platforms (DSPs) that leverage audiences provided by data brokers perform poorly in terms of identifying consumers who have a certain set of demographic characteristics.

Study Two finds that [inlinetweet prefix=”” tweeter=”” suffix=””]when asked to measure the characteristics of consumers arriving at a certain website, data brokers show high inconsistencies and perform poorly again in correct audience identifications.[/inlinetweet]

Study Three finds that the[inlinetweet prefix=”” tweeter=”” suffix=””] actual underlying audience information on individual users appears to be flawed, with inconsistent and poor classification results across 14 leading data brokers. Specifically, we find that on average 62.5% of user attributes are wrong for our data…”[/inlinetweet]


I analyzed a big ad buy from a client and I noticed something odd right away.  A lot of traffic was coming between the hours of 2-4AM.  How could that be?  When I dove deeper I found that the traffic was fraudulent.  They were either BOTS or took me to bogus sites.  Of their $500,000 online ad buy I estimated $250,000 was totally worthless as the bounce rates for these ads was close to 90%.

[pullquote]Bots account for 56% of overall website traffic and 25% of publishers have no way to detect non-human traffic. 72% of the loss associated with the web’s fraudulent traffic happens on desktops and 28 percent on mobile.[/pullquote]

How to get around most of this?

Have your agency do the work they were hired to do.  This means aligning your online ads to your audience via information from the publishers rather than using programmatic buys. If your agency is recommending programmatic, find another agency.