As an expert witness that spent years analyzing the validity of claimed display ad traffic, I recently ran a test on Taboola display traffic for a client who was interested in switching from Facebook and Linkedin. The results are in and are shockingly disappointing: most certainly not in Taboola’s favor.
Below is a snapshot of ad behavior that suggests the traffic was completely invalid and should be refunded by the card provider. This is in case the ad network (an Israeli outfit) attempts to persist with the charge after we informed them:
Meaningful engagement: NONE
Bounce Rate (BR): 100%
Average time spent per user (Time): 0.00 seconds
Comparative performance of the same landing page with other channels:
Google PPC: (BR 50%, Time 0:45)
Organic: (BR 30%, Time 0:50)
Email: (BR 30%, Time 1:00)
Expert statement and recommendation to clients:
When testing an ad platform, the above metrics are crucial to measure and compare against other channels. This negates the potential argument by fraudsters or those exploiting SME’s, as well as ad networks, that there is a problem with the website landing page. The behavior of other channels prove that it was not the case and other channels performed well.
The fact that zero seconds was spend on-site, consistently across all visitors from the cohort, raises the following questions on four critical factors we seek to assess:
- Ad Fraud: Can we state that the traffic represents authentic users who intended to visit the website?
- Accidental Clicks: Does it seem like “users” genuinely intended to click on the ad?
- Real human users: Does the behavior of this traffic seem like a human user?
- Overall authenticity: Is there any chance that this traffic was authentic but a mistake was made by the advertiser?
The answer to all questions is NO.
Now it is not clear to us whether the clicks were as a result of ad fraud, or accidental clicks where the ads were placed in a manner that interferes with natural navigation, or whether these were not real human users, or rather something akin to a bot farm purporting to be real American visitors. This is a question Taboola and it’s advertisers can explain better. Yet what is clear, is that no user intended to visit the website and that the authenticity of the visitors score an overwhelming ZERO.
Report recommendations:
To Adam Singolda the founder and CEO at Taboola (NASDAQ: TBLA): Stop ad serving until such time you can prove authenticity to advertisers. (This report was also shared with Lynda Clarizio, a board member at Taboola).
To advertisers: Any advertiser seeing a similar result should raise a dispute with their card provider and bank, presenting their analytics. Stop ads until the problem can be resolved.
To agencies: Regardless of the incentives an agency may receive from Taboola – it is not worth risking customer trust by promoting such questionable traffic. There is a real risk of hurting client trust.
To regulators and banks: Sustaining charges in support of Taboola traffic – where it equates to the above metrics, is unthinkable, especially at a time where SME’s are in an area of low growth.
To investors in Taboola: Should this be part of a wider organizational issue at Taboola, expect to see advertisers and regulators take action.
Note that researchers made efforts to uncover ad fraud for nearly a decade now. In 2019 already this problem was labelled an accountability crisis – due to the fact that ad platforms get away with this. It is partly blamed on poor accounting practices and internal communication breakdowns between the customers’ bank, accountants and CMO team who is supposed to querry invalid traffic and claim approproate refunds.
Additional observations:
Taboola was informed that this report will be published should they attempt a charge. The invalid traffic – in this case 100% of the ads, were reported to them. They sought to ignore this and pass off this traffic as valid and to charge for it. Even though it was clear which account and client they dealt with and that only two campaigns were run for two days, the types of questions asked by their support team seemed aimed at frustrating the process and of poor internal communications, and even account access.
Although Taboola attempts to showcase a long list of credible publishers – in reality, there appears no obvious way to target these publishers – except for by first observing (and paying) for poor publishers and then removing them from a list after paying, which still benefits both Taboola and the poorer publishers, at the expense of the advertiser. How convenient?