In one of our earlier blog articles, we discussed why TV advertising is immune to bot traffic issues that affect many digital platforms. Bots generate fake clicks on websites and never lead to sales (as there is no human behind these clicks), and therefore falsely suggest that a marketing campaign is performing. TV, on the other hand, isn’t impacted by bot traffic for two reasons: (1) bots can’t generate fake TV ad impressions, and (2) they don’t fake traffic on the advertiser’s website at the exact time of a TV ad airing.
That said, to measure TV correctly, we need to have a reliable baseline for each advertiser. If the advertiser’s baseline is affected by bot traffic, we have to implement a robust method of handling this issue to correctly measure TV performance. We will describe our approach using a recent example from a Tatari client, in which we initially noticed an unusual increase in website traffic.
In the first step, we filtered out known data centers and bots. Third-party data providers often have a list of known bots, as well as their associated IP addresses, which facilitates filtering on our side. Interestingly, we still noticed unusual increases in traffic that could not be attributed to TV.
In the second step, we followed several suspicious IP addresses that had not been filtered, and discovered that they were related to a massive ad fraud called Methbot. Methbot is a highly sophisticated, Russia-based bot net center that provides large volumes of low-cost premium video inventory. According to a MarTech article from 2017, Methbot used over 6,000 domains and 250,000 specific URLs to trick advertisers all over the world into thinking they were buying video placements on premium publisher sites. Fake ad impressions were then generated on these domains using over 500,000 IP addresses.
In the third step, and once we had recognized these IP addresses as part of the Methbot operation, we built our own filtering algorithm and based it specifically on Methbot traffic patterns. At the same time, we developed it to be a generic filtering algorithm as well so that any suspicious traffic undetected in the second step would be automatically filtered at this point. This ultimately decreased the client’s registered traffic by 30%, and, in turn, improved our measurement of their TV performance.
It is worth noting that our approach to TV measurement makes the filtering process more robust. Our baseline is dynamic, which means we look at traffic in absence of a particular TV ad, not in absence of TV in general. We can therefore accurately measure the impact of each TV ad on advertiser’s website traffic. As mentioned earlier, website bot traffic is diffuse (it cannot be timed to correspond with timestamps of TV airings), so if we notice spikes in traffic that correspond to TV airings, we know they can be attributed to TV. This is not the case with digital ads because bots are designed to easily trick digital platforms.
So, even though TV itself is not impacted by fraudulent traffic, TV advertisers and agencies still have to be cognizant of the fact that baselines can be affected by fraudulent traffic from digital platforms. If not dealt with properly, this can compromise TV measurement and related optimizations.
