When programmatic advertising doesn’t work

Google has encountered a serious problem in the United Kingdom after the government and some large advertisers said they would stop placing institutional advertising on YouTube due to the appearance of government announcements on video pages with content qualifying as incitement to hate, racist, anti-Semitic, extremist or simply bad taste.

The problem, which has forced the company to promise more stringent controls in the labeling of such content, makes evident the shortcomings of the advertising model that is growing fastest on the internet: programmatic advertising or real-time bidding, whereby advertising is bought and sold on a per-impression basis via programmatic instantaneous auction, following a double auction model similar to the one in place in most financial markets.

The idea of getting the right ad to the right person and at the right moment, managed through a system that allows the most interested to offer the highest price, faces many problems in its application. One of them is that content cannot be defined solely by the audiences that it attracts, and much less so if those audiences are based on a few sociodemographic variables. Reality, as anybody knowns, is much more complex than a few variables.

In the case of YouTube, it is clear that Google has a problem: with around three hundred hours of video published every minute, manual supervision is impossible: an estimated fifty thousand people would need to watch video constantly for eight hours a day to do so. Instead, YouTube employs a system of alerts based on tools for the user community: when several people mark a video as offensive, it goes to a manual monitoring system that could lead to various measures, from being removed, to being excluded from the advertising system, local locks, exclusion from certain audiences, or other possibilities.

A government or company cannot evade responsibility by simply saying: “I did not know” or “I have no idea where my advertising comes from,” because although there is no reason placing a particular advertisement means agreeing with the content of a page, the issue goes beyond aesthetics: the advertising impression contributes to funding the content creator. Hence the concern that public money or private profit is funding racism, xenophobia, hate, terrorism or anti-Semitism.

On the other hand, qualifying content is far from the only problem facing programmed advertising. As it spreads, an ecosystem is emerging in which there is a feeling that “everything is a lie”: audiences generated by bots, increasing amounts of click bait, simulated or non-existent video reproductions, falsified analytics … some time ago, I wrote about the need for the advertising industry to reset: in a very short time, the internet has gone from offering the promise of a perfect medium for advertisers to reach the people they are most interested in to becoming a kind of chamber of mirrors in which traffic is made, bought, and sold and even faked.

Once again we see an attractive ecosystem that has experienced strong growth, where the vultures are waiting to exploit it, spoil it, to cheat the unwary and to destroy its value proposition. Same old story.

(En espa?ol, aquí)