Thread · 5 tweets · 18 May 2021

Sure — but what the final customer (advertisers here) pays for and how the value gets distributed upstream are two different things. Maybe people will pay CoffeeSpace $5 for a beet frappuccino. But how much did CoffeeSpace pay for frappuccino-grade beets?

If Robin's Chill Beets are the only beets in town (or distinctly better), I'll extract more of that $5. If Ben's Bestest Beets is also in market, whoever wins the tender will extract less and CoffeeHouse will pocket the diff. (In a perfect market it would be different, but…)

The question for advertising is: what goes into a good conversion predictor? If I have a strong conversion predictor (you're currently comparing frappuccino machines on Wirecutter), sharing that away isn't going to increase how much I can sell it for.

That's before considering that it's data, and therefore that I can sell it multiple times in parallel, which will decrease the value I get but increase someone else's ability to arbitrage the value. (What constitutes data monopsony conditions is a fun question too!)

Of course, that's a loss when you have conversion predictors, but if your content is interchangeable and your audience is one-and-dones, then you might benefit from the system. But all that does is incentivise whoopie cushion and content farm "publishers". Should we care?