Ben Thompson Teaches Strategy Comms

I am constantly telling candidates to use the rule of three AND to summarize. Listening to Ben Thompson do the same weekly, I wanted to grab one of his conclusions as an example. Notice First, Second and Third. The takeaways are clear BEFORE he dives into the details and he ends on the conclusions, leaving you focused on his takeaways. You should use the same during your interviews.

There are a few big picture takeaways from this reality.

First, Google is undergoing the same transformation. On the product side search is increasingly driven by probabilistic models built from machine learning, instead of deterministic models driven by explicit data; on the advertising side Google spent a lot of time on yesterday’s earnings call discussing its new ad products that were built on probabilistic models that required advertisers to effectively trust Google to figure it out.

Second, assuming these new ad approaches work, the less insight advertisers will have into why or how, and the less ability third-party measurement companies will have to verify Google or Meta’s data. On one hand this may make advertisers reticent to pay for Google or Facebook inventory; on the other hand, the more they come to rely on lift measurements (i.e. when I paid for ads I had more sales, and when I stopped they fell), the more leeway Google and Meta will have to capture margin.

Third, these AI investments are extremely expensive. Both Google and Meta continue to announce significant increases in capital expenditures driven largely by server costs. This is almost completely driven by these investments in machine learning. Narrowly speaking this is a bad thing because their fully weighted cost of goods sold is increasing. Strategically, though, this is deepening the duopoly’s moat: the more elusive explicit data becomes, thus making deterministic advertising less viable, the more important it is to find patterns in the detritus that remains, and so build better probabilistic models; this capability will be a function of the engineers you can hire and the infrastructure you can afford to build, which is to say that the biggest players will have the advantage.

From: Meta Earnings, Meta Spending, AI Costs and Moats April 28, 2022