What Happened is recommended reading for anyone who hopes to understand the 2016 U.S. Presidential election. In this review I will focus on one particular aspect of the story: Hillary Clinton’s description of her data and analytics team. This is a small part of the overall election but an important one to understand for anyone interested in technology and our statistical understanding of the world.

Surprisingly, the book contains only four mentions of the Clinton campaign’s data team. First, it describes how the leaders of the team were hired based on advice from Silicon Valley CEO’s (p. 70). Shortly thereafter, Clinton rejects the idea that her team relied too heavily on big data, suggesting that going forward Democrats need even better data and analysis (p. 75-76). It is possible to agree with this conclusion without accepting the premise; in the future data analysis will likely constitute even greater portions of campaign spending, but it is still possible that the Clinton campaign was too focused on polling than on their “ground game.”

One particularly revealing sentence in the book is that, as late as election night, “all our models… gave us an excellent chance at victory” (p. 384-5). This suggests that the analytics team’s approach was fundamentally flawed. Whether this is because poll respondents were less willing to express a preference for Trump, or because sampling mechanisms themselves were biased, or some other reason, this is a major oversight on behalf of the Clinton campaign and a cautionary tale for future election analysts.

The final mention of the campaign’s data team is a claim that they were “outgunned” by spending on the Republican side (p. 421-22). But that seems to disagree with many other facts. For example, the Trump campaign paid Cambridge Analytica less than $6 million out of a total campaign budget of $322 million. By contrast, the Clinton campaign’s total spending amounted to $565 million. Presumably they would have been willing to spend another 1 percent in order to increase their odds of winning.

Furthermore, recruiting talent matters mroe than overall spending. By her own admission Hillary had a large, technically savvy staff in her Brooklyn campaign headquarters. Anecdotally, it also seems that the most talented data scientists and software engineers would be far more willing to work for Democratic than Republican campaigns.

This abdication of responsibility fits with the larger context of the book. For example, there are numerous mentions of Clinton’s victory in the popular vote. While it’s true that a sizable majority of Americans who went to the polls in 2016 wanted her to be president, that’s not the contest she signed up for–and she knew that better than anyone.

Two of the other most significant themes of the campaign–Russian hacking and her emails–each got their own chapter. But at the time of the campaign, the composition and performance of her data team was much more directly within her control. This lack of attention suggests there is more to learn before 2020.