Dear Zack and Becky,

I am writing in response to your article, "Hillary Clinton’s Vaunted GOTV Operation May Have Turned Out Trump Voters."  I have several issues with the article, most of them regarding the substance of your argument, but I also take issue with the style of your critique.  I review my concerns here with the hopes that my thoughts help contribute to an improved post-election retrospective.

1.  At the start of your argument, you state that "Anecdotal evidence points to anywhere from five to 25 percent of contacts were inadvertently targeted to Trump supports." Later you say, "Volunteers reported as many as 30% of the replies they received from voters they were urging to get out were Trump supporters."  I do not believe that anecdotal data is sufficient to support your conclusion that micro-targeting should not be used for creating GOTV universes.  As a movement, we take pride in our fact based approach to both policy and also campaigning.  If you don't have the data to back up your accusations, then you shouldn't go public with your critique.

2. Even if, for the sake of argument, we assume that your opening premise is true, the increased contact rate with Trump supporters demonstrates correlation, not causation.  In order to demonstrate that a strategy based in "big organizing" beats out "small organizing" you'd need to present a comparison proving that real IDs beat a model based GOTV universe.  You do not present this comparison.

3.  You should not need to tear down someone else's work in order to build credibility for your own conclusion.  If the "big organizing model" was so successful for you during the primaries, your argument would be more powerful by publishing metrics showing this success.  How good were your IDs at predicting turnout for Senator Sanders?  Even if you didn't use a comparison model at the time, it wouldn't be that hard to ask an agency to run a historical snapshot of your data through their model and compare your IDs to a model.  It's ironic that you cite a "lack of actual data" as the reason for your skepticism of micro-targeting in an article that cites no data to support it's own conclusion.

4.  Given the weight you both carry within the progressive movement, I worry that you've set your philosophy of big organizing, a philosophy that I support, up against the science of data.  Models aren't just used for defining a GOTV universe.  We also depend on that same math to help us listen properly to constituents when we have millions and millions of messages that couldn't possibly be read by a small staff.  Models help us understand what kinds of issues are important to a voter so that we can send a nurse more detailed information about the health care policy she cares about without inundating every single supporter with every detail about everything.  Models help us understand whether a voter might prefer to be contacted via the phone, email or text.  Models help us serve our supporters better.  Don't make math the enemy.

5.  Finally, I take issue with the meanness of this article.  I will always support after-action examination of our work with an eye to determining how we can improve.  However, your thesis did not require you to attack the work of our colleagues and friends, and you chose to do so anyway.  Your critique wasn't data-driven.  And you chose to publish three days after a devastating loss.  I can forgive a lot, but I struggle to forgive unkind.  

No doubt there will be a lot of public and private conversations about the 2016 campaign moving forward. I'd like to remind those of you who read this that truly productive post-mortems are NOT about highlighting failures with the purpose of assigning blame.  They are about reviewing the work and the outcome in service of improvement. The conversation needs to be constructive, not destructive.  And, most importantly, our conclusions should be based in fact.