Matt's Ten Campaign Data Rules

Here are TEN rules that you should abide by when doing campaign data work:

1. An approximate answer to a precise question is better than a precise answer to the wrong question. -  Make sure your data is answering the question and don't be afraid to report on not having the data sufficient for anything other than an approximate answer.

    

2. Simple is better than complex, complex is better than complicated - The more complicated the model, the more energy you and your team will need to understand and use the model. Expect members of your team to get lost (for good).

    

3. Start with stupid models. Then, try a Generalized Linear Model (linear or logistic regression). If the next model you try can't outperform a GLM, reconsider if the time investment is worth it.

    

4. However smart you think your stakeholders are, pretend they are only half as smart. Dumb down the message to it's kernel - If you're dealing directly with the candidate, explain how it leads to a win.

    

5. Your job is not to be a scientist, it is to use scientific approaches to guide decisions - campaign wins are measured in days, if not hours. If you're building the main model while the campaign is underway - you're essentially building a model for the next election using old data. Prepare your data and test/train your models well ahead of time. 

    

6. Clean, stored, secure and relevant data will always lose to the incumbent with a well-stocked address book -  Never underestimate the power of human interactions to sway donations and votes.

    

7. Bots don't do favours - It takes a human to understand human concerns. Capture, analyze and act on the concerns so that the candidate can deliver the result personally.

    

8. Fancy dashboards don't win campaigns - Always remember that the goal is votes and donations. Dashboard the targeting/acceleration of each. Anything extra is extra.
   
   

9. You're not the website developer.
   
   

10. Automate the dailies well ahead so you can focus on what matters - If you're spending most of your time cleaning data, inputting contacts, analyzing unsubscribes, etc, you're not guiding decisions. Automate the boring stuff well before the campaign starts so you can focus on data-driven decisions.