What are some good "beginner level" data modeling/analytics approaches to kick start a data science/analytics team?

What type of voter data did Obama's election team start out with, and exactly what data did volunteers collect, and how did they do it?

• In 2008, I understand that there was a problem in the campaign: they had collected data on millions of people — from online, from donations, and from in-person volunteering — but they realised, before it was too late, that they had no way of combining the data. This was solved in 2012. Did the Obama campaign, however, start off with other data sources on voters? Did they purchase this data from polling companies and other sources? If so, what data, and from whom? There are many, many articles on the data management and the technological side of it, but I want to know what data the volunteers canvassing — those calling door-to-door — collected. The whole process of walking vote sheets was done through iOS/Android (or web apps), I know, but what exactly did they ask each voter? Did they ask their names? Ages? Political party? Who they would vote for, etc.? It's difficult to visualise how much the campaign knew without knowing what data they collected: yes, a lot of it was online, but there is a huge chunk of middle class voters who voted Obama, and aren't active on social media. Just how was the in-person data collected on them?

• Answer:

  [note: I am going to tread a little carefully here. I was a staff member of both the 2008 and 2012 campaigns. As a low-level field person, I never knew much that could be considered secret, but I'm going to stick to things that are common to all campaigns, or are already common knowledge about the Obama campaigns.] There was indeed a much better system for integrating data from other sources (fundraising, etc.) with field data in the 2012 campaign than in the 2008 campaign. But, as you suggest, that certainly does not mean that there wasn't a lot of data from other sources. (Some of the hagiographies of the 2012 tech department make it sound as if we were banging on doors at random before Narwhal. This is not the case.) The most important source of data is the voter rolls, which are public domain. Voter contact, unsurprisingly, targets registered voters. For reasons of transparency, a great deal of information about voters and voting is available to anyone who wants it. You asked if canvassers collected data about voters' names, ages, and party affiliation. There was no need -- everyone already knows who is registered at what address, how old they are, and, in states where it is relevant, what party they are registered as. And that's not all: you can look up any registered voter and see in which elections, general and primary, they voted. (The details may vary from state-to-state.) Of course, that data is very far from perfect. When people move, reregistering at their new addresses is not always their top priority. Which addresses are bogus is thus a very good piece of data to collect. You can also track which addresses are simply inaccessible to someone trying to go around knocking doors. That information is supplemented with phone numbers and demographic data from other sources. I don't know much of anything about the specifics, and I wouldn't share them if I did. Although, come to think of it, the vendors at least are probably publicly disclosed somewhere. As to the data collected by canvassers, I will not go into any specifics that might constitute privileged information, but the bulk of it is common to any campaign: canvassers ask people if they are planning to vote and for whom they intend to vote. They try to persuade undecided voters to vote for their candidates. They try to get supporters to volunteer. Sometimes they ask questions about what issues the voter cares about. Occasionally there is other, less obvious stuff. The answers to all these questions are stored, and that data is incredibly valuable. (A final note. Only a tiny percentage of canvassing was done with the aid of the mobile devices mentioned in the details. And news reports suggesting that volunteers were wandering around with private information about everyone in the state/country in the palms of their hands were, to use a technical term, wrong.)