We read every research paper
so your campaign
doesn't have to.

Find the studies — including the ones nobody searches for

Keyword searches find the famous papers. They miss everything else. So we also walked the bibliographies of every paper and then the bibliographies of those — pulling out work that no search term would ever surface.

Throw most of it away

The bar is deliberately brutal. It's not about total number, It's about getting to the most scientifically rigorous and useful papers. To get in, a study has to be a real experiment on a real electorate — randomised, in the field — and it has to report real results, not observations or opinions, but results that were measured and reviewable. 

Take every paper apart

A study is not one number. A single paper might test four different mail pieces on three different kinds of voter, and the headline figure is an average of things that behave nothing alike. Our objective was not 'does mail work' but when does mail work? When?  For what kind of campaign?  And in what type of election cycle? To get there, we break down every individual tactic of every experiment, separately, with the conditions it ran under. The result is an information map indexed on dozens of tactics across hundreds of papers.  

Every arm is tagged the same way: what tactic, which channel, what kind of election, how loud the race was, which voters, how long before election day. That's what makes two studies comparable — or proves that they aren't.

A continuous loop of quality checking 

Tactic names vary across papers. Efficacy indexes differ. And tactics change over time. So every record gets re-read against, arbitraged, and tested. One pass flags inconsistencies. A second tries to prove the first one wrong, A third goes back into the study to judge the conflict. An entire system built to rigorously pull truths out of a mountain of non-standard information. 

What you end up with

Not a library of PDFs but a structured, queryable map of what political science actually knows about persuading and turning out voters — and, just as usefully, where it knows nothing.

Combine it carefully — or refuse to

Pooling results is where most "evidence-based" claims go wrong. We weight each study by how strong it is, how recent it is, and how close its world is to yours — and we refuse to blend results that aren't comparable, even when that leaves a thinner answer. A study of a presidential year doesn't get to speak for your school board race.

A tactic with weak evidence is shown as a tactic with weak evidence. We'd rather tell you the number is shaky than hand you a confident one we made up.

And point it at four questions campaigns actually ask