Victory Lab by Sasha Issenberg
What happens when our professions finally tests itself
Masters in Public Affairs goes back to the foundational books in this field and extracts the principles that the best practitioners return to again and again. Join the growing number of subscribers on YouTube, Spotify, and Apple.
This episode’s book is one that had a significant impact on my approach to campaigns. While it was written years after I led already led several political digital campaigns, it nonetheless validated my approach to constant experimentation and innovation, which I first learned while volunteering on the 2008 Obama campaign. Of course, that campaign gets a lot of coverage in The Victory Lab, and its lesson hold true nearly two decades later.
“Most things are done with only one check. People’s guts.” — Steve Rosenthal, quoted in The Victory Lab
Those of us in public affairs explain things for a living. We are good at the story after the fact: why the campaign worked, why the policy landed, why the vote broke the way it did. What we rarely do is run the test beforehand that would tell us whether the story is true in the first instance. And there is a comfortable reason we don’t, one that Sasha Issenberg’s The Victory Lab surfaces at length: work that is never measured can never be shown not to work.
If you heard our last episode, on Joseph Napolitan’s The Election Game, the contrast will be obvious. Napolitan won by instinct, out in the field, a generation before the academic research caught up to what he already knew in his gut. The Victory Lab is the story of that research catching up, and then, in a few places, overturning the instinct.
The book runs in a few different directions: the data, the targeting, the operatives, decades of turf wars between consultants. However, the single biggest takeaway I got from this book, and have applied often since, came from medicine: the randomized field experiment.
In 1998, two Yale political scientists, Alan Gerber and Donald Green took a large group of voters, split it at random into a contacted group and a comparison group that got nothing, waited for the state’s record of who voted, and counted the difference. In other words, a clinical trial, with the voter standing in for the patient.
The results were extremely eye opening:
Phone calls, the workhorse every campaign spent fortunes on, moved turnout by essentially nothing.
The mail nudged it a sliver.
The oldest, least impressive tool in the kit worked: a person on the doorstep making the ask in person. That moved turnout about seven points, more than the winning margin in most races that matter.
The most expensive, most modern tactics were the weakest, and the thing that moved people was presence: one person asking another. Three decades of polished broadcast had gone into replacing the very thing that worked best. And the method had a pull of its own. You can only cleanly isolate a one-to-one contact — you can’t show a TV ad to one neighbour and hide it from the family next door — so once the field got serious about measurement, it was dragged back toward the individual human being. The tool chose the subject.
Then there’s the experiment that is often cited in our world, but worth revisiting: A Michigan consultant named Mark Grebner, working with Gerber and Green, mailed people a letter that printed their own voting record next to their neighbours’, and promised an updated copy after the election so the whole street would see who had shown up.
Turnout in that group jumped to nearly 38 per cent, against 30 for the group left alone: about three times the lift of any mailer they had tested, at roughly two dollars a vote.
You may already know why it works. It is social proof, the mechanism Robert Cialdini lays out in Influence: our pull toward whatever the people around us are seen to be doing. But here’s the catch: Voters found the letter menacing, and no candidate would put their name on it. The blowback was way too intense for any political campaign to stomach. And so, the most effective tool anyone had measured is one nobody can use. How, Donald Green asked, do you take that static electricity and turn it into useful current? You take it down a notch, and stick to the principles, of course. Which is always easier said than done.
The book is more than a decade old now, and it shows. The firms and tools it celebrates have mostly been overtaken, and a reader in 2026 can’t help measuring its sunny middle chapters against what micro-targeting turned into after 2016 (the book has no idea what’s coming, which gives it a strange innocence in hindsight).
It is also a long parade of American campaign operatives whose names blur together by the back half, and the rivalries between data shops get deep into the weeds. None of that is a reason to skip it. The core idea — test what you believe, or admit you’re guessing — holds as well for a regulated-industry coalition in Ottawa as for a turnout drive in New Haven.
Why This Book Matters
Any field that runs on inherited belief instead of evidence pays a tax, in money spent on things that don’t work. The campaign that books the phone bank every cycle because campaigns book phone banks is paying it; so is the trade association that flies its members to the capital each spring because it did last spring. (I’ve helped plan a few of those trips. The cost is real, and invisible, because nobody runs the comparison group that would expose it.)
The fix is almost embarrassingly cheap. We report motion — meetings taken, op-eds placed, alerts sent — and call it results, when the job was to move a number. So next time you run an advocacy push, hold back a random slice of your list, send them nothing, and compare. That one move separates a team that can report its activity from one that can prove its effect. The catch is temperament — we’d rather not find out our best tactic does nothing.
This one cost me something. On a provincial campaign I worked on in 2011, we tried things that were new for a Canadian race, much of it drawn from my time on the 2008 Obama campaign. We lost, so we never sat down afterward to ask what had worked. Plenty of it had. But to the victor go the spoils, and the losing side’s innovations get thrown out with everything else. Issenberg describes a literal dumpster where a campaign’s hard-won data ends up. It’s real. I’ve watched it get filled.
What We Cover in the Episode
The New Haven experiment, and why it was the first field experiment of its kind in 70 years
The three mechanisms that make the book work: the experiment itself, micro-targeting, and turnout as a measurable craft
Hal Malchow’s polling flip — more about who voters are, less about what they think — and Alex Gage’s “consideration score”
The neighbours letter, social proof, and the line between what works and what you can decently use
Three mental models: the Lore Tax, the Untestable Channel, and the Margin Game
Four ways the book gets misread, and four ways to use it in public affairs
The mastery lesson: why the best operators are the ones willing to be proven wrong
Listen and subscribe here:
