The networks we cannot see

The thing that wins the next election will not come from consultants like me. It will come from somebody in a network we cannot see.

The arithmetic is worse than the version people repeat at conferences.

NBC's final poll before the 2024 election split respondents by whether they follow politics closely. Among the people who do, Harris won by five, fifty-two to forty-seven. Among the people who do not, Trump won by fourteen, fifty-four to forty. A nineteen-point swing on one variable, and the variable is whether somebody pays attention to us.

Catalist went back through the voter file afterward and found the same shape in the behavior rather than the self-report. Clinton took more than fifty-four percent of people who had voted in two or fewer of the last four elections. So did Biden. Harris took about forty-eight. Against Biden she lost five points among people who do not vote in every election, and nine among Latino voters, and six among voters under thirty.

Meanwhile the electorate itself got more devoted. Nearly half of everybody who voted in 2024 had voted in all four recent federal elections, a share nine points higher than in 2020.

We are winning the room we are already standing in, and the room is getting smaller relative to the country.

Democratic campaigns and their allies spent four and a half billion dollars on advertising in that election. A billion more than the other side. They lost every swing state.

So what have we proposed doing about it?

More sophisticated ways of getting at people who want nothing to do with us. Better targeting. More texts, sent at better hours. Different creative, tested against four variants. A new CRM, again. Somewhere in this country there is a deck about reaching disengaged voters and every option on it is a way of arriving in their life uninvited, more precisely than last time. Well, in fact, many decks.

The option that is never on the deck is asking the people who are already in those communities. And I am not talking about a focus group, where what people say gets handed back to the same strategists, repackaged, and sold as the same old crap that assumes the answer is some monolith of a demographic rather than the reality that every community network is different from every other one.

I think I know why that option is never on the deck. It would mean the strategist is not the one in charge. Not the one in control. Not the one with the idea.

What that costs is the part we have never properly counted.

We fell in love with social networks because we could see them. A graph. A follower count. A share, a reach figure, a number that moves when you do something. Visible, countable, and therefore, in our world, strategy.

The networks that actually carry a political conversation are invisible to us. The soccer field. The school gate. The WhatsApp group with the other parents from the swim club. The church hall after the service. The kitchen at Christmas, where somebody's brother-in-law says a thing and somebody else decides not to let it go.

We discounted all of that because it does not produce a dashboard. That is a methodological failure and it has been running for thirty years.

I wrote in another book that the TV or digital ad will never be as powerful as the dinner table conversation. I meant it as a claim about persuasion. I think now it is also a claim about access. Those tables are the only route into the households that stopped listening to us, and there is no version of our current practice that gets us a seat at one.

A supporter has that seat. She also has something no file contains.

Not just who she knows. What would land, and what would not be welcome. She knows the church will not take a partisan flyer and will turn out forty people for a food drive. She knows which of her neighbors will hear this from her and from nobody else on earth. She knows what happened at the school board in March that makes a particular argument either obvious or insulting in her town, and she knows it without having to look it up.

That is the half of the work that cannot be modeled. It is also why the same message is a gift in one community and an intrusion two towns over.

Which is why the relational tools we built were never going to be enough. Connect your contacts. Tick the people you know. Here is a suggested message. That extracts the one part of her a database could almost have guessed at, and throws away her judgment, her language, her stories, and her read of the room.

If you want to know how much that costs, the best evidence comes from two researchers who are usually quoted for the opposite purpose.

Joshua Kalla and David Broockman went through forty-nine field experiments and found that the persuasive effect of campaign contact in a general election is approximately zero. Mail, calls, ads, doors, all of it. That finding gets waved around every cycle by people arguing against field programs.

Read further and they name their own exceptions. Contact persuades when voters have no partisan label to lean on, which is primaries and ballot measures. It persuades when a campaign has done the work to find the specific people whose position the message actually meets, which produces between five and nine additional votes per hundred conversations. Broad outreach fails. Knowing exactly who you are talking to, and why, does not.

Then in 2020 those same two researchers evaluated a deep canvass program run by People's Action. Long conversations, unscripted, two ways, where the canvasser tells their own story and then listens to the voter's. Six hundred and ninety-five of them.

It moved presidential vote choice by 3.1 points. Among independents, 6.7. Among independent women, 8.5. Their own summary called it a hundred and two times more effective per person than the average presidential persuasion program.

The two people who proved that campaign contact does not persuade also measured the one thing that does. It was not a media buy, a creative refresh or a targeting model. It was a person bringing their whole self into a real conversation, six hundred and ninety-five times.

The forty-nine experiments do not indict canvassing. They indict scripted canvassing, which is a different thing, and which is what we have mostly been sending people out to do.

I wrote a guide on distributed organizing years ago and one line in it has bothered me ever since, because we still seem to not build anything that acts on it. Supporters who participate bring their full and best human potential to a campaign's strategic capacity. Their creativity, their insight, their enthusiasm, their charisma, their leadership. When you let autonomous groups develop their own strategy, their own content and their own issues, you get value that has nothing to do with money or volunteer hours.

The same guide says something else that belongs in this chapter. A campaign with real constituent participation is harder to attack. When people can see that their neighbors are inside it, the charge that the thing is fake, astroturfed, or run by people who despise them stops landing. Participation is not only reach. It is armor.

The problem is that we do not let people come to the table as themselves. We hand them a walk list, a script and a suggested caption, and then we call what comes back capacity.

None of which is an argument for sending people out with nothing.

I have spent two weeks refusing that version. An untrained volunteer improvising in your name is exactly the disaster the whole sector reorganized itself to avoid, and the fear is legitimate. What I am describing is different. Coached through the method, and trusted with the judgment. The practice is ours to teach, and it is well studied and hard won. The reading of who, and where, and whether this is welcome here belongs to her, because she is the only one holding it.

Harvey Milk understood the second half of that better than most people in this business. He did not launch by announcing that a gay man wanted representation in San Francisco. He got reporters to meet him in a park, stepped in dog shit in front of them, talked at length about the state of the parks, and understood perfectly well that the daily nuisance people actually complained about was the door into everything else he intended to do. Decades later Danica Roem became the first openly transgender legislator in the United States by running on a road, Route 28, because in her district that road was the thing everybody had an opinion about.

Neither of those came out of a message-testing exercise. Both came from somebody who lived there.

Somewhere in a network I cannot see, somebody already knows what would work in their town. They are not going to write it up. Nobody is going to ask them. The reason is not that we lack the budget to find them, because we have four and a half billion dollars and we found the parts of them that fit in a spreadsheet.

The reason is that asking would mean giving up the part of this we like the most, which is being the one with the idea.

What we leave on the table was never only the people who could not participate. It is everybody they are closest to, which is to say the people who do not follow politics, which is to say the people who decide.