Political campaigns and NGOs have been trading methods for two decades, and the trades have not been equal. Campaigns took organizing training from movement veterans and a conversation technique from an LGBT community center. NGOs took split testing and the small-dollar ask from campaigns. The transfers that held up kept their method intact. The ones that failed kept the tactic and dropped the constraint, and for an NGO the constraint is trust. A campaign spends its credibility over a few months and a vote settles the account. An NGO has to keep the same credibility for decades, in front of donors and of the communities it says it speaks for.
Test it, and know what the test can tell you
In December 2007 the Obama campaign’s director of analytics, Dan Siroker, ran an experiment on the sign-up page. Four button labels and six pieces of media (three images, three videos) made 24 combinations, each shown to random visitors. The staff liked the videos. Every video lost to every image. Siroker later estimated that the winning version brought in 2.88 million extra email addresses and, at $21 in average giving per address, about $60 million in donations. The dollar figure is his own extrapolation and deserves a discount. The finding does not need it: the people closest to the material picked wrong.
Any communications office can fall into the same trap, wherever the person with the strongest feelings about a video tends to win. NGOs can test cheaply, and mostly should.
The limit is what a test measures. Joshua Kalla and David Broockman pooled 40 earlier field experiments, added nine of their own, and put the best estimate of the effect of campaign contact and advertising on candidate choice in general elections at zero. A sign-up test shows which button gets clicked. It does not show that anyone changed their mind, and an NGO that reports the first as the second misleads itself before it misleads anyone else.
The stronger version of testing goes to the audience that has not made up its mind. By 2009 the US marriage equality movement had lost 30 statewide votes, even though its rights-based ads polled well with viewers (SSIR). Research in Oregon found that undecided voters heard the language of rights as a rewrite of marriage itself, and responded instead to couples talking about love and lifelong commitment. Thalia Zepatos, Freedom to Marry’s director of research and messaging, studied how different audiences thought about marriage and used the results to change how the movement made its case. Ads built on the new language were then tested in real media markets, including a 2012 Maine spot in which four generations of the Gardner family sat around a kitchen table. Pew’s national polling moved from 42 percent support in 2010 to 55 percent in 2015. The SSIR account was written by the head of the foundation that funded the research and a former communications executive there, and it is careful to say that messaging was one factor among several. Hundreds of thousands of volunteers telling personal stories did much of the work.
Discipline is not silence
“Take Back Control” is three words, and voters could hear their own grievance in it. Vote Leave’s campaign director, Dominic Cummings, is credited with the slogan. The same campaign’s bus claimed that Britain sent the EU £350 million a week that could go to the NHS instead. The UK’s official statistician found the claim misleading because it ignored the rebate. Vote Leave never retracted it, and Cummings later said the campaign could not have won without it. NGOs can borrow the slogan’s craft. The bus is a poor bet for an organization whose credibility depends on its numbers surviving a fact check.
The other lesson from the same era runs the opposite way. California’s Proposition 8 passed 52 to 48 in 2008. David Fleischer of the Los Angeles LGBT Center later analyzed the No on 8 campaign’s own daily tracking polls and found that about five percent of voters moved toward the ban in the final six weeks, mostly parents. The push came from a Yes on 8 ad in which a schoolgirl tells her mother she learned she could marry a princess. No on 8 answered with a rebuttal ad that worked, about two weeks too late. One Advocate commentator drew the rule that rebuttals should be built before the attack arrives and used at once, whatever consultants say about message discipline. The distinction is fair. Discipline governs what a campaign says about itself. It does not require silence under attack. The NGO instinct to verify before answering is sound, but it needs a pre-approved holding response, or the other side finishes the story first.
Ask small, ask often, and never trick anyone
The Sanders campaign built its 2016 fundraising on a very small ask. By May 2016, according to its own announcement, it had received more than 7.6 million contributions from more than 2.4 million donors, at an average of about $27. The lesson for NGOs is about frequency and friction: a donor who gave $27 can give again next week.
The bad version of the same idea appeared in 2020. The New York Times reported that the Trump campaign, working with the processor WinRed, made weekly recurring donations the default for online donors, added a second prechecked box that doubled the gift, and made the opt-out text steadily harder to see. Banks were soon fielding fraud complaints from the campaign’s own supporters (Times investigation, reprinted by the East Bay Times). The Trump operation and the Republican Party ended up refunding roughly $122 million, more than ten percent of what was raised through WinRed, according to the Times. NGOs with monthly giving programs feel the same pull. A prechecked box raises income this quarter and produces refund requests next quarter.
A/B testing complicates this, because it optimizes whatever it is pointed at. In Norway, the student aid fund SAIH parodied charity appeals in its 2012 video “Radi-Aid: Africa for Norway,” then began handing out annual Rusty Radiator awards to the worst charity videos. SAIH’s vice president, Martine Jahre, acknowledged that the clichéd imagery had raised money. Jennifer Lentfer of Thousand Currents told Nonprofit Quarterly that “Pity and shame are easy emotional levers to pull.” A test scored on donations will find those levers. The guardrail has to be a rule about dignity, fixed before the test and not put to a vote by it. The awards appear to have worked as a norm: by 2015 SAIH struggled to find a third Rusty nominee because organizations had changed course.
What campaigns took from NGOs
Marshall Ganz spent 16 years on staff at the United Farm Workers before training organizers for unions, nonprofits and campaigns. The model he developed with a Harvard psychology professor for a Sierra Club project became Camp Obama in 2008, which trained about 3,000 organizers. Its core was public narrative: a story of self, a story of us and a story of now, with relationships ranking above platform. What the campaign got was a way to turn volunteers into people who could persuade their neighbors. The ground game that campaign is remembered for was, in origin, movement technology.
Deep canvassing has a similar pedigree and a rougher path. After Proposition 8, Fleischer and volunteers at the Los Angeles LGBT Center spent years trying approaches to reduce prejudice at the door, and by his account almost all of them failed. What remained were long, nonjudgmental conversations in which canvassers ask voters to recall a time they were treated unfairly for being different. A 2014 Science paper appeared to validate the method, then was retracted after Broockman and Kalla found the data had been fabricated by a graduate student, Michael LaCour, who has denied misconduct. Fleischer called every reporter who had covered the study to say the evidence was gone. The real test came in 2016. In Miami, working with the group SAVE, 56 canvassers spoke with 501 voters, and a single ten-minute conversation reduced anti-transgender prejudice, with the effect lasting at least the three months studied. About one voter in ten changed, which the researchers noted was larger than the average national drop in homophobia between 1998 and 2012. Against a literature in which ordinary contact averages zero, that is a striking result. It is also one topic in one city, and campaigns adopting it should treat it as promising rather than settled.
The line between the sectors was thin from the start. MoveOn began in 1998 as a one-sentence email petition from Wes Boyd and Joan Blades. In 2003 it held what it billed as the first online presidential primary. Howard Dean won it with 44 percent, and his rivals complained that MoveOn had advised him on how to market himself to its members. Dean raised $7.5 million in the second quarter of that year, at least $3 million of it online.
The strongest NGO case of campaign tempo is probably Ireland’s 2018 referendum. Together for Yes, an umbrella of more than 70 civil society groups, launched on 22 March and faced the vote on 25 May, winning 66.4 percent (Amnesty Ireland). Amnesty’s own contribution, the “It’s Time to Talk” campaign, rested on the bet that conversations among friends and family would decide the vote. A crowdfunding appeal for posters set a target of €50,000, reached it in two hours, and closed at ten times that. More than 70 organizations worked to a nine-week clock, which is the harder part.
Where NGOs copy campaigns badly
Kony 2012, from the US group Invisible Children, passed 100 million views in six days. Its maker, Jason Russell, told Reuters, “It definitely oversimplifies the issue,” and said the simplicity was deliberate. Simplification is a campaign’s trade because opponents do the correcting and the contest ends on a fixed date. Kony 2012 got its correction from Ugandans, analysts and journalists who examined the group’s accounts and structure, and people in northern Uganda said the film offended them. In the assessment of Justice in Conflict, the attention it created was squandered, and the organization’s credibility went with it.
Data is the other risk. Microtargeting is now cheap enough for any NGO to try, and the cautionary case is Cambridge Analytica. The UK Information Commissioner’s inquiry into data analytics in political campaigns ended in 2019 with Facebook paying the maximum £500,000 penalty, without admitting liability, over app developers’ access to user data without informed consent. An estimated 87 million users worldwide were affected (Silicon Republic). NGOs often hold data about beneficiaries that is more sensitive than a voter file, which is a reason to move more slowly than campaigns do.
The framework NGOs built for themselves
Not everything flowed through campaigns. In 2010 Tom Crompton of WWF-UK and staff from COIN, CPRE, Friends of the Earth and Oxfam published Common Cause, a report arguing that civil society groups should strengthen values such as concern for others and social justice, and reduce the pull of status, wealth and image. The Public Interest Research Centre’s 2011 handbook turned it into a manual. Its central claim is that every appeal strengthens some values over others. The wager is that a message which wins by triggering status or fear leaves the next cause a harder audience, which is a direct challenge to any donation page optimized on conversion alone.
Scorecard
| Case | Direction | Verdict | Why |
|---|---|---|---|
| Obama sign-up test, 2007 | Campaign to NGO | Good | Random assignment showed the staff favorite lost |
| Sanders small-dollar model, 2016 | Campaign to NGO | Good | Frequent, low-friction asks; figures are campaign-reported |
| WinRed prechecked boxes, 2020 | Campaign to NGO | Bad | Default recurring gifts led to roughly $122 million in refunds |
| Take Back Control and the £350 million bus | Campaign to NGO | Good slogan, bad claim | Simple idea; number found misleading, never retracted |
| No on 8 rebuttal, 2008 | Campaign | Bad | Reply ran about two weeks after the attack ad |
| Camp Obama, 2008 | NGO to campaign | Good | Organizing model from farmworker and Sierra Club work |
| Deep canvassing | NGO to campaign | Good, early | One replicated study, one topic, one city |
| Freedom to Marry research | NGO | Good | Tested audiences and rebuilt the message; messaging was not the only factor |
| Together for Yes, 2018 | NGO coalition | Good | 66.4 percent Yes on a nine-week clock |
| Kony 2012 | NGO copying campaign | Bad | Deliberate oversimplification, public backlash |
| Radi-Aid and the Rusty Radiator | NGO critique | Good | Shaming shifted norms even though the tactic raised money |
The best-documented exchange ran both ways at once. Proposition 8 was lost, an LGBT community center built a canvassing method from the wreckage, and a marriage movement that had lost 30 statewide votes went back to voters and tested what they said. The worst exchanges skipped a step: they dropped the test, or dropped the constraint. The working rule that follows is dull. Decide what is off limits before running the test, because the test will not decide it.
Sources
- Dan Siroker, “Obama’s $60 million dollar experiment,” Optimizely
- Kalla and Broockman, “The Minimal Persuasive Effects of Campaign Contact in General Elections,” American Political Science Review, 2018
- Cuff and Clark, “Funding Research for Advocacy,” Stanford Social Innovation Review, 2016
- “How Data Helped Win the Battle Over Same-Sex Marriage,” FiveThirtyEight
- ITV News, “Brexit group which campaigned under the slogan ‘Take Back Control,'” 2018; “Who is Dominic Cummings,” 2020
- Religion Dispatches, “New Report: Prop. 8 Was Parent-Approved,” 2010
- Windy City Times, “Parents big part of Prop 8 support,” 2010
- The Advocate, “Prop 8 lessons,” 2010
- Sanders campaign, fundraising release, May 2016
- The New York Times investigation on WinRed donations, reprinted by the East Bay Times; Snopes summary; Sen. Klobuchar’s release quoting the Times
- SAIH, “Rusty Radiator Awards 2014;” NPR, “Radiator Awards Salute ‘Manpons,’ Freezing Norwegians, Sad Babies,” 2015; Nonprofit Quarterly, “Poverty porn awards,” 2017
- Marshall Ganz, Harvard Kennedy School; Harvard Magazine, “Marshall Ganz,” 2010
- NPR, “Study finds deep conversations can reduce transgender prejudice,” 2016; Christian Science Monitor, “In-depth conversation can reduce anti-transgender prejudice,” 2016; ScienceDaily, “Canvassing conversations reduce transphobia,” 2016
- Time, “Quotes of the Day,” 2003; Center for Public Integrity, “Democrats likely to choose Howard Dean as leader,”
- Together for Yes, Wikipedia; Amnesty International Ireland annual report, 2018; Petrie-Flom Center, “The abortion referendum in Ireland,” 2018
- TechCrunch, “Makers of most viral video ever, Kony 2012,” 2012; Reuters via CSM, “Kony 2012: Director of video agrees with critics,”; Justice in Conflict, “Kony 2012: How 100 million clicks went to waste,”
- Silicon Republic, “Facebook to pay £500,000 over misuse of personal data in political campaigns,” 2019
- Common Cause: Handbook (PIRC); 2011 handbook PDF
