After disappointing but unsurprising losses earlier this month, Republicans are asking what it will take to reverse course. How can they hold their House and Senate majorities in 2026 and win more governors’ races? The honest truth is that nobody knows.
Still, pundits will argue that the answer lies in embracing certain policies or speaking directly to key groups. Others will say it is about tactics such as digital outreach or ground game. Some will insist that campaigns need to stop listening to consultants altogether and recruit better candidates. The fact is, they may all be right, so let’s put it to the test.
Modern election campaigns have become more sophisticated and data driven, yet also more polarized and less predictable. Voters are harder to persuade, and old formulas no longer guarantee success. The path forward is not guessing the next big message or discovering a new, revolutionary tactic.
It is experimentation – of everything.
Republicans need to test it all: messages, strategies, tactics, and platforms. Every assumption must be open to scrutiny, and allow the data to pave the path forward.
In recent years, campaign research has begun to use randomized controlled trials, the same scientific method that pharmaceutical companies rely on to develop new drugs. Some specific sample of voters receive a targeted treatment, such as mail, door knocks, or messages, and other voters in the control group do not. This approach can and should be applied to how we communicate with voters, from what we say to where and how we say it.
Take television advertising. Online testing panels now allow campaigns to compare different versions of ads with real voters before committing millions of dollars to airtime. This type of pre-testing has been proven in our own experiments to save money and improve effectiveness. Yet many campaigns still rely on gut instinct for the most expensive line items in their budgets. If Republicans want to win in 2026, they must commit to rigorously testing their ads.
Polling faces similar challenges. Finding representative samples of an electorate has grown more difficult and more expensive than ever before, and, even then, forecasts are often well off the mark.
In Virginia’s major off-year elections this month, dozens of public and internal polling data proved inaccurate. Campaigns should borrow from academic research by experimenting with new methods of data collection and weighting. That means accepting incomplete surveys, calibrating results against past vote history, and combining polling data with other indicators of voter behavior.
The Center for Campaign Innovation is running large-scale randomized trials to evaluate the effectiveness of tactics such as digital ads, direct mail, canvassing, relational organizing, and texting. Turnout provides a clear and objective outcome. We can measure whether a voter who received a specific treatment actually voted – and when they didn’t.
This kind of testing runs against our instincts as campaigners. When we believe something is effective, campaigns want to do it for as many voters as possible. But without a control group of voters who do not receive the treatment, we cannot know what truly made the difference. That is why independent organizations focused on experimentation are so important.
The reality is that campaigns do not actually know what works. They say they do, but they just don’t. As the old saying goes, half of the money we spend on advertising is wasted, but the trouble is we don’t know which half.
If Republicans want to win in 2026 and beyond, they must embrace experimentation. Test your messaging, your strategy, and your creative. Set ego aside. Follow the data. That is how we will finally learn what works and how to win the next election.