← NotesWave / Notes · APR 07 2026

Strategy

How Do Campaigns Know If Their Message Is Working Before Election Day?

Every campaign thinks its message is working. Most find out they were wrong on election night. The gap between "my team likes this message" and "this message moves voters" is where campaigns go to die. The problem isn't that campaigns don't test — it's that they test the wrong things, in the wrong order, on the wrong people.

The Polling Problem: Why Traditional Surveys Don't Tell You What You Need to Know

Campaigns have relied on polling as their primary message-testing instrument for decades. The intuition makes sense: ask a representative sample of voters what they think, see which message performs best, run with it. The problem is that this model has broken down in two distinct ways.

First, polling accuracy has deteriorated significantly. The AAPOR Task Force report on 2020 pre-election polling documented the largest polling error in 40 years, with polls systematically underestimating Republican support. The error was not random — it was directional and concentrated among voters with lower trust in institutions. AAPOR's preliminary 2024 analysis found similar patterns, with polls again underestimating Republican performance across most competitive Senate and presidential battleground states.

Historical Polling Error in Competitive Races

Election CycleError in Competitive Statewide RacesNotable MissesSource
2016Significant underestimate of Republican support in MidwestWisconsin, Michigan, Pennsylvania presidential pollsPost-election analyses; no comprehensive AAPOR statewide error report
2018Generally more accurate than 2016; some notable state-level missesGeorgia governor raceIndustry reporting; no systematic AAPOR evaluation for this cycle
20204.5 pt average error in competitive Senate/Governor races (AAPOR)Largest polling error since 1980; systematic Republican undercountAAPOR 2020 Task Force Report
2022Varied; "red wave" failed to materialize but state-level errors persistedFlorida, New York substantially wrong in different directionsPost-election industry reporting; no comprehensive AAPOR evaluation
2024Systematic underestimate of Republican performance repeated (AAPOR preliminary)Most competitive Senate and presidential battleground statesAAPOR 2024 Preliminary Analysis

The 4.5 pt average error figure is from the AAPOR 2020 Task Force Report and applies specifically to competitive statewide races. AAPOR did not publish equivalent systematic analyses for all cycles, so direct cross-cycle comparisons should be treated as approximate. Methodology for measuring "polling error" varies across sources.

Second, the structural basis for polling has eroded. Pew Research Center documented that telephone survey response rates dropped from 36% in 1997 to just 6% in 2018 — a collapse that fundamentally changes who is captured in the sample. The people willing to answer a cold call from a pollster are not a random cross-section of the electorate.

The deeper problem is conceptual. Polls tell you where the race is. They do not tell you whether your message is moving it. A tracking poll showing you up 3 points doesn't tell you if that's because of your healthcare message or despite your crime message. For message testing, you need a different instrument.

Focus Groups: Useful But Dangerous

Focus groups remain a staple of campaign message research, and for legitimate reasons. They are genuinely useful for specific tasks: testing the emotional resonance of language before it goes into a poll, surfacing objections you didn't anticipate, understanding how voters talk about issues in their own words rather than your words.

The danger is that campaigns use focus groups to do things they cannot reliably do. A room of 8 to 12 people is not statistically meaningful. A single dominant personality can reshape the group's expressed views. The moderator's framing, tone, and follow-up questions materially affect what participants say — research on the moderator effect in political focus groups suggests that subtle facilitator biases can shift reported preferences by 10 to 20 percentage points.

Focus Group Strengths vs. Weaknesses

Good ForNot Good For
Finding the right language and framingPredicting vote share impact
Surfacing unexpected emotional reactionsRepresentative sampling of the electorate
Understanding how voters narrate issuesIsolating message from messenger effects
Pre-testing before larger quantitative workDistinguishing genuine opinion from social pressure
Identifying unforeseen vulnerabilitiesSegment-level message differentiation

Cost is also a limiting factor. A single professionally facilitated focus group runs $5,000 to $15,000 depending on location, recruitment, and analysis. Campaigns that run serious message research typically conduct four to eight sessions — a $40,000 to $120,000 commitment before a single dollar goes into media. For congressional campaigns or below, this is frequently the entire research budget.

The most sophisticated campaigns use focus groups for what they are: a qualitative diagnostic tool that feeds into quantitative work, not a standalone decision mechanism. The findings from a focus group should sharpen the hypotheses you test in a poll — not replace the poll.

A/B Testing in Digital: The Best Tool Most Campaigns Misuse

Digital advertising is the first medium in campaign history that allows real-time message testing at meaningful scale. Facebook, Google, and programmatic platforms will tell you, within 48 hours, which of two ads drove more clicks, lower cost-per-acquisition, or higher video completion rates. This is genuinely powerful — and genuinely misused.

The 2024 election saw over $3 billion in digital campaign advertising according to OpenSecrets tracking. The vast majority of that A/B testing was on creative variables: image versus video, one headline versus another, button color, ad format. These are legitimate optimizations. They are not message tests.

What Campaigns A/B Test vs. What They Should Test

What Most Campaigns TestWhat They Should Test
Image vs. video formatAffirmative frame vs. contrast frame
Headline word choiceEconomic message vs. values message
Call-to-action button textIssue prioritization (healthcare vs. economy vs. safety)
Color paletteCandidate biography emphasis vs. policy emphasis
Audience age targetingMessage resonance across voter segments

There is also a fundamental sampling problem. The people who see your Facebook ads skew younger, more urban, more educated, and more engaged online than the median voter in most competitive districts. Pew Research data consistently shows that Facebook's active political ad audience is not a representative cross-section of registered voters. If your message tests well with Facebook users, you have learned something about Facebook users.

Digital A/B testing is a legitimate optimization tool once you have established which messages deserve investment. It is not a reliable way to discover which messages to invest in.

The Emerging Practice: Segment-Level Message Simulation

The question every campaign should be asking is not "does this message work?" It is: "does this message work on swing voters in suburban precincts who voted in 2020 but skipped 2022?"

Those two questions have different answers. A crime-and-safety message may move soft Republican men in exurban counties while actively repelling college-educated women in swing suburbs. Running a single poll asking "which message is more effective" averages across those groups and produces a number that is accurate for no one.

The voter file makes this tractable. Every campaign has access to individual-level data on voting history, party registration, geography, and demographics. These data points segment the electorate into meaningful subgroups — base supporters who need mobilization, low-propensity supporters who need persuasion to turn out, swing voters persuadable on specific issues, soft opposition reachable on crosscutting concerns.

Research from Broockman and Kalla (PNAS, 2016) on deep canvassing and message targeting demonstrated that persuasion effects vary substantially across voter segments — generic untargeted messaging showed minimal persuasion effects, while segment-matched messaging showed persistent vote-choice changes. Subsequent research has consistently found that segment specificity is the dominant variable in message effectiveness, outperforming message content when the two are held constant.

The practical implication: before spending on production and distribution, test your message against the actual voter file segments you are trying to move. Not a poll of registered voters. Not a focus group of demographically matched participants. The specific precincts, turnout histories, and demographic profiles that make up your path to victory.

Want to test your message against real voter segments before you spend?

The most effective message testing connects directly to your voter file — so the segments you test against are the same segments you'll be spending media to move.

See How Wave Works →

How Much Should Campaigns Spend on Message Testing?

Campaigns consistently under-invest in message research relative to media. The conventional rule of thumb among political consultants is roughly 3 to 5 percent of budget on research and 50 to 60 percent on paid media — though actual allocations vary widely by race type and competitiveness. Whatever the specific ratio, the pattern is the same: research is treated as an afterthought rather than as the foundation for media investment decisions.

Message Testing Methods: Cost, Sample Size, and Turnaround

MethodTypical CostSample SizeTurnaroundBest For
Traditional polling$25,000–$50,000 per poll500–800 LV2–3 weeksHorse race, broad message test
Online survey panel$8,000–$18,000800–2,0003–5 daysRapid message framing tests
Focus groups$5,000–$15,000 per session8–12 per group1–2 weeksLanguage and emotional resonance
Digital A/B testing$5,000–$20,000 in ad spendThousands of impressions48–72 hoursCreative optimization
Segment simulationVaries by platformFull voter file24–48 hoursTargeted message allocation

Cost ranges reflect 2024–2025 market rates for competitive campaigns. Polling costs from Campaigns and Elections.

Recommended Research Budget by Race Type

Race TypeTotal Budget RangeRec. Research AllocationMinimum Research Spend
City Council / Local$50K–$250K8–12%$5,000–$15,000
State Legislature$200K–$1M6–10%$15,000–$50,000
Congressional$1M–$5M5–8%$75,000–$200,000
Statewide$5M–$50M+4–6%$300,000–$1M+

Benchmarks based on Federal Election Commission filings and industry reporting. Race competitiveness increases research ROI substantially.

When Messaging Went Wrong: Real Examples

The most instructive cases are campaigns that had high confidence in their message — based on polling and focus groups — and still lost because the message moved differently across segments than the research suggested.

"Defund the Police" — 2020

The 2020 cycle produced a canonical case study in segment-level message divergence. National polling on the phrase "defund the police" showed majority opposition overall. But the aggregate number masked dramatic segment variation: the phrase polled favorably among younger Black voters in urban areas, unfavorably among older Black voters, and extremely unfavorably among swing suburban voters of every demographic.

Democrats who ran in suburban swing districts against the backdrop of that framing — regardless of whether they personally endorsed it — suffered measurable losses according to Brookings Institution analysis. The message that was intended for one segment became the message attributed to all candidates by another segment. The failure was not in the research — it was in the assumption that segment-level findings could be averaged into a single strategy.

2024 Battleground Senate Races

Multiple 2024 Senate campaigns in battleground states ran message strategies rooted in economic contrast — inflation, cost of living, corporate profits — that polled well among their base and among general registered-voter samples. The final results showed underperformance relative to polling in almost every case, with the gap concentrated among voters who identified as economically anxious but ultimately prioritized other concerns.

The AAPOR preliminary 2024 analysis notes that polls systematically overstated Democratic performance, particularly in Senate races. Part of the explanation is likely that campaign message research — which informed targeting decisions — shared the same sampling biases as the horse-race polls. The campaigns were testing their messages on a sample that did not accurately represent the voters they needed to move.

The Down-Ballot Coattail Assumption

A recurring error in down-ballot races is assuming that top-of-ticket message research transfers to lower offices. State legislative and congressional campaigns routinely adapt presidential or gubernatorial message frameworks — framing that was tested on a statewide or national electorate — and apply it to much smaller, more idiosyncratic districts. The segment composition of a competitive state house district is rarely a miniature version of the statewide electorate. Message research that doesn't account for the specific district's voter composition is research conducted on the wrong population.

Building a Message Testing Program: A Practical Checklist

Campaigns that build systematic message testing programs outperform those that treat it as a one-time pre-campaign exercise. The following sequence applies at any budget level, scaled to available resources.

  • Establish a baseline poll early. Before any message testing, you need to know where the race stands and how voters currently perceive the candidates. A baseline poll establishes the starting point against which all subsequent message movement is measured. Without it, you cannot know whether a message is working.
  • Define your voter segments from the file, not from intuition. Pull your actual voter file and segment based on turnout history, party registration, geography, and available demographic data. Your path to victory is built from specific segments — base mobilization targets, low-propensity persuasion targets, swing persuasion targets. Each segment needs a message strategy.
  • Test 3 to 5 distinct message frames, not variations of one frame. Most campaigns test executions of the same underlying message. Test genuinely different frames: economic contrast versus values contrast versus biography versus record. Discover which frame moves which segments before committing to production.
  • Measure vote-choice movement, not agreement. Message testing that asks "do you agree with this message?" tells you about message content, not persuasion. The relevant question is whether exposure to the message shifts vote preference. Pre-post ballot test designs — where respondents are asked vote preference before and after message exposure — are the gold standard.
  • Test against segments, not just the overall electorate. Aggregate results mask segment divergence. Report message performance separately for base voters, swing voters, and soft opposition. A message that underperforms overall but strongly moves swing suburban women may be your most important message — you just need to see the segment-level numbers.
  • Iterate monthly during campaign season, not just at kickoff. Campaigns that test messages in February and don't revisit them until October are operating on stale intelligence. The issue environment shifts. Opponent behavior creates new vulnerabilities and opportunities. A message that tested well six months ago may be neutralized or even counterproductive by the time the media spend is at its peak.
  • Build a message matrix, not a single message. Different segments need different messages. Your base needs a mobilization message. Your swing targets need a persuasion message. Different demographic groups within the swing need different frames. A message matrix lets you distribute different content to different segments while maintaining coherent campaign themes.

Build your message matrix before you commit to media.

Wave maps message resonance across your voter file segments — so you know what to say to base voters, swing voters, and soft opposition before the money is spent.

See Wave's Message Intelligence →

The Bottom Line

The campaigns that win in 2026 won't necessarily be the ones with the best message. They'll be the ones that know which message works on which voters — and found out before they spent the media budget.

Traditional polling tells you where the race stands but cannot isolate message effects. Focus groups surface qualitative insights but cannot predict vote share movement. Digital A/B testing optimizes creative execution but tests the wrong population. These tools are valuable in the right applications. None of them, alone or together, answers the question that determines whether a media strategy will work.

The answer requires testing messages against the actual voter segments in your actual district, measuring vote-choice movement rather than message agreement, and updating the research as the campaign environment changes. That is a higher bar than most campaigns set for their message research. It is also the bar that separates campaigns that run confident strategies into election night losses from campaigns that know exactly which voters they are moving and why.

The gap between "my team likes this message" and "this message moves the voters I need" is not bridged by conviction. It is bridged by research designed to find out the answer before the dollars are committed.

Filed under strategy on April 7, 2026.