A PR survey earns press when it answers one question a reporter's readers care about, uses a sample large enough to defend, and produces a headline that works as a single number. Field about 1,000 respondents for a national claim and at least 400 for a narrow one, write 8 to 12 neutral questions, test three headline versions on a small pilot, and release the results with the full methodology and a data table.
Updated October 21, 2026
Summary
- Build the survey backward from one headline and one core question.
- Use 1,000 respondents for a national claim (margin of error about 3.1 points) and 400 or more for a niche audience.
- Document sample design, mode, weighting and response rate, because journalists and analysts ask for all four.
- Write neutral questions, randomize answer choices, and test the draft on three outside readers.
- Draft three headlines, run a four-point check, and pilot with 100 to 150 respondents.
- Release with a methodology note, the full question wording, and a data table.
Contents
- What makes a PR survey earn press coverage?
- How many respondents does a PR survey need?
- How do sample design, weighting and mode change survey results?
- How do consumer surveys differ from B2B surveys?
- How do you write survey questions that hold up?
- How do you test the headline before you field the survey?
- What does a worked example look like?
- How do you release the results to reporters?
- What mistakes sink a PR survey?
- Frequently asked questions
What makes a PR survey earn press coverage?
A PR survey earns coverage when it gives a reporter a fresh, quotable number about something their audience already cares about. Reporters do not cover surveys. They cover findings that explain a trend, settle an argument, or surprise readers.
Three things separate surveys that travel from surveys that sit unused:
- One clear question. The best surveys exist to answer a single question, such as how many employees plan to switch jobs this year.
- A defensible sample. Reporters and editors check who answered and how many. A stated method protects the finding when someone challenges it.
- A number that stands alone. If the headline needs a paragraph of context, it will not be quoted.
The recurring franchises on the survey research hub, such as the Edelman Trust Barometer, show the long game: the same question asked on a fixed schedule becomes a citable record. A first survey can still earn a strong single cycle if it follows the steps below.
How many respondents does a PR survey need?
Most PR surveys need 1,000 respondents for a national claim and 400 or more for a narrower audience. The margin of error shrinks as the sample grows, but each extra respondent adds less accuracy than the last.
The table shows the margin of error at a 95% confidence level for a simple random sample, using the standard worst case of a 50/50 split.
| Respondents | Margin of error (points) | Best use |
|---|
| 400 | ±4.9 | Niche audience or early pilot |
| 500 | ±4.4 | Single industry or one metro area |
| 1,000 | ±3.1 | National consumer claim |
| 1,500 | ±2.5 | National claim with sub-group cuts |
| 2,000 | ±2.2 | Trend tracking across several segments |
Two cautions apply. First, every sub-group has its own margin of error. If 1,000 respondents include only 100 parents, the parent finding carries about ±9.8 points. Second, online panels are not random samples, so treat these figures as a guide and state the panel provider and weighting in your methodology. How Large Should a Survey Be? covers sample size decisions in full.
How do sample design, weighting and mode change survey results?
Sample design, weighting and mode decide whether survey data measures what it claims to measure. Teams that document all three, plus response rate and question wording, produce findings that survive scrutiny from journalists and analysts.
Probability sampling gives every member of the population a known chance of selection, and non-probability sampling does not. Only probability samples support a formal margin of error. Most commercial research, including most surveys run as press tactics, uses non-probability online panels because they field in a day at a fraction of the cost. A team that releases non-probability results as if they were projectable takes on credibility risk, so disclose the sample source and the weighting.
Weighting adjusts each respondent's influence so the sample matches the population's known makeup. Raw samples usually underrepresent younger and non-college respondents, and the standard fix compares the sample to Census benchmarks and gives underrepresented groups more weight. When you publish weighted results, disclose the weighting variables, the population benchmark and the unweighted sample makeup. Polling Errors That Change Headlines explains why pollsters added education and, in some cases, recalled prior vote as weighting variables.
Mode is how the survey is delivered, and it changes answers. People give more socially desirable answers to a live interviewer than to a screen, so the effect is strongest in person and weakest online. An online result of 47% and a phone result of 51% on the same question may not disagree at all, so check for mode effects before announcing that your data contradicts a major franchise.
Low response rates matter because the people who answer can differ from the people who do not. That gap is non-response bias, and weighting does not fully repair it. Telephone response rates once topped 50% and are now often below 10%, and researchers generally agree a low rate does not invalidate a survey by itself. Address non-response directly in the methodology note.
How do consumer surveys differ from B2B surveys?
Consumer and B2B surveys use the same statistics but differ in sample source, sample size, incentives, length, cost and press cycle. A consumer design applied to executives produces data that does not survive scrutiny, so match the design to the audience.
| Dimension | Consumer survey | B2B survey |
|---|
| Sample source | Large online panels such as Cint, Lucid, Toluna, Dynata and YouGov | Association lists, trade-publication subscribers, client lists, LinkedIn recruitment and specialist B2B panels |
| Typical sample size | 1,000 to 30,000 | 200 to 2,000 |
| Typical incentive | A few dollars in gift cards or points | Honoraria, donations or report access worth more per respondent |
| Questionnaire length | 10 to 15 minutes, with essential questions first | 20 to 45 minutes, with operational detail |
| Working cost estimate | About $5 to $30 per complete | About $50 to $500 per complete, higher for C-suite audiences |
| Press cycle | Front-loaded, peaking within 48 hours in consumer and morning media | Slower, with citations building over months in trade press and analyst reports |
The cost estimates are EPR working figures for 2026 and vary with method and audience. B2B designs also need enough respondents inside each industry segment to support segment-level claims, which pushes multi-industry studies toward 1,000 or more. Rarely does one survey serve both audiences well, so run two coordinated studies if you need both views.
Write questions that a skeptical reader would call fair. Neutral wording is the single biggest factor in whether a survey survives scrutiny.
Follow this order when building the questionnaire:
- Start with the headline you hope to publish, then work backward to the one question that supports it.
- Add 7 to 11 supporting questions that give context, such as age, region, and a "why" follow-up.
- Write each question in plain language, one idea at a time.
- Put general questions before specific ones, so early answers do not steer later ones.
- Randomize the order of answer choices.
- Run the questionnaire past three people outside your team and fix every question they stumble on.
Avoid these wording traps:
- Leading phrases, such as "Don't you agree that..."
- Double-barreled questions that ask about two things at once
- Loaded terms that signal the answer you want
- Missing options, such as no "none of these" or "not sure" choice
Small changes in wording, order and answer options can also move results by large amounts. Framing effects appear when logically equivalent questions get different answers, order effects appear when an earlier question shapes a later one, and a five-point scale can produce a different distribution than a seven-point scale. Test draft questions with cognitive interviews before fielding.
How do you test the headline before you field the survey?
Write three versions of the headline before you spend the budget, then check which one the data could support. A headline test catches weak ideas while they are cheap to fix.
Draft each version around a different angle:
| Angle | Pattern | Example (illustrative) |
|---|
| Number first | "X% of [audience] plan to [action]" | "6 in 10 renters plan to move in 2027" |
| Comparison | "[Group A] is twice as likely as [Group B] to..." | "Renters under 30 are twice as likely to..." |
| Surprise | "Most [audience] still [unexpected behavior]" | "Most renters still pay by check" |
Then apply a four-point check to each version. Does it contain one number? Would a reporter in your target outlet's beat want it? Can you name the question that proves it? Would it still be true if the result came in 5 points lower? Keep the version that passes all four, and field a pilot of 100 to 150 respondents to confirm the answer is not a 90/10 split that leaves no story.
What does a worked example look like?
The example below is illustrative, not a real study. A regional credit union wants coverage in local business media about how families handle rising costs.
- Question: Are households in the region cutting savings to cover everyday bills?
- Audience and sample: 800 adults in the credit union's metro area from an online panel, giving a margin of error near ±3.5 points.
- Questions: 10 in total, including one core question ("In the past 6 months, have you reduced your monthly savings to cover everyday expenses?") and follow-ups on age, income band, and the main cost driver.
- Headline versions tested: "4 in 10 local households cut savings," "Younger households cut savings twice as often," and "Housing, not groceries, drives savings cuts."
- Result of the pilot: The first version clears the four-point check, and the third offers a second story for a follow-up pitch.
- Release: The credit union sends a one-page findings summary with the data table to three local business reporters, with the full method attached.
The point of the example is the sequence: one question, one defensible sample, three headlines, and one release package.
How do you release the results to reporters?
Release the findings with a package that makes a reporter's job easy. A survey that arrives with the method and data attached is far easier to cover than one that arrives as a claim.
Include these items in every release:
- A headline finding in the first sentence of the pitch, with the number
- A methodology note covering sample size, dates, panel provider, weighting, mode, response rate and margin of error
- A data table with the full question wording and results
- A named spokesperson who can explain the finding
- One exclusive offer to the outlet that best fits the story, with a clear timing window
Methodology note checklist
State the sample size, field dates, who was surveyed, how respondents were recruited, the mode, any weighting applied, the response rate, the margin of error, and the exact wording of every question. Publish the full data table with the release.
For pitch structure, see the press release guide and the data on what happens to off-beat pitches.
What mistakes sink a PR survey?
Most failures come from a small set of avoidable errors.
- Starting with the answer. A survey designed to prove a point reads as marketing and gets ignored.
- Skipping the method. Reporters ask where the number came from, and a missing methodology ends the conversation.
- Over-slicing small samples. Reporting results for a 60-person sub-group invites an easy correction.
- Burying the finding. A strong number in paragraph six gets cut.
- Running once and stopping. The surveys that compound repeat on a schedule, as covered on the survey research hub.
Frequently asked questions
How much does a PR survey cost?
Cost depends on audience and sample size. EPR's working estimate is about $5 to $30 per completed consumer response and about $50 to $500 per completed B2B response, so B2B surveys cost roughly an order of magnitude more. Get quotes from at least two panel providers before you set the budget.
How many questions should a PR survey have?
Eight to twelve questions works for most PR surveys. Shorter surveys have higher completion rates, and every question should support a possible headline or give context for one.
Can a small survey still earn press?
Yes, if you state the sample honestly and claim only what it supports. A 400-person survey of a defined niche can earn trade coverage when the headline stays within the margin of error.
Can a non-probability survey be published as press content?
Yes, if the release states the sample source, the weighting and the limits of the claims. Non-probability samples support most commercial purposes but not the same projection claims as probability samples.
What should the methodology note say?
State the sample size, field dates, who was surveyed, how they were recruited, the mode, any weighting, the response rate and the margin of error. Publish the exact question wording alongside the results.
How long does it take to run a PR survey?
Plan roughly three to six weeks from questionnaire to release: about a week to draft and test, a few days to field a panel survey, and one to two weeks to analyze and prepare the release package.
Should I use AI to analyze open-ended answers?
AI can speed up first-pass coding of open-ended answers, but a person should review the codes before publishing. Disclose where AI was used, as noted on the survey research hub.
