In short
Start from the story you hope to tell, write neutral questions that could prove you wrong, and ask a sample that matches the country by age, gender and region. Check every response, weight the results to the country's profile, and publish a methods note with who ran it, when, how many answered, how they were reached and the exact question wording. SampleLabs Agency runs it for you from €1,500, or you can order 1,000 five-minute responses from Recruit for €2,619 before VAT and run it yourself.
Step by step.
Ten steps from the story you hope to tell to the tables you publish with it, with how long each one takes.
-
Start from the story
Half a day
Write the headline you hope for, and the one you'd publish if the answer goes the other way. If only one of them is a story, journalists will treat the survey as an advert. Pick three to five findings worth asking about.
-
Decide who and how many
1 hour
Choose the population you'll write about, like all adults in one country, and the groups you want to quote on their own, like under-30s or each region. Every group you quote needs enough people of its own, so size the sample for the smallest one.
-
Write neutral questions
1 day
Ask one thing per question, give balanced answer scales and a "Don't know". Don't steer people towards the answer you want, and ask about the topic before you mention your brand. Journalists may ask for the exact wording.
-
Set quotas to match the country
1 hour
Set quotas on age, gender and region so the people who answer look like the population. Pew Research Center showed in 2024 how an online opt-in poll can mislead: 20 in 100 under-30s agreed the Holocaust is a myth, against 3 in 100 in a carefully drawn sample.
-
Test it and run a small pilot
1 to 2 days
Have colleagues take it, then open it to a small first group and read every answer. Fix any question people misread before everyone sees it.
-
Field it
1 to 2 weeks
Open the survey to the full sample. On SampleLabs Recruit, 1,000 five-minute responses arrive within 6 to 14 days, and our Agency team fields 1,000 to 5,000 responses a week.
-
Check every response
Half a day
Remove bots, speeders, failed attention questions and nonsense open answers, using rules you wrote down before fieldwork, and note how many went for each reason. Our guide to cleaning a dataset with bots has the steps.
-
Weight and analyse
1 day
Weight the results to the population's known profile from official statistics, so a group that answered more than its share doesn't tip the result. Then check that every finding you want to quote has enough people behind it.
-
Write the methods note
1 hour
Put it under the release: who commissioned the survey, the dates, how people were reached, who they represent, how many answered and where. The British Polling Council asks its members to publish exactly these, with the full question wording when it's practical.
-
Publish the tables
1 hour
Put the full tables and question wording online and link them from the release. British Polling Council members publish them within 2 working days of the results going public.
Times are a rough guide for one person doing the work.
What this costs with SampleLabs.
Two ways to pay for it, before VAT.
- Done for you by our Agency team A senior researcher plans the sample and the questions with you, runs fieldwork, checks every answer and writes the methods note
- From €1,500
- Most Agency projects Bigger samples, more countries, or weighting and a write-up
- €3,000 or more
- Done yourself with Recruit 1,000 five-minute responses, delivered within 6 to 14 days
- €2,619
- Your survey built in another tool Protect's check page for every response, €0.05 each
- €50
Above 500 responses, a managed Agency project usually works out better than ordering the responses yourself. Your brand goes on the survey, and we can stay out of it or be named as the fieldwork partner.
Questions people ask.
How many people should a PR survey ask?
Work back from the smallest group you want to quote on its own, like one region or one age group: it needs enough people for the finding to hold. Our sample size calculator works out the number.
What do journalists want to know about a survey?
Who paid for it, who ran it, when, how people were reached, who they represent, how many answered and the exact question wording. The British Polling Council asks its members to publish these with every result that goes public.
Which companies run surveys with news value?
Research agencies run them for you, and survey tools sell you the responses to run it yourself. SampleLabs does both: our Agency team runs the whole survey under your brand, or you order responses from Recruit.
Can I buy a representative sample for a survey I built in another tool and still get help with weighting?
Yes. Recruit sells general population samples for a survey in any tool, priced before you pay. If you want the quotas, weighting and tables done for you too, our Agency team does that as part of the project.
How long does a PR survey take?
About three weeks is comfortable: a week to write and test it, one to two weeks in the field, and a few days to check, weight and write it up. A short survey with a small sample can be faster.
How much does a PR survey cost?
Run for you by our Agency team, projects start at €1,500, and most are €3,000 or more. Done yourself, 1,000 five-minute responses from SampleLabs Recruit cost €2,619 before VAT.
Do I need to weight the results?
Usually, yes. Even with quotas, some groups answer more than their share. Weighting to official population figures evens that out, and the methods note should say what you weighted to.
Can the survey go out under my brand?
Yes. It's your survey and your story. We can stay out of it, or be named as the fieldwork partner in the methods note.
What should I ask before I hire a survey research agency?
Who will answer and how they're found, how every answer is checked for bots and AI, the fixed price and the timeline, and what you get at the end: data, tables, a methods note. A good agency also tells you which questions push people towards an answer.
Are online polls reliable enough for a press release?
They can be, if the sample matches the population and the answers are checked. Pew Research Center showed in 2024 that an online opt-in poll can badly overstate some views among young people, so quotas, checks and weighting matter.
Where the outside facts come from
- British Polling Council, Objects and rules, what members must disclose, checked 22 November 2026
- Pew Research Center, "Online opt-in polls can produce misleading results, especially for young people and Hispanic adults", 5 March 2024
More guides
- Run a 300-person B2B survey in 10 days. A day-by-day plan for a 300-person B2B survey: screener, quality checks, pilot, fieldwork and cleaning, with what each part costs on SampleLabs.
- Clean a survey dataset that has bots in it. Step by step: find bots, AI-written answers, speeders and repeat entries in survey data you already have, remove them fairly, and keep them out next time.
A survey journalists can check.
Tell us the story you want to test, and a senior researcher comes back with a plan and a fixed price.
We tell you if it's doable, by when and for how much.