Blog

How to run a content marketing survey people can trust

Design and run a survey, from the question it must answer to a checked, publishable result, so that what you publish says only what the responses support.

By Andre Guelmann · Updated · 12 min read

Contents

In The B2B Numbers Census, a website-content audit of 614 B2B companies observed in September 2026 and published in October 2026 by ClaimTheStory Research, 7 in 10 companies used someone else’s numbers, and 1 in 3 never published a number from their own data.

A survey gives you publishable results when you start with one sharp question, reach a defined group, screen out people who can’t answer it, remove duplicates and unfinished submissions before counting, and report only the groups that meet your minimum base. This guide walks through that in seven sections, in order: deciding when a survey is the right route, writing the core question, choosing who to invite, drafting the questionnaire, running fieldwork and cleaning, checking the finding, and setting your publication rules.

What a survey for content is, and when it is the right route

A survey for content measures something about a defined group that outside readers care about. That’s different from an audience-feedback survey sent to your own subscribers to plan your editorial calendar. This guide covers surveys built for publication, not internal editorial polls.

A survey measures what respondents report: attitudes, habits and their own account of what they did. It can investigate behavior, but it records what respondents say happened; in the example below, those answers aren’t checked against billing or accounting records. If your question needs verified records of behavior, check your own data or public records first. Our guide to how to get data for content covers choosing between internal records, public data and new collection.

What a survey can measure

Can measure

What respondents report: attitudes, habits, and their own account of what they did.

Can’t measure

Verified behavior. For that, check your own records or public data first.

TrendCandy’s guide to survey research for content marketing makes the point that the same survey can support both content marketing and PR. If you plan to hire an outside research firm instead of designing and running the survey yourself, see PR surveys for how to commission and judge a survey supplier. If you’re still weighing the overall investment, start with original research for content marketing.

Write the one question the survey must answer

Start with one core question, not a wish list of topics. A question is worth investigating when a real audience decision depends on the answer and your study adds something to what’s already published: updated evidence, a different population or a better-supported analysis. B2B research ideas covers how to choose one.

Turn your question into a concrete plan before you draft any items: who is in the group, the exact figure that answers the question (a share of what, counted how), the period the answers refer to, and the comparison you plan to make.

The survey plan, before any question is drafted
  • Group Independent restaurants, one owner or manager answering for each.
  • Figure Restaurants answering “Yes” divided by retained responses; “Don’t know” stays in the base.
  • Period The 12 months before each restaurant answered.
  • Collection 3–24 March 2026.
  • Comparison By restaurant size.

Before fieldwork starts, write down the headline you’d publish whichever way the result goes. If neither outcome would be worth publishing, change the question before you spend time building the survey.

Decide who to ask and how you will reach them

Separate the group you want to describe from the people you can actually reach. Your findings describe only the people your invitations reached who chose to answer.

Each common invitation source shapes who shows up in your results:

  • Your own customer list: easy to invite, but over-represents businesses that already buy software like yours and leaves out non-customers.
  • A paid respondent panel: reaches beyond your own list, though opt-in panels reflect people who signed up to take surveys rather than a random sample of every business in an industry.
  • Social followers or newsletter readers: reaches people already paying attention to your brand or a partner’s, and misses everyone outside that circle.

Use screening questions at the very start to keep only respondents who belong to the group your findings describe, and record how many people are screened out. When your findings describe organizations rather than individuals, allow one response per organization and remove duplicates across every invitation source.

There’s no universal minimum number of responses for a content survey. What you need depends on what you’re measuring and how many groups you compare. Set your own minimum group size before fieldwork starts, label it as your choice for that dataset, and apply it the same way everywhere.

Write questions that measure what you plan to publish

Give every question a job before you write it: screening, the core measure, the grouping variable (such as company size) or a single follow-up. Cut any question without one.

Follow four wording rules for each item:

  1. Ask about a defined period: use “in the last 12 months” rather than “recently”.
  2. Avoid leading wording: don’t nudge respondents toward an expected answer.
  3. Cover every option: include “Don’t know” so respondents aren’t forced to guess.
  4. Ask one thing at a time: don’t combine switching suppliers and finding a “better” supplier in the same question.
The core question, before and after the four rules

Before

“Have you, like many restaurants, switched to a better supplier recently?” It leads, asks two things at once, and has no fixed period.

After

“In the last 12 months, has your restaurant changed its main produce supplier?” Yes / No / Don’t know.

Answer options decide what you can count later. Single-answer options give you clean shares. An optional open-text box can give you direct quotes, but those aren’t counted as percentages unless they’re coded. Put screening questions first, general questions before specific follow-ups, and grouping questions at the end. Before launch, test the draft with a few people from your target group and note any wording they misread.

Run the fieldwork and clean the responses

Fix your open and close dates before launch and record them as the collection period. Keep that separate from the reference period the questions ask about. Don’t peek at early totals and close the survey early when a percentage looks good; the close date stays fixed.

Cleaning is never optional, and every removal is counted and recorded:

  • Screened out: people who didn’t pass your opening rules (recorded separately from eligible submissions).
  • Unfinished responses: removed completely rather than partly counted, applied the same way across every table.
  • Duplicates: extra responses from the same organization across invitation sources (keep the first complete submission).
  • Speeders: submissions finished too fast to have been read.
The cleaning log
  • Screened out 63: not an owner or manager ordering produce at an independent restaurant.
  • Eligible 412 submissions before cleaning.
  • Unfinished 17 removed.
  • Duplicates 11 removed: same restaurant, matched on name and ZIP code.
  • Speeders 0 removed.
  • Retained 384 restaurants, one response each.

Check the finding before you publish it

Apply your pre-set minimum group size before you publish a single percentage. In the restaurant example, the 384 cleaned responses split across three size groups: 148 restaurants with up to 30 seats, 214 with 31–50 seats, and 22 with 51 or more seats (148 + 214 + 22 = 384). Because 22 is below the illustrative minimum of 40 responses, the 51+ seats group is left out of every published table and total, and its switching count and share are never reported anywhere.

Horizontal bar chart of the share of responding independent restaurants that reported changing their main produce supplier in the last 12 months, by size: up to 30 seats (base 148) 25%, 31–50 seats (base 214) 38%, total, up to 50 seats (base 362) 33%. 51 or more seats: not reported: below the 40-response minimum.
Larger independent restaurants were more likely to report changing their main produce supplier. Self-reported; each group’s base is in its label. Illustrative data.

Check what your table can’t show before you write a headline. A share of restaurants that switched suppliers doesn’t measure how often they switch or how much produce they buy. It can’t say whether switchers buy more or less than restaurants that didn’t switch, or why a higher share of the larger restaurants reported switching. Nor does it show that switching caused anything. Answering those questions would need order-volume fields and a way to test causes, and this table has neither. For the full method on checking findings and building charts, see turning data into content.

What the table can and can’t show

Can show

The share of responding restaurants that reported changing their main produce supplier in the last 12 months, by size.

Can’t show

  • How often restaurants switch, or how much produce they buy.
  • Why more of the larger restaurants switched.
  • That switching caused anything.

Decide what you can publish, and how others can check it

Set a clear publication policy with the data owner before you release any findings. A minimum of 40 responses never makes survey data anonymous on its own, so always assess disclosure risk with the data owner.

When you publish, attach a complete methods note that covers eight items:

  1. Source of respondents and invitation channels
  2. Collection dates and the reference period asked about
  3. Exact question wording and answer options
  4. Bases for every published group and total
  5. How responses were screened and cleaned, with counts removed
  6. Limits of the sample and self-reported answers
  7. Suggested citation line
  8. What data access is offered, without confidential records

Writing the finished survey report, including that methods note in full, is covered in writing up survey and research findings. One survey can feed several pieces of content, as long as each piece stays within what the responses support.


If you know what you want to say, we find the data that would support it, and tell you if somebody has already published it. Book a demo