A survey is what you run when you do not have usable internal data. Done well it produces a citable number that journalists reach for repeatedly. Done badly it produces a press release nobody covers and a line item somebody has to explain.
The determining choices are made early — before design, before fieldwork, and certainly before anyone writes the report.
The three questions that decide everything
1. Would a journalist write this headline without your company in it?
Write the headline first, before designing anything. If it names your company or your category's benefit, the survey will not be cited.
Not citable: "72% of operations leaders say automation improves efficiency."
Citable: "Operations teams now spend 11.4 hours a week on manual data reconciliation — up from 8.6 two years ago."
The second is useful to a writer who has never heard of you. That is the whole test.
2. Is anyone currently asserting this without evidence?
The best survey topics are claims your industry repeats as fact with no source behind them. Find articles stating a figure and follow the citation trail — it frequently leads to a blog post from 2017 citing another blog post.
Supplying the missing source is a durable position, because the claim keeps being made.
3. Can you get 300 qualified respondents?
Below roughly three hundred, trade press treats findings as anecdote. And qualification matters more than volume — three hundred verified practitioners beat three thousand from a general panel comprehensively.
The disqualifier most companies miss
Surveying your own customers is not research. It is a customer survey, and a careful journalist will treat it as one — correctly, because your customers are self-selected for agreeing with your premise.
It can still be published; it simply has to be described accurately, and it will earn a fraction of the citations. If the goal is links, the panel must be independent of you.
Sample size, honestly
| Respondents | How it is received |
|---|---|
| Under 100 | Anecdote. Will not be cited by serious publications. |
| 100–300 | Usable for a narrow, well-defined population if you say so plainly |
| 300–600 | The workable range for most B2B categories |
| 1,000+ | Only necessary if you intend to segment heavily |
The instinct to maximise sample size is usually misplaced. Six hundred well-qualified respondents in one function beats two thousand loosely qualified ones, and costs less.
Qualification
Screening questions matter more than anything else in the design, because they determine whether the population you claim to have surveyed is the population you surveyed.
- Role verification — not "do you work in operations" but "which of these are you personally responsible for", with a list where only some options qualify.
- Company size band — B2B findings vary enormously by headcount; state the band and stick to it.
- Practical exposure — "how many times in the last quarter have you done X". Screens out people who recognise the topic but do not live it.
- Attention checks — at least one. Panel data contains straight-liners and it is your job to remove them.
Designing questions that produce findings
Three principles that separate a survey producing numbers from one producing a story.
Ask about behaviour, not opinion
"How many hours last week did you spend on X" produces a citable figure. "How important is X" produces a percentage nobody quotes.
Include at least one comparison
Against last year, against a benchmark, between segments. Findings that show change or difference get covered; static ones do not.
Include one question you might not like the answer to
A survey designed so every result flatters your category reads exactly like one. Including a genuine risk — a question where the honest answer might be inconvenient — is what makes the whole instrument credible.
The single most-cited finding in a study we ran for a client was the one contradicting their own marketing. It was covered eleven times, largely because it was surprising coming from them.
The methodology section nobody writes
Without it, a careful journalist cannot responsibly cite you. It takes twenty minutes and it doubles the citation rate.
What it must contain:
- Sample size, and the qualified population it represents
- Fieldwork dates
- How respondents were recruited, and by whom
- Screening criteria, stated explicitly
- How many responses were discarded and why
- Geographic and company-size distribution
- Who funded the research — you, stated plainly
That last line worries people and should not. Disclosed funding is normal in industry research. Undisclosed funding, discovered later, is what damages a study.
Costs and timeline
| Item | Typical range |
|---|---|
| Panel fieldwork, 300–500 qualified B2B respondents | $4,000–$12,000 |
| Design and analysis | $2,000–$6,000 |
| Report page, charts, press kit | $2,000–$5,000 |
| Outreach and pitching | $1,500–$4,000 |
| Total | $10,000–$25,000 |
Timeline: two weeks design, two to three weeks fieldwork, two weeks analysis and build, two to three weeks pitching. Call it nine to ten weeks from decision to coverage.
Against a market average of roughly $500 per bought editorial link, a study yielding 20–40 referring domains plus continued accrual for a year sits at a comparable or better unit cost — with the difference that the asset keeps working.
Where survey studies fail
- The finding is promotional. Fatal, and decided at question one.
- The sample is unqualified. Careful writers check and stop reading.
- No methodology published. The number may be sound; without the method it cannot be cited responsibly.
- Gated. Astonishingly common and it removes nearly the entire value.
- No launch. Published with no outreach earns a fraction of what it should. The pitching is a separate exercise, not an afterthought.
Making it an annual thing
The economics change substantially on repetition. Each edition links back to the last, the archive ranks for the underlying query, and by the third year journalists cite it without prompting.
It also gets cheaper: the instrument exists, the panel specification is written, and the analysis is a comparison rather than a first look. If you are going to run one survey, plan for three — and if you are not prepared to repeat it, consider whether the budget is better spent elsewhere.