Question 1

How many respondents do I need for a survey to be credible?

Accepted Answer

More is better, but niche B2B surveys with 50–100 well-defined respondents are frequently cited. The credibility threshold depends on the niche: in a field where the total addressable population is small (e.g., CTOs of companies between 50 and 200 employees), 80 responses is a representative sample and will be treated as credible by journalists and AI systems. In broad consumer markets, 500+ respondents is usually the minimum for credible reporting. The key is matching your sample size claim to your stated population: "survey of 87 UK ecommerce operators" is appropriately scoped; the same 87 responses presented as "survey of ecommerce operators globally" would be misleadingly small. Scope your claims to your sample.

Question 2

Do AI systems distinguish between primary data and secondary analysis?

Accepted Answer

Partially. AI systems prefer to cite specific, attributed statistics over general claims — and a statistic with a clear source ("according to X's 2025 survey") is more likely to be cited than an unsourced assertion. Primary data (data you collected) is more inherently unique than secondary analysis (analysis of data others collected), which is why it gets cited more — anyone can analyse the same publicly available data, but no one else has your survey responses. That said, secondary analysis with a distinctive methodology or finding that isn't available elsewhere is also highly citeable. The critical factor is uniqueness, not whether the underlying data collection was yours.

Question 3

Is client case study data safe to publish?

Accepted Answer

With client permission, yes — and most clients are willing to permit publication of anonymised or named results if you ask during or after the engagement. Named case studies (with the client's name, specific metrics, and their testimonial) are the most powerful format and the most citeable. Anonymised case studies ("a UK ecommerce brand in the outdoor category") are slightly less compelling but still valuable as primary data. Aggregated client data ("across 23 clients, average improvement was X%") requires no individual client permission since no single client is identifiable. The risk: publishing specific client metrics without permission can breach client confidentiality agreements. Review your contracts before publishing anything client-specific, and ask permission explicitly before naming a client.

Question 4

How do I promote original research content to maximise citations?

Accepted Answer

Treat the research launch like a PR campaign. Identify journalists and publishers who cover your topic regularly (those who cite similar statistics already), and do direct outreach with the key finding stated in the first line of the email. "Our survey of 87 UK ecommerce operators found that 64% report mobile conversion rates below 1% — the data is publicly available if you'd like to cite it." This approach generates earned links and citations that extend the reach of the data beyond your own audience. Distribute via your email list on publication day; share the key finding (not just the link) on social to make it shareable; and consider contributing a data-driven opinion piece to a relevant trade publication that links back to the full dataset on your site.

Question 5

Can I create original data from a tool I already use, without surveying anyone?

Accepted Answer

Yes — this is often the lowest-effort path. Product analytics, CRM data, ad account performance, and project management history all contain aggregate patterns nobody outside your company has seen. A SaaS company can publish time-to-value benchmarks from product usage data; an agency can publish average results across past campaigns; an ecommerce brand can publish seasonal conversion patterns from its own store. None of this requires recruiting respondents — it requires querying data you already have and presenting the aggregate cleanly, with appropriate anonymisation where individual customers could otherwise be identified.

Question 6

What happens if someone disputes or misquotes my data after it's cited widely?

Accepted Answer

Keep your methodology page or section permanently accessible and unchanged at its original URL — this is your reference point if a citation misrepresents the finding. If a misquote happens, a brief, factual correction (not a dispute) published as an update to the original page, with a changelog note, is usually sufficient; most secondary citations will eventually update or fade. Avoid editing the original numbers after publication except to correct a genuine error, and always note the correction explicitly — silently changing a widely cited statistic damages credibility more than the original error did.

AI citations

Original Data as a Citation Magnet: How to Create Research Content Without a Large Budget