How to write a customer case study: eight blocks, one evidence table

To learn how to write a customer case study, interview the customer, collect before-and-after evidence, and fill eight blocks: the result, the starting point, the trigger, the work, the measured change, what failed, a named quote, and the lesson another reader can copy. Publish only claims you can source. A case study is evidence, not a testimonial with a logo on top.

Most case studies fail at collection, not writing. The team has a happy customer and a deadline, so it writes a story with no baseline, no date and no number that survives a skeptical reader. The fix is to decide what evidence exists before you write a sentence.

A case study is only as strong as the customer's least flattering verifiable fact.

What makes a customer case study worth reading

A reader gets value when the page answers three questions in order: what changed, how do you know, and what would I do differently? Logos and adjectives answer none of them. The page has to show a starting point, a change that can be checked, and a limit the reader can recognize in their own situation.

This is why a case study and a testimonial are different assets. A testimonial says the customer is happy. A case study shows the work that made the result plausible, including the parts that did not go to plan. The second one is harder to write and much harder to fake.

Start with the evidence. If the customer cannot give you a before number, a date and a way to verify the after number, you can still publish a process story. You cannot publish it as a measured result.

How to write a customer case study: the eight blocks

Use these eight blocks in this order. Each block has one job, and the evidence column tells you what to collect before drafting. If a block has no evidence, either leave it out or label it as the customer's recollection.

Block 6 is the one teams skip. It is also the block that makes the other seven believable. A page where everything worked reads like a brochure; a page with one failed attempt reads like a report.

Which numbers you can publish, and which you cannot

Publish a number only when you can name its source, its time window and its unit in the same sentence. "Support tickets fell 38% in the eight weeks after the migration, according to the customer's helpdesk export dated 2026-09-30" is publishable. "Support improved dramatically" is not.

In the United States, the Federal Trade Commission's Endorsement Guides cover testimonials and case studies, including the need to disclose material connections and to avoid presenting unusual results as typical (FTC Endorsement Guides, read 2026-10-20). Google's guidance for helpful content asks for first-hand experience and original information, not recycled claims (Creating helpful content, read 2026-10-20). Neither source replaces written permission from the customer.

Ask for the export, not the summary. A summary has already lost the unit and the date, and those are the two things a skeptical reader checks first.

If the customer wants to stay anonymous, replace the company name with a checkable description such as "a 30-person cross-border payments team" and keep the industry, size and time window. Anonymity removes the identity, not the unit. Send the draft back before publishing and ask three questions: is the quote exact, do the numbers match the export, and which sentence would you rather not see public? The sentence they cross out is often the boundary the page needs.

The interview: seven questions

One call is enough if you prepare it. Ask for permission to record, send the transcript back for correction, and keep the customer's words for the quote block instead of rewriting them into marketing language.

1. What was happening before you started, and what did it cost you in time or money?
2. What event made you look for a change?
3. What did you try first, and why did it not work?
4. Walk me through what you did, in the order you did it.
5. What number did you watch, where did it come from, and when did you read it?
6. What part of the result would you not claim was caused by the product?
7. What should someone in your position do first, and what should they not copy?

Question six is the one that produces the limitation paragraph. Customers often volunteer it when asked directly, and it protects you from writing a claim they will later dispute.

Before recording, state the use: this case only, reviewed by the customer before publication, not passed to sales as a testimonial. Send the transcript within a day, while the details are fresh. Keep the customer's corrections. That is not proofreading; it is the customer taking responsibility for what they said.

What to do when the customer will not share numbers

Write a process case study and say so in the first paragraph. Describe the starting point, the steps, the elapsed time and the customer's own account of what changed. Keep the quote. Drop the percentage.

What we cannot tell you is how much of any change the product caused. A case study is not a controlled experiment. Even with a clean before-and-after number, seasonality, a pricing change or a second tool can move the same metric. If the customer did not measure a baseline and a comparison, write "the customer attributes this change to" and stop there.

That is not a weakness in the page. It is the difference between a page a reader trusts and a page a reader discounts.

Sometimes the customer will only give a range, such as "20 to 30 hours a month". Publish the range, but say it is the customer's estimate and name how many people reported it. Do not collapse a range into a precise number. A range that is honest is easier to trust than a decimal that is not.

Common questions

How long should a customer case study be?

Long enough to carry the evidence and short enough to read in one sitting. Most product case studies need one screen for the result, one for the work, one table for the numbers and one paragraph for the boundary. If the page needs a second interview to explain the first, split it.

Do I need written permission?

Yes, for the name, the logo, the quote and any number the customer supplied. Ask before drafting, not after publishing. Keep the approval in the same folder as the export it refers to.

Can I write one without an interview?

You can write a product story from public information, but you cannot call it a customer case study. Without the customer's account you have no trigger, no failed attempt and no quote. Label it as an example or a scenario instead.

How do I know the case study worked?

Use it as a sales and support asset with a trackable link, then read the referrals and the questions it answers. We do not have a universal benchmark for case-study performance, and we would not trust one that came without the page's traffic and sales context.

Next step

List the eight blocks and mark which ones you already have evidence for. Anything you cannot mark becomes an interview question, not a sentence to invent. When the evidence is complete, write the page. If the result is public news as well as a customer story, use the press release format for the announcement and keep the case study as the proof. The same discipline applies when you adapt a product description for every channel: one claim, one source, one place to check it. If you want the asset, the channel and the follow-up in one workflow, keep the promotion record in one place.