How to Write a B2B Case Study: The Complete Process (2026)

The short answer: there are eight steps, and only three of them are writing. Pick a customer who matches a segment you sell into, get their agreement, interview them, confirm the result, draft it, get sign-off, publish it as a web page, and distribute it. Nearly every guide to this expands the middle and compresses the ends. That is backwards, because the ends are where case studies die.

Why most guides get the emphasis wrong

Read the top-ranking guides on this subject and you will notice something odd. They devote most of their length to structure, headlines, and narrative arc, and dispose of "conduct your interview" and "get approval from your customer" in a line each.

That distribution is exactly inverted from where the time goes.

A 2024 survey of 115 SaaS customer marketers by Uplift Content put around 12.7 hours of active work into a case study, spread across roughly two months. It also named the top three blockers: finding customers willing to participate, identifying metrics, and getting approval. Not one of those is writing.

Twelve hours of work stretched over two months is not an effort problem. It is a waiting problem, and it is caused by the two steps the guides skip.

Put the two distributions side by side and the mismatch is stark:

Forty-five percent of the advice is aimed at five percent of the delay. The three bolded rows in the right-hand column carry roughly 85% of the elapsed time between deciding to do a case study and having one live.

(Word-count shares are our read of the top-ranking guides for this query, not a measured study. The elapsed-time shape follows from the Uplift survey's blocker findings.)

There is a second oddity worth naming. The page currently ranking first for this question is published by a video testimonial company. Nothing wrong with that, but it tells you how little of this category's advice comes from people whose job is the written asset.

The eight steps, honestly costed

Look at the elapsed-time column. Three rows carry almost all of it, and only one of them is a step most guides bother to explain.

Step 1: Choose a customer who matches a segment you sell into

The default is to pick your happiest customer, and it is usually the wrong instinct. Your happiest customer is often your oldest, smallest, or most unusual, and a prospect reading their story cannot see themselves in it.

Start from the segment you want more of. Which industry, which company size, which use case is your pipeline weighted toward right now? Find a satisfied customer inside that box, even if they are less effusive than someone outside it. Recognition beats enthusiasm.

Two practical filters. Pick someone who can name a specific before state, because a customer who cannot describe what things were like beforehand will not produce a usable story. And ask more people than you need — if you want three stories this quarter, ask six, because some will go quiet, some will hit a legal wall, and some will simply be busy. Treating it as a funnel rather than a set of individual requests is the single most useful reframe in this whole process.

Step 2: Make the ask small enough to say yes to

This is the first place case studies die, and the cause is almost always that the ask is bigger than it needs to be.

"Would you be willing to do a case study with us?" asks the customer to commit to an unbounded amount of work: a call to schedule, questions to prepare for, a draft to review, an internal approval to chase. Most people cannot estimate that, so they defer, and deferral looks identical to refusal.

Three things reliably help.

Name the exact time cost. "Twenty minutes, whenever suits you" is answerable. "A case study" is not, so put the number in the email that makes the ask.

Remove the scheduling. The single biggest reduction in friction is not making the conversation shorter, it is removing the calendar from it entirely. A link the customer can open at 9pm on a Tuesday will get taken up by people who would never have found a mutual slot.

Get permission at contract time. Experienced teams add a line to the initial agreement saying the customer may be quoted in marketing subject to their review. It costs nothing when nobody is emotionally invested, and it means legal is not assessing the question from scratch a year later.

For the full version of this step, see how to get customers to say yes.

Step 3: Interview for material, not for quotes

The purpose of the interview is not to collect a testimonial. It is to gather more material than you need, so you can cut a case study, a customer story, and several pull quotes from one conversation without going back.

The questions that produce usable material are almost all about the before state. People are poor at articulating their improvement and excellent at describing what their job used to be like. Ask what this used to take, who else was involved, how often it went wrong, what they stopped doing, and whether there was a moment they noticed things had changed.

That last question is the highest-yield one in any case study interview, because anecdotes are specific by nature.

Listen especially for a plain-language before and after — "it used to take about six weeks, now it's around nine days." That is a number, and it came from them without you asking anyone to calculate anything. The full question set is at case study interview questions.

Step 4: Confirm the result, and accept it if there isn't one

Sometimes there is no number, because nobody measured the before state or the customer is not allowed to share it. This is common and it is not a failure.

What you must not do is invent one, or reach for the softer version and write "significantly reduced." Both read as a number-shaped hole, and the second is worse because it pretends. How to write a case study when the customer will not give you a number covers the qualitative path in full.

Step 5: Write it — the easy part

Two or three hours, and the structure is well established: the hook, who the customer is, the before state, why they chose you, the result, quotes throughout, and a close. The template that converts sets this out section by section.

Three rules worth repeating because they are where drafts go wrong.

Give the before state more room than the solution. The before state is the only part a prospect can recognise as their own, and recognition is what makes them keep reading.

Do not polish the quotes. A sentence that has been tidied into marketing language reads as manufactured, because it is. Slightly awkward real speech is the most credible thing in the document.

Write about the customer, not the product. A case study that reads as an extended feature list gets skimmed. The product should appear as the thing that changed, not the subject of the piece.

On length: 600 to 1,200 words covers most B2B software stories. There is no evidence for an optimal figure — Ahrefs' analysis of 174,048 pages cited in AI Overviews found the correlation between word count and citation position is 0.04, effectively zero. Specificity is the lever, not length.

Step 6: Get sign-off — the second place case studies die

Approval routinely takes longer than every other step combined, and the reason is that it is asynchronous and unowned. You send a draft, the customer forwards it internally without telling you, and from your side it looks like silence.

Two fixes that experienced teams reach independently. Bring the customer something nearly finished rather than a request for help, because correcting a draft is a ten-minute favour and helping build one is a project. And attach a visible source to every number, because what holds drafts up in legal review is the figures and the wording, and an unsourced figure invites a question by default.

The case study approval process covers this step properly, including what to do when the customer goes quiet.

Step 7: Publish it as a page, not a PDF

Publish each story as its own indexable web page. A PDF cannot be found by search, cannot be cited by an AI assistant answering your buyer's question, cannot be linked to from a product page, and is unpleasant on a phone.

There is a maintenance argument too, and it is the one that bites hardest. If the case study is a design file, then every time a figure changes you are editing two artefacts and re-exporting both. Keep each story as structured content in one place and treat the page, the deck slide, and the PDF as exports of it.

Then place it where the doubt lives: on the relevant product and solution pages, on the pricing page, and in a filterable library. How many case studies you need and where they belong covers the placement question.

Step 8: Distribute it in the same pass

Distribution is the step that gets scheduled for after publication and therefore never happens. By the time the story is live the team has spent weeks on it and moved on.

The fix is sequencing rather than discipline: cut the LinkedIn post, the short social post, the pull quote and the sales email while the material is still in front of you. How to repurpose a case study without doing the work twice covers which formats are worth cutting.

What actually changes the output

If your team publishes four case studies a year and wants twenty, the lever is not a better writing process. Steps 2, 3 and 6 hold almost all the elapsed time, and all three are permission and scheduling problems.

That is the specific thing StoryVoice was built for. The ask becomes a link rather than a meeting: your customer talks to a voice AI for about five minutes, on their own schedule, from any device, with no call to book — which collapses steps 2 and 3. The publish-ready story is written from what they said the moment they finish, with any headline figure computed from a before and after they described in their own words. Then it goes back to them on a link needing no login, where they can approve it, request a change, or edit it themselves, and every metric card opens the passage of the recording it came from so a reviewer can check a number rather than query it — which is step 6. Every interview also returns a LinkedIn post, an X post, a pull quote and a sales email, so step 8 happens in the same pass.

What it does not do is design. If your four hours per case study go into layout rather than words, that is a template problem and a different tool.

Send your first interview link free — no credit card required, first case study free.

Production time and blocker data is from a 2024 Uplift Content survey of 115 SaaS customer marketers. Word-count and citation data is from Ahrefs' analysis of 174,048 pages cited in Google AI Overviews. Where no credible research exists — an optimal case study length, participation rates — this article says so rather than estimating. Product behaviour described for StoryVoice is current as of 2026-08-16.