How to Write a Case Study When the Customer Won't Give You a Number

The short answer: write it without the number. A case study persuades when a reader recognises their own situation in it, and specificity does that job better than a statistic does. Use the concrete before state, a named person, real quotes, and honest detail about how the work went. What you must not do is invent a figure, or dress a missing number up in words like significant and dramatic. Both are more damaging than the empty space they fill.

Why won't the customer give you a number?

It is worth understanding which of these you are dealing with, because three of the four are not about you at all.

They are not allowed to. Public companies, regulated industries, and anyone in the middle of a funding round or an acquisition often cannot disclose operational figures. This is a policy decision made above your contact's head and no amount of rapport changes it.

They never measured the before state. This is the most common reason and the most under-appreciated one. Your customer knows things are better. They do not know by how much, because nobody was tracking it when it was worse. Asking "what was your conversion rate before?" gets you an apologetic shrug, not a number.

They have a number but do not trust it. Internal dashboards disagree. The figure came from a spreadsheet somebody left the company maintaining. Attaching a name and a logo to a number you privately doubt is a career risk, so the safe answer is to decline.

The improvement is real but tangled. They adopted your product the same quarter they hired two people and changed their pricing. The result is genuine and they cannot honestly attribute it to you alone. A customer who says this is being rigorous, not difficult.

Notice that pushing harder helps in none of these cases. What helps is changing what you ask for.

The second row is the most common and the most fixable. A customer who cannot give you a percentage can almost always tell you what a Tuesday used to look like.

What to ask instead

The single most useful shift: stop asking about the improvement and start asking about the before state.

People are bad at knowing their percentage gain and remarkably good at describing what their job used to be like. "What did a Tuesday morning look like before?" gets you an answer. "What was your efficiency gain?" does not.

Questions that produce usable material without requiring a metric:

• ●What did this used to take, start to finish?
• ●Who else had to be involved, and what were they doing instead?
• ●How often did it go wrong, and what happened when it did?
• ●What did you stop doing once this was in place?
• ●What was the thing that made you finally go looking for a solution?
• ●Is there a moment you remember where you noticed it was different?

That last one is the highest-yield question in a qualitative interview. Anecdotes are specific by nature, and a specific anecdote is more convincing than an unsourced percentage.

The pattern in the right-hand column: every question asks about the past, in concrete terms, and none of them asks the customer to calculate anything.

There is also a quiet arithmetic opportunity here. If a customer says onboarding used to take about six weeks and now takes around nine days, you have a number, and it came from them. You did not ask them to calculate anything. This is the pattern worth listening for: a before and an after described in plain language, from which the figure follows.

Why inventing a number is the worst option

It is worth being blunt about this, because the temptation is real and the reasoning is usually well-intentioned. The story feels thin. A plausible, conservative figure would make it land. Nobody is going to check.

Three problems.

It fails in the exact situation you built it for. A case study earns its cost in a sales conversation with a sceptical buyer. That is precisely the moment someone asks "how was that measured?" A number you constructed cannot survive the follow-up question, and the failure is public and expensive.

Your customer reads it. They know what they told you. Seeing a figure they never gave you, attributed to their company, over their name, is how you lose a reference. It is a particularly bad way to lose one, because you lost it while trying to celebrate them.

Buyers are now primed to disbelieve fluent, confident, unsourced claims. Two years of AI-generated marketing copy has trained people to treat polished specificity with suspicion. An unattributed statistic in a case study reads differently in 2026 than it did in 2022. A traceable claim is worth more than an impressive one.

The same logic rules out the softer version of the same move. "Significantly reduced" and "dramatically faster" are not honest alternatives to a missing metric. They are a number-shaped hole with an adjective in it, and readers hear exactly that.

How to write the story without the number

The structural point: every row on the right is more specific than the row on the left, and none of them contains a number. Specificity and quantification are not the same thing, and it is specificity that persuades.

The rest of the structure is unchanged from any good case study. A named person with a real job title. A before state a reader can recognise. What they tried first and why it did not work. What actually changed. A quote that sounds like speech rather than copy. If anything, a qualitative story needs the quote to work harder, so resist the urge to tidy it into marketing language. Slightly awkward phrasing reads as real.

When there is genuinely no story

Sometimes the interview produces nothing publishable. Not a modest result, but no endorsement at all: the customer is neutral, or says it is much like any other tool they have used, or their positive comments are about something you did not do.

The honest answer is not to publish. A case study assembled from a lukewarm interview is worse than no case study, because a buyer can feel the absence of substance and the customer can see their real experience misrepresented as enthusiasm.

This is worth building into your case study process rather than leaving to willpower at the end of a long week, when a blank page and a deadline make manufacturing something feel reasonable.

StoryVoice makes it a product decision. Before writing anything, it judges whether the interview contains a genuine, product-specific endorsement. If it does not, it declines to produce a story rather than manufacturing one. The marketer sees the full transcript so they can see exactly why, and the interview does not count against their monthly quota, so the honest outcome costs nothing. A modest but real positive still produces a story, and reads as genuinely positive but modest rather than an inflated triumph.

What StoryVoice does when the numbers aren't there

The qualitative path is a first-class output, not an error state.

If an interview produced no quantified result, StoryVoice emits an empty metrics list. The metric cards simply do not render, and the story runs on the customer's own words and quote. It never fills the space with a plausible figure.

Where the customer does state a before and after in plain language, StoryVoice does the arithmetic rather than asking them to. Someone who says onboarding took six weeks and now takes about nine days gets the computed headline, derived from their sentence rather than supplied by a marketer. And every metric card opens the passage of the recording it was read from, so the customer, her legal team, and anyone reviewing it can check the claim against what was actually said. If a figure is edited afterwards, the card says so and keeps the original passage, so a number never silently keeps a claim it has lost.

That last part matters more than it sounds. The reason to care about traceability is not compliance theatre. It is that a claim someone can check is a claim they will believe.

Run your first interview free — no credit card required. Or see the questions that get specific answers when a customer can't give you a number.

This article describes StoryVoice product behaviour current as of 2026-08-16. It contains no third-party statistics; where we could not verify a figure, we left it out rather than estimating one. Nothing here is legal advice.