How to write a case study

The symptom

Thirty percent improvement. Eighteen months. A quote at the bottom from a delighted client saying the team was a pleasure to work with.

Every number in it is true. Legal cleared it. The client signed off. It sits on the site under a heading that says Results, and the design is clean.

And in all the time it’s been up, no prospect has ever asked a question about it. Nobody emailed to ask how you got there. Nobody brought it into a meeting. It gets read the way a plaque on a wall gets read, which is noticed and then walked past.

That same buyer will spend forty minutes on a reference call asking what went sideways in month four.

Why it keeps happening

Outcomes are the part that survives legal review.

They’re also the part the client will let you print. A number in a chart doesn’t require anybody to admit anything. Judgment is a different transaction, because judgment means saying what nearly went wrong, and that requires a conversation with the client nobody on either side wants to schedule. The account lead has a renewal coming. The client contact has a boss who was told the project went well. So the firm publishes the half it can clear and calls it proof.

We’ve published the version with the clean arc, more than once, because the clean arc was the version the client would approve. The approval process rewarded it. Nothing in a review chain ever flags a case study for being too smooth.

What gets cut in that process is the only section a buyer reads slowly.

What the buyer is doing at the Proof gate

Working out how the firm behaves under conditions that resemble theirs.

A result tells them the engagement worked once, inside a set of constraints they can’t see. What they’re mining for is that constraint set: what the firm was working with, what it traded away, where the decision points were and which way they went. Conditions are the thing a buyer can map onto their own situation. An adjective gives them nothing to map.

There’s a stage problem underneath this. Early on, proof can be directional and illustrative, because the buyer is still deciding whether the firm is worth an hour. Later it has to carry weight: numbers, documentation, references, process detail. Firms get it wrong in both directions, opening with documentation nobody asked for and then, once the committee is assembling its file, handing you the narrative you already read.

Nobody reads a spotless account and concludes the work was spotless. They conclude the failures were edited out, which puts every other claim on the page under suspicion.

The correction

Take one of your case studies and rebuild it. The rest of the library can wait.

Seven parts, in order: the situation, the stakes, the constraint set, the firm’s interpretation, the action taken, the result, the lesson. Most published case studies contain three of those, which is why they read like a receipt.

The interpretation is the load-bearing element and it’s the one most often missing. It’s the paragraph where the firm says what it made of the situation before it knew whether the read was right. Give it more room than the result gets. If your result paragraph is longer than your interpretation paragraph, you’ve written a scoreboard.

Then the move that does more work than all the others: include the decision the firm had to correct partway through. The read that turned out to be wrong, the point at which it became obvious, and what changed as a result. A difficulty that conveniently resolved itself doesn’t count. A case where nothing went wrong reads as marketing. The correction is where a buyer finds the answer to a question nobody puts in writing...how does this firm behave when it’s wrong?

Which means going back to the client and asking permission for the correction, not only for the outcome. The framing that tends to work is what the two organizations figured out together, because the client contact has the same defensibility problem the firm does. If confidentiality means stripping the name, keep enough sector, scale and constraint detail that the story stays credible. An anonymous case study with the specifics sanded off proves nothing about anybody.

Be specific about the tradeoffs. Keep hero language out unless something in the record supports it.

Worth borrowing from the field stories: they hold back the identity until the reader is invested, put the science in only after the narrative has earned it, and close on a details block that carries the sourcing honestly, including what the record doesn’t settle.

A case study built from outcome adjectives also compresses to nothing when a model summarizes it, where one built on a named constraint and a specific decision leaves something the summary can carry.

Before and after

Before. Facing margin pressure, we implemented a comprehensive operating model redesign that delivered a 30% improvement over eighteen months.

After. They had eleven weeks until the board met and one analyst who understood the cost model. We read the margin problem as a pricing problem and spent the first month there. That was wrong. The leak was scope creep on three accounts, and finding it cost four weeks we’d promised to the rollout.

What backfires

Writing up the correction as the client’s error. The buyer reads that as a preview of how the firm will describe them later.

Disclosing one tidy mistake inside otherwise flawless prose fails the same way, since a staged imperfection reads as a technique. And if the correction doesn’t carry the reasoning that produced it, you’ve published evidence of a mistake with no evidence of judgment.

Run this on your own material

Open the case study your firm is proudest of. Find the sentence that describes a decision that could have gone the other way.

A decision, with alternatives that were live at the time, made by people who couldn’t see how it would turn out. Challenges and milestones don’t count.

If there isn’t one, what you have is a testimonial with a chart in it. Fine as a testimonial. Adding more results won’t fix a document that never shows anyone thinking.

Do this before the next proposal goes out, since that’s where the case study is going to end up.

Signals you’ve cleared it

Advance when prospects quote the constraint back to you rather than the number. When someone asks what you’d have done if the timeline had been half as long. When the correction section is the part that comes up on the call.

Hold when the questions keep circling capability. That means the material demonstrated an outcome and left the judgment question open, and sending another result won’t close it. Hold too when the only response the case study gets is congratulation. Congratulation means it was read as a claim about you. It was supposed to be read as information about them.

Where this connects

Story from the Field · The $4.3b Man Crush, on a decision that was logically airtight and wrong about people.

Research Library · Entry 01, bounded rationality.

Next move · Go to How to know when proof arrives too late.