The definitive guide to case studies

The promise a case study makes, and why it is different

Every other piece of sales material makes an argument. A case study makes a different kind of claim entirely: this happened, to a real business, and here is the proof. A reader approaching a case study is not asking “is this well argued?” the way they would with a deck or a sell sheet. They are asking “is this actually true?” And the moment they sense it might not be, the document does not just fail to convince them. It teaches them to distrust every other number on your site, because if this one was softened, rounded or invented, why would the rest be any different?

This is why a case study with an invented number is worse than having no case study at all. A missing proof point is an honest gap, and a buyer can decide for themselves how much that absence matters. A fake proof point is a trap waiting for exactly the buyer who checks hardest, which, inconveniently, tends to be the buyer closest to actually signing. The customers least likely to take your numbers on faith are the ones who do the most diligence before a purchase. A fabricated case study does not just risk getting caught; it specifically risks getting caught by the audience whose trust was worth the most.

The discipline, stated as a rule rather than an aspiration

Here is the whole craft, compressed into one sentence. If the number is not in your source, it does not appear. Not a rounded version of it. Not a plausible estimate standing in until the real figure shows up. Not “roughly” or “approximately” dressed up to sound precise. If you do not have the real number, confirmed, from somebody who can stand behind it, the slot where that number would go either stays visibly empty, flagged as a gap rather than hidden as an oversight, or the case study simply is not ready to publish yet.

This sounds almost too simple to be a discipline worth naming. It is not simple in practice, because the pressure runs entirely the other way. A thin case study with a visible gap looks unfinished. A business under pressure to show proof, any proof, before a launch or a pitch will feel a real pull to round “somewhere around a third faster” into “35% faster,” or to write a quote that captures the spirit of what a happy client said rather than the client's actual words, because the polished version reads so much better. Every one of those small moves is the same mistake wearing a different disguise: trading a true, checkable document for a persuasive, unverifiable one.

The discipline holds the line exactly where the pressure is strongest. It is not a rule about being modest. It is a rule about which kind of document you are building, because a case study that has been improved past what the source actually supports has quietly become a different thing, a testimonial written in the third person wearing a case study's clothes, and it no longer does the one job a case study exists to do.

Permission is a separate question from truth, and both have to clear the bar

A case study can be entirely accurate and still be unpublishable, because truth and permission are different questions. Being named in a private pitch deck, a proposal, or on a sales call is not the same disclosure as being named on a public page anyone can find, and it is a genuine mistake, made more often than it should be, to treat “we have mentioned this client before” as though it already settled the public question.

The honest version of this process asks for permission as its own explicit step, separate from the writing. Who exactly is this client. What specifically is being asked: being named at all, a described scope of work, a number, a direct quote, a logo, because a yes to one of those is not automatically a yes to the others. And does the permission still hold today, for this exact wording, or was it given for a different version of the story at a different point in the relationship.

A client who was comfortable being named informally two years ago, when the relationship was warm and the engagement was recent, may feel very differently about a page that will outlive the relationship, get forwarded to strangers, and sit on your site indefinitely. The honest move is to ask again, specifically, for what you are actually about to publish, and to get sign-off on the exact final wording, not the brief, not the interview notes, the actual sentences.

Challenge, approach, outcome, and why outcome keeps losing

The structure a case study should follow is almost universally known and almost universally mis-weighted: challenge, approach, outcome. The problem is not that anyone disagrees with the order. It is that the three sections do not get equal attention, and they lose it in a predictable direction, toward approach and away from outcome.

This happens for an understandable reason. Approach is the section where the business gets to describe its own work, in its own voice, at whatever length it likes, and it is naturally the part the people writing the case study are proudest of. Outcome, by contrast, requires somebody outside the business to confirm a number, which takes an extra step, an extra conversation, sometimes an extra few weeks of waiting for a result to exist at all. The path of least resistance is to let approach expand to fill the space outcome should have occupied, and the result reads like a portfolio piece dressed as proof: a confident account of what was done, thin or silent on what it actually achieved.

The fix is structural, not stylistic. Decide, before writing a word, roughly how much of the page each section earns, and protect outcome's share deliberately. A buyer skimming the page is not reading to admire the process. They are reading to find out if it worked, for somebody like them, in a way they can check.

And be clear what counts as an outcome. “We redesigned their site” is a description of the work. “Their quote-to-close time dropped from eleven days to four” is a result. A number, a named change, a measurable before and after. If the only thing you can say is that the client was pleased, you have a testimonial, which is a fine thing to have and a different document.

The quote test

A quoted client is doing real work in a case study. It is the one place in the document where the voice is not yours, which is exactly why it carries weight a third-person claim does not. That only holds if the quote is genuinely the client's own words, not a line drafted on their behalf and sent for a rubber-stamp approval.

A useful, honest test: read the quote out loud, and ask whether it sounds like something a person actually said in conversation, or like a sentence built to contain a keyword and a happy adjective. “We saw a 40% increase in qualified leads within the first quarter” is a sentence a marketing team writes. “Honestly, I did not think this would work as fast as it did” is a sentence a person says. Both might be true. Only one of them reads like a human being said it, and a reader can tell the difference even when they cannot articulate why.

The gap, handled honestly

Every case study, at some point, hits a spot where the real number is not available yet. The client has not measured it, the engagement is too recent, or nobody asked the right question at the right moment to capture it. The answer is not to avoid the gap by softening the claim into something vague enough that nobody can check it. It is to mark the gap, visibly, and either leave it flagged or hold the whole piece until the number exists.

A flagged gap costs you something real: the document looks less finished, and a stakeholder under pressure to publish something will feel that cost directly. It still costs far less than the alternative, because a reader who notices one invented number stops trusting every other number on the page, and once that trust is gone it does not come back by apologizing for it later. An honest gap is a cost paid once, up front, in public. A discovered fabrication is a cost paid later, in a way you do not get to control the timing of.

Start the hunt the day the deal closes

This is the one closing-set row whose gathering can begin before the rest of the set exists, because its real dependency is a happy customer rather than a finished brand kit. The logo kit and the asset library style the finished page; they do not gate when you can start asking.

So ask early. Which of your customers would actually say yes if you asked to tell their story publicly, not who you wish would say yes. What was broken, missing or costing them time before you started, in their own words rather than yours. What is the one number that proves it worked, and who specifically could confirm that number if a stranger asked. If you do not have the number yet, who is the right person to ask, and does this ship with a flagged gap for now or wait until the number exists. Who at the client has to approve the exact final wording. Six questions, asked while the engagement is still warm, is the difference between reporting a story and reconstructing one.

A worked gallery: patterns, not real businesses

Every business and number below is explicitly fictional, built to illustrate one craft failure or success each. None describes a real client, a real engagement or a real result. That is not a disclaimer bolted on; it is the same rule the guide just taught, applied to the guide.

1. RoofClear Exteriors, the approach that ate the page. A roofing contractor's case study opens with four paragraphs on crew process and materials before a single sentence states what changed for the customer, and that sentence is unconfirmed: “the homeowner was very satisfied.” Real craft, described in detail, with nothing checkable behind it.
2. Driftline Analytics, the rounded number. The case study claims the client “cut reporting time by half.” The real, confirmed figure was a drop from eleven hours a week to seven: a genuinely good result, but not literally half. Rounding a real figure into a cleaner one is the same category of error as inventing one, because the number printed is no longer the number that was true.
3. Marrow & Co., the quote that reads like copy. The client quote reads: “their innovative, results-driven approach exceeded all our expectations and delivered exceptional ROI.” No specific detail a person would plausibly say out loud. Compare: “I told them I did not believe the timeline, and they hit it two days early.” Less polished, far more convincing, because it sounds like somebody remembering a specific thing that actually happened.
4. Harrow Legal Partners, the honest gap done right. The results section reads: “the client has not yet measured this engagement's impact on caseload capacity; we will update this page once they have.” Visibly less impressive than its competitors' pages, and the only one of the four a careful buyer could actually trust.
5. Briarcroft Dental Group, permission asked for the wrong thing. The page publishes a client's name, logo and revenue figure on the basis that the client agreed, a year earlier, to be “mentioned as a reference in sales conversations.” The client never agreed to a public page with a dollar figure attached, and is blindsided when a prospective patient mentions reading it.