How to write a freelance case study
A case study is a story with a receipt. It tells a prospective client three things, in order: the situation a past client was in, what you actually did about it, and a result that client confirms. Written that way — five short blocks, the client's own numbers, their words at the end — one honest page outsells a gallery of logos, because the reader is not admiring your past. They are checking whether your past looks like their future.
This guide is about the writing. Getting the raw material — asking at the right moment, the three questions, permission in writing — is covered in our guide to how to get testimonials and case studies. The two are halves of one job: that guide fills the notebook; this one turns the notebook into a page.
Pick the right project
Not every good project makes a good study. You need three things at once: a problem the next client recognises as their own, a result you can state without squinting, and a client willing to confirm both in public. A technically impressive project nobody else buys teaches the reader nothing. A modest project with a familiar problem — the backlog, the stalled website, the inbox nobody owns — is the one that gets read to the end.
Pick toward the work you want more of. A case study is an advertisement for a kind of project, and it will attract more of that kind — so build the set around the direction you are steering the business, the same evidence-first logic as choosing a niche. Three studies aimed at your next client beat a shelf of studies for work you have stopped wanting.
And settle permission before you write a word. The conversation is the one the testimonials guide sets out: the client agrees the level — named, with name, role and company; anonymised, described by role and type of business; or not at all. Treat the permission as three separate yeses — the name, their words and the figures — because a yes to one is not a yes to the others, and the draft goes back to the client for approval of the facts before anything publishes. Writing first and asking later wastes the writing whenever the answer is no, and publishing on a verbal yes is how a happy client becomes a surprised one. The time to plant the seed is the delivery itself: a good wrap-up email names the result while it is fresh and promises the check-in at which you will ask.
The structure that works: five blocks
Every strong one-page study follows the same skeleton. Five blocks, each with one job:
1. Situation. Who the client is and what normal looked like before you arrived — two or three sentences of scene-setting. Its job is recognition: the reader should be able to say "that is a business like mine" before anything goes wrong.
2. Complication. What changed, or what broke — the cost of the situation in time, money or dread. This is the block everyone skips, because it feels like dwelling on the client's failure. It is also the block readers recognise themselves in, because nobody hires help for a situation that is merely untidy. Skip it, and the result further down has nothing to be measured against. Close the block with the decision, in a line: why they hired help at all, and why they chose you — the referral, the specialism, the smaller job that went well. Writers skip it because it feels like boasting; readers hunt for it, because choosing is the task they are in the middle of — which is why the proposal template guide puts its proof at the decision point rather than in an appendix.
3. What you did. The work itself, plainly: what you built, fixed, wrote or cleared, in the order you did it. Specific beats decorative — "rebuilt the invoice run" persuades; "delivered a comprehensive solution" does not. Keep it the shortest block that stays truthful: the reader is checking competence, not reading a manual. Show the shape of the method, never the full recipe — the step-by-step how is what the next client pays for, and a study that gives the reusable detail away has crossed the same line as the spec work described in handling free work requests: the result and the outline are the proof; the recipe stays yours. Where you faced a real choice — rebuild or patch, fast or thorough — name it and the reason in a line: prospects read this block for judgement, and a decision explained persuades more than a task listed.
4. Result. What is different now, stated the way the next section insists on: the client's confirmed figure, or the concrete before-and-after state. One solid result outranks three vague ones.
5. Their words. One quote, the client's own, approved for publication — the recommendation line from the testimonials guide's third question is usually the strongest candidate. It closes the study because it converts your account into their testimony: everything above it is your claim; this block is their signature.
Numbers, honestly
Use the client's numbers, confirmed by the client. If the result is a figure — invoices paid, hours saved, a queue cleared — it comes from their records or their mouth, exactly as they stated it. Never round a client's result up: "about a third" does not become "40%", and a figure from one good month does not become the typical month. Never borrow, either: a result from a different project, a different client or a different year has no business in this study, however similar the work felt. And some figures never enter a study at all: the ones a client shared in confidence so you could do the work — their margins, their own rates, their customer details — stay out, however warm the yes to the study itself. It is the same confidentiality discipline a proper project handover document works by: what you were trusted with is not yours to publish.
Where the result is not a number, use the before-and-after state. "Invoice backlog cleared in the first month, and nothing has been older than 30 days since" is a fact pattern a reader can picture and a client can confirm. "Transformed their business" is fog — it could mean anything, so it persuades nobody who reads carefully, and careful readers are the clients worth having. The test for any result sentence: could the client read it over your shoulder and nod? If they would frown, qualify or laugh, it does not go in.
That is the standing rule for all client proof on this site, and a case study is where it matters most, because a study asks to be believed at length. A study that would survive a phone call to the client is an asset. One that would not is a liability with your name on it.
A worked miniature
Here is the whole structure at its smallest. The example is fictional — the bookkeeper, the client and every number are invented — shown so you can see the shape, the size and the honesty rules working together. The whole study runs to 150 words:
Complication: By the time bookkeeper Ana Reyes was taken on, 38 invoices worth $21,400 sat unpaid, finished jobs were still unbilled in email threads, and in two of the previous three months the owner had covered wages from a personal account.
What she did: Ana rebuilt the run end to end: every finished job invoiced the same week, a weekly aged-debt list reviewed every Monday, and polite chase emails sent on a fixed ladder.
Result: Inside the first month, 31 of the 38 invoices were paid — $17,900 of the $21,400 — and the remaining seven, worth $3,500, went onto agreed payment plans. Three months on, no invoice is older than 30 days.
Their words: "I stopped dreading the bank balance. Ana just fixed it." — Tom Hales, owner, GreenScape Landscaping
Notice what the miniature refuses to do. The 31 paid invoices are reported, not inflated into all 38. The payment-plan remainder is stated, not hidden. And the quote claims a feeling, while the numbers claim the facts.
Two further mistakes survive even honest numbers. The first is casting: in a working study the client is the hero and you are the guide they hired. A study written as your own success story leaves the reader a spectator, when its whole job is to let them sit in the client's chair. The second is burial: readers skim, so the result cannot live only in the fourth block — the opening lines of the published study should carry it too, the way this page's first paragraph states its result before its structure.
Where it lives, and how it gets used
Give each study its own page on your site or portfolio, beside the service it proves — the same placement rule as testimonials, because a reader considering bookkeeping wants the bookkeeping study, not your greatest hits. Link it from the matching line of your rate card as well, so the proof sits beside the price it justifies. One page per study also means one link per study, which is what makes the next two uses possible.
In a proposal, link the one relevant study; do not attach five. A proposal is read under time pressure, and a single well-matched study says "this exact problem, solved", while a bundle says "homework". The proposal generator leaves the credibility slot where the link belongs, and the proposal template guide shows the about-you section it sits in — reassurance at the decision point, one click away, at the reader's choice. Keep the link in the proposal skeleton in your template library, and picking the matching study becomes a swap, not a search.
After a discovery call, the follow-up email is where a study does quiet work: two lines of recap, then "the closest thing I have done to your situation is this — the short version is on this page". It answers the question the client was too polite to ask on the call: have you actually done this before? The discovery call script covers the call itself; the study is the evidence you send after it.
Keep them fresh
A study has a shelf life, because you do. Work written up three years ago sells the practitioner you were then — the older tools, the simpler problems, the lower rate. Once a year, read the set through and retire or rewrite anything that no longer describes the work you sell now. The natural moment is the same sitting as the quarterly review, where the stop-doing decision gets made: a study for a service you have decided to stop is not proof, it is advertising for a closed shop. And fresh studies get easier to produce each year, because the projects you finish this year are next year's material — provided the permission habit is already in place. Date each study when it publishes, and let the date show: an undated result from three years ago, presented as your current form, is a small dishonesty a careful reader catches. Keep the live set listed with your other standing records — the habit behind documenting your processes — so the annual read-through actually has a list to work from.
The honest limit
One limit deserves stating plainly: a shelf of case studies shows your wins, because wins are what clients approve and prospects ask to see. That is the nature of the shelf, not a scandal — as long as you remember the missing half exists. The projects that went sideways still teach, but privately: the project retrospective is where a hard project gets mined for what it can fix, away from public view. And a study a client has not approved is not a study at all. It is a draft, and it stays a draft until they say yes.
Questions
Related: the receipt for your best claim
A case study is not decoration; it is the receipt for the claim your proposal makes — we did this, and here is what changed. Write it in five blocks, let the client confirm every number and every word, link the one study that matches the job in front of you, and keep the set as current as the work itself. Do that, and your proof stops being a gallery and starts being an argument: the next project, half-sold before the proposal is even opened.