Favors.dev
A startup case study is one customer's before-and-after story, structured as situation, friction, change, result, and published on your own site with their permission. That is the entire form. You need exactly one good one, not a library, and you can produce it with three customers, no design team, and a twenty-five minute call.
Nearly every guide to writing one assumes you are a marketing team at a company with a hundred logos, a customer-advocacy program, and a designer who lays out the PDF. The templates you can download are slide decks: title page, executive summary, background, objectives. None of that is the hard part. The hard part is getting a real customer to say something specific enough to be worth publishing, and knowing what to do when the honest answer is that they save a couple of hours a week.
This is the version for that founder. The six questions, the fill-in-the-blanks structure, the rule for handling unglamorous numbers, and the five places one story earns its keep.
Why one story out-converts ten testimonials
A testimonial asserts an outcome. A case study demonstrates a mechanism. That difference is why one detailed story moves more skeptical buyers than a wall of five-star quotes: the skeptic is not asking whether anyone likes you, they are asking whether their particular mess resembles a mess you have already fixed.
Different proof assets do genuinely different jobs, and it is worth being precise about which one you are short of:
| Proof asset | What it proves | Effort | What it can't do |
|---|---|---|---|
| Star rating | That someone bothered. | Minutes | No mechanism. A five-star average tells a skeptic nothing about whether it works for them. |
| Testimonial quote | That a named human vouches for you. | One email | It asserts an outcome without showing the path to it. Easy to discount as cherry-picked. |
| Case study | That a situation like the reader's changed, and how. | One interview | Only covers one situation. You need the right one, which is why choosing the customer matters more than volume. |
| Usage numbers | Scale. | A query | Meaningless below a few thousand users, and it never answers the question the buyer is actually asking. |
There is a second reason to invest in the long form in 2026, and it has nothing to do with your landing page. Answer engines are now a real discovery surface, and they do not all cite the same things: the citation analysis in 5W's State of AI Citations 2026 found Wikipedia and Reddit alone supplying more than a quarter of US ChatGPT citations, while Perplexity leans on attributed, named-author pages and pulls from roughly three times as many sources per answer. A star rating on a directory cannot be cited as evidence that your product solves a specific problem for a specific kind of team. A named, dated, detailed customer story on your own domain can.
If you are still collecting the raw material, start with how to get testimonials for a new product and how to get your first reviews. A case study is what you build when one of those short quotes turns out to have a story behind it.
Which customer should you feature?
Pick the customer who passes three tests, in this order: they used the product recently, they have already said something positive unprompted, and they look like the buyer you want more of. Recency matters most, because memory for the before state decays fast. Ask someone in month nine what their old workflow felt like and you get a summary, not a scene.
With three customers you are not really choosing, you are checking which one can name a specific moment. The one who says "the week we onboarded two people at once and I realised the spreadsheet was going to break" is your case study. The one who says "yeah, it's great, much better" is a testimonial, and that is fine. Different asset, different job.
Three customers to skip, even when they are tempting:
- The biggest logo you can't get approval from. A story stuck in someone's legal review for six weeks is worth less than a published story about a two-person agency.
- Your friend. Readers can smell it, and you will soften the friction section to avoid embarrassing them, which removes the only part that persuades.
- The outlier. The customer whose results you could never repeat sets an expectation you will spend the next year failing to meet.
The six-question interview script
Book twenty-five minutes, ask permission to record, and ask these six questions in this order. Do not send them in advance. Prepared answers come back in marketing language, and the specificity you need lives in the unrehearsed ones. Your only job during the call is to keep saying "can you give me an example?" every time an answer turns abstract.
What were you doing about this before you found us?
Establishes the situation in their words, not your category's words. You are listening for the workaround: the spreadsheet, the intern, the three tools taped together, the doing-nothing-and-hoping.
What finally made you go looking for something else?
The single most valuable answer in the interview. Almost nobody shops on a normal Tuesday. There was a trigger, and that trigger is the exact moment your future buyer is living through right now.
What almost stopped you from signing up?
Their objection is your reader's objection. A case study that names the hesitation and then shows what happened to it converts skeptics far better than one pretending nobody hesitates.
Walk me through the first week. What did you actually do?
This is where you get the mechanism, and mechanism is what separates a case study from a testimonial. Push for concrete steps until you could re-enact the week yourself.
What is different now that wasn't before?
The result, framed widely enough that it works even without a metric. Stopping a Sunday-night dashboard check is a publishable result. Let them define the win before you go looking for a number.
Who should not buy this?
The disqualifier. One honest line about who this is not for buys more credibility than three more paragraphs of praise, and it pre-filters leads that would have churned anyway.
Six questions fill twenty-five minutes comfortably and produce roughly two thousand words of transcript. You will publish maybe six hundred of them. That ratio is normal, and it is the reason the interview cannot be replaced by a feedback form: the good line is almost always the aside, the thing they said after they thought they had finished answering.
The fill-in-the-blanks template
Here is the whole structure. Every bracket is something the interview gave you. If a bracket is empty, you did not push hard enough on the matching question, so go back to the transcript before you go back to the customer.
The template
- Headline
- How [customer] [did the thing] with [product]. Name the customer and the outcome. No wordplay.
- Summary (2 sentences)
- [Customer] is a [size, kind of company] doing [job]. After [change], they [result].
- Situation
- Before [product], [customer] handled [problem] by [the workaround, described concretely]. That worked while [condition held].
- Friction
- Then [the trigger event]. The old approach cost them [time, money, missed thing, or stress, in their words].
- The change
- In the first week they [step one], [step two], [step three]. The part that surprised them was [detail].
- The objection
- They almost didn't start because [hesitation]. What actually happened was [outcome of that fear].
- Result
- Today, [what is different], plus [a real number with its source, or an honest qualitative change]. Verbatim quote goes here.
- Fit note
- This is likely to work if [conditions]. It probably won't help if [disqualifier from question six].
Two rules govern the filling-in. First, keep the customer's nouns. If they said "the Sunday spreadsheet", do not upgrade it to "their legacy reporting process". The specific noun is the thing a reader recognises from their own week. Second, publish the objection. Almost no founder does, and it is the highest-return paragraph on the page, because it is the only one written in the reader's current emotional state.
Then send the finished draft to the customer before it goes live, with the exact quotes highlighted, and ask for a written yes. One email. Every time.
When the numbers aren't impressive
Use real numbers or use none. There is no third option, and the temptation to round a modest result up into a headline percentage is the fastest way to build a page that persuades nobody and exposes you legally. The FTC's Rule on the Use of Consumer Reviews and Testimonials, in effect since October 21, 2024, lets courts impose civil penalties for knowingly deceptive reviews and testimonials, and it carves out no exception for small startups marketing on a budget.
The related trap is the exchange you forgot to mention. If the customer got a discount, a free year, credits, or points in return for taking part, say so in the piece. The FTC's Endorsement Guides treat an undisclosed material connection as deceptive, and one plain sentence costs you nothing while its absence costs you the whole page.
Results you can publish honestly when there is no clean metric:
- A customer-reported number, attributed as theirs. "Ana estimates it saves her about three hours a week" is honest. The same figure on your homepage as a general claim, extrapolated from one customer, is not.
- A behaviour that stopped. They no longer check something at the weekend, no longer keep a parallel spreadsheet, no longer chase people for status updates. Ending a chore is a result.
- A thing they can now do at all. Onboard a second person, ship on a Friday, take a week off. Capability is often more persuasive than efficiency.
- Time to first value. How long it took them to get the first useful output. Concrete, verifiable, and exactly what your reader is trying to estimate for themselves.
One thing you should never do: write the case study yourself and ask the customer to sign off on words they never said. That is not a grey area under the rule above, and it reads as invented anyway.
Five jobs one case study does
The reason one story is enough is that it is not one asset. Published properly it does five distinct jobs, and most founders stop after the first.
A page on your own domain
Not a PDF, not a Notion doc. A real indexable URL. This is the page an answer engine can cite when someone asks whether your kind of tool works for their kind of team.
The proof block on your landing page
One verbatim quote plus the single most concrete detail, linking to the full story. Replaces the generic three-testimonial carousel every reader skims past.
The reply to "do you have anyone like me?"
The most common question in a sales or support thread. One URL to send, rather than a paragraph you retype, closes the loop in a single reply.
Launch and directory copy
The concrete detail from question four is the strongest line you can put in a Product Hunt first comment or a directory description. Specificity is what stops the scroll.
Six weeks of build-in-public posts
One post per interview question: the trigger, the objection, the first week, the disqualifier. You already did the reporting; publishing it in pieces costs nothing.
All five depend on one input you may not have yet: a customer who has used the product enough to have a before and an after. That is the real bottleneck for a founder at three customers, and it is the gap Favors.dev was built to close. It is a founder marketing co-op with a points economy: you earn points by doing verified marketing favors for other founders, and you spend them requesting the same back. The testimonial and feedback actions in the queue put your product in front of founders who will actually use it and then tell you, in detail, what changed. Some of those become quotes. One of them becomes your first case study.
The wider argument for proof as a channel you earn rather than buy is in SaaS marketing on no budget.
Frequently asked questions
What should a startup case study template include?
Four sections, in this order: situation (what the customer was doing before), friction (what triggered the search and what the old way cost them), change (what they actually did in the first week), and result (what is different now, in real numbers or plain qualitative terms). Around that core, add a headline naming the customer and the outcome, one verbatim pull quote, the objection they had before signing up, and a short line on who this is for and who it isn't. That is the whole template. Everything else is design.
Can you write a case study with only one or two customers?
Yes, and one is the right number to start with. A library of ten shallow studies is worse than one story told in detail, because depth is what makes a case study credible and depth is the thing a template library cannot fake. Pick the customer who can name a specific moment when things changed, interview them for twenty-five minutes, and publish that. Write the second one when you have a customer in a visibly different situation from the first.
How long should a startup case study be?
Long enough to carry the mechanism and no longer, which in practice is 600 to 1,200 words. The failure mode is not length, it is abstraction: three hundred words of concrete detail beats fifteen hundred words of adjectives. If a paragraph could appear in a competitor's case study with the names swapped out, cut it.
What if the customer's results aren't impressive?
Publish the unimpressive result. A modest, specific, verifiable outcome is more persuasive to a careful buyer than a spectacular one they suspect you invented, and inventing or exaggerating one is a legal problem as well as a credibility problem under the FTC's rule on consumer reviews and testimonials. Where there is no number, use a qualitative change the customer described in their own words, and say plainly that it is one customer's experience.
Do you need permission to publish a customer case study?
Always, in writing, and it costs you one email. Send the finished draft with the exact quotes you plan to use and ask them to reply with approval or edits. If you gave the customer anything in exchange for participating, such as a discount, a free year, credits or points, disclose that connection in the published piece. The FTC's Endorsement Guides treat an undisclosed material connection as deceptive, and readers tend to spot the omission anyway.