A case study gets read by someone about to sign
A B2B case study gets opened near the end of a purchase, by someone with a shortlist and a deadline. That may be you checking a supplier, or your own buyer checking you.
Nothing on a supplier’s own site counts as independent. Google makes a similar call about reviews. When the business being reviewed “controls the reviews about itself”, its pages with Organization or LocalBusiness markup are ineligible for review stars. A case study sits in the same position, so we judge it by what it says.
We look for 5 things. A source you could reach, a before number with its method, a customer close to your own, a date, and a problem with its fix. Design portfolios use “case study” for a design write-up, and none of this applies to them.
Federal buyers already wrote the checklist
US federal buyers grade a supplier’s track record, called past performance, from a written list.
The FAR Part 15 text asks evaluators to weigh the “currency and relevance of the information, source of the information, context of the data, and general trends in contractor’s performance”. Section 15.104 brings in past problems. A request for proposals must give bidders the chance to “provide information on problems encountered on the identified contracts and the offeror corrective actions”.
Your buyer is unlikely to be a federal agency. The questions travel anyway, and we built the 5 signs below from them, leaving trends aside.
A case study from us is written for this reader, with every figure approved by the customer in writing before publication. Our case study page lists what we need from you.
Sign 1: a source you could reach
Logos with no role behind them prove nothing. Look for a person by role, a company by type and size, and a sign that someone could pick up the phone.
Under UK CAP Code rule 3.47, the marketer must “hold documentary evidence that a testimonial or endorsement used in a marketing communication is genuine”. Marketers must also “hold contact details for the person who, or organisation that, gives it”.
A missing name isn’t a fail by itself. We’d accept it if the page states the reason, and 2 US rules give suppliers a good one. On US defence contracts with the standard disclosure clause, releasing non-public contract details needs the Contracting Officer’s written approval, asked for at least 10 business days ahead. Federal employees can’t lend their title to endorse a supplier, outside narrow exceptions.
Rule 3.45 adds one more check: an ad must make clear “where consumer reviews have been incentivised”. A customer who got a discount for taking part has a stake, and you should be told.
Sign 2: a before number, taken before the change
Every result needs 4 parts: the number before, the number after, the period, and who counted. Without the first, a percentage has nothing to be a percentage of.
We read the before number first. A figure pulled from the customer’s own log before the change is evidence. A figure recalled 6 months later is an estimate with a decimal point.
Google’s review guide asks writers to “Share quantitative measurements about how something measures up in various categories of performance”. The EU directive on advertising to businesses lets an authority “require the advertiser to furnish evidence as to the accuracy of factual claims”. If that evidence falls short, the claim can be treated as inaccurate.
Look for a caveat next to the number, since it shows somebody counted and noticed what else changed in the same period.
Sign 3: a customer close to your own
Check whether the customer looks like you. Compare size, sector and the conditions behind the result, since a result from a much bigger plant says little about yours.
Then check it’s the same product. CAP rule 3.48 runs to 7 words: “Testimonials must relate to the advertised product”. A case study about an older version, or a bundle you aren’t buying, fails that test even if every number is true.
FAR calls this relevance and “context of the data”. A supplier’s best customer, shown without that context, gets read as its typical one.
Sign 4: a date, and a customer who stayed
Look for when the project ran, when the result was measured, and any sign the customer still uses the product.
Currency (how recent it is) leads the FAR list. An undated case study can be 5 years old, about a customer who left 2 years ago, and we wouldn’t forward one before asking the supplier.
Sign 5: a problem on the record, with the fix
Find the part that was harder than planned and what the supplier did about it. A project where nothing went wrong looks like no project you’ve ever run.
Section 15.104 of the FAR text makes room for this. If you’re on the writing side, what to collect before a case study covers capturing it while the project is live.
One invented result, written twice
Invented for this page: no real plant, supplier or sensor is behind either version.
The showcase version:
A leading food producer cut downtime by 40% with our condition-monitoring platform. “The results speak for themselves,” says the operations director.
The checkable version:
A dairy packer runs 3 filling lines on 2 shifts. Line 2 logged 26 unplanned stops a month from January to March 2026, counted in the plant’s own stop log. Vibration sensors went on line 2 only, in April. July to September, the same log showed 15 a month. The first 6 weeks brought 9 false alarms a week, until the maintenance team reset the thresholds. The plant paid list price, still runs the system, and its maintenance lead takes reference calls.
26 down to 15 is 11 fewer stops a month, a 42% drop by our arithmetic. The 40% in the first version could be the same result, rounded, and you can’t tell from the page.
Run the 5 signs on both versions. The second has a source you can reach, a before number from a log, a plant you can compare with yours, 2 dated quarters and a problem with its fix. The first has an adjective in each place.
Where the case study in front of you goes next
We’d put it in 1 of 3 piles, and each pile has its own next step.
Forward it when the source and the before number both hold, since your committee can check both without you.
Some case studies hold up except for the source. For those, ask the supplier for a 20-minute call with the customer, which any supplier with a real customer can arrange.
Set it aside when every sign is an adjective, because your engineer will read it as marketing. For writing samples, see B2B copywriting examples, and for other formats, B2B content marketing examples.
Our own composite, held to the same 5
Our logistics case study is a composite with illustrative figures, and its first line says so.
By our reading it passes sign 2. The baseline was timed before anything changed, and the churn figure carries a caveat about a price change in the same quarter. It passes sign 5, with a section called “What didn’t work” about a setup wizard that was built and then dropped.
Where it’s weaker: signs 1 and 4 fail by design. There’s no customer to call and no real project date, because there’s no real project. A case study you publish about a real customer shouldn’t fail those 2.
Sources
- ASA, CAP Code section 3, rules 3.45, 3.47 and 3.48. Read 18 October 2026. UK only.
- GSA, FAR Part 15 overhaul text, sections 15.104 and 15.202, last updated 28 September 2026, and DFARS 252.204-7000, last updated 23 September 2026. US federal acquisitions only.
- eCFR, 5 CFR 2635.702, as of 1 October 2026. US federal employees only.
- EUR-Lex, Directive 2006/114/EC, Article 7, read 18 October 2026. Applies through national law.
- Google Search Central, Write high quality reviews, updated 10 December 2025, and Review snippet, updated 8 September 2026. Written for reviews, used here as a parallel.
For a case study of your own, we start from the customer, the change and the figures they’ve cleared for publication. With those in hand, a scope and a price reach you within 1 working day.