Why this signal matters for Google
A case study shows your work, not your pitch. A services page says what you can do. A case study says what you did, for whom, when, and what came of it. Four things make the difference: a named client, a dated starting situation, the method you followed, and numbers measured before and after. A locksmith writes "fast response"; their case study describes the shop broken into on a Sunday morning, the door replaced in four hours, and the $1,600 invoice the owner agrees to have quoted. That is also what separates it from a lived testimonial: there, you tell your own story; here, you tell a client's, with their permission and their numbers.
Google has its raters look for exactly that kind of trace. Human evaluators rate page quality using a public manual, the Quality Rater Guidelines (QRG for short). Section 3.4 tells them to check whether the author has genuine experience with the topic, rather than knowledge copied from elsewhere. Section 7.3 recognises everyday expertise: the knowledge of someone who has actually done the thing, degree or no degree. A case study is the densest form of that proof, because it carries both halves at once: the account of the work and the measured result. Section 5.1, on the other side, files a page as low quality when its author lacks experience of the topic. A "Our work" page listing ten logos and not one line of narrative lands squarely in that box.
A number without its frame proves nothing. "+300% traffic" can mean three visitors became twelve. Three things turn a number into evidence: the starting value, the timeframe, and the measurement tool. "From 340 to 1,020 monthly visits between January and June 2026, measured in Google Analytics" can be checked; "we tripled their traffic" cannot. Then add what did not work. Most people learning how to write a case study skip this part, and it is the one that buys trust: a study that admits a missed deadline or a wrong assumption reads as more honest than one where everything went to plan. If your site touches health, money or safety (Google calls those topics YMYL, for Your Money or Your Life), section 3.4.1 expects a higher bar of trustworthiness, and an uncheckable number costs you more there than anywhere else.
AI tools cite what they can cross-check. When ChatGPT, Perplexity or Google's AI summaries recommend a provider, they look for details that show up in more than one place: a company name, an industry, a figure, a year. An anonymous case study gives them nothing to match. The same study with the client's name, sector and year becomes quotable, especially when the client says the same thing on their own site or in a verified review. In practice, when you run a site through all 72 signals, the case study is often the only page that carries lived experience and hard data together. That is what makes it so easy to cite.
The NeuroEEAT scoring grid (0-10)
Score what a sceptical prospect can verify on their own in your case study, without calling you first.
| Score | Description | Concrete indicators |
|---|---|---|
| 0-2 | Nothing to verify | An "our work" page with logos or broad statements. No client named, no dates, no numbers. No way to tell whether the project happened. |
| 3-4 | A story with no evidence | The project is described, but the client stays "a mid-sized manufacturer". Results come as percentages, with no starting value and no timeframe. |
| 5-6 | Decent | Client named, or anonymised with sector and size given. Starting problem described. At least two before and after numbers with their period. No measurement tool cited. |
| 7-8 | Solid | Context, step-by-step method, a dated before and after table, a named measurement tool, a client quote with full name and job title, and at least one obstacle described. |
| 9-10 | Verifiable end to end | All of the above, plus photos or screenshots taken during the project, a link to the client's site, written publication consent on file, and a follow-up a year later with updated numbers. |
How to audit it on your site
Five checks on the case studies you already publish. Start with the one you send most often to prospects.
- 01
Look for the client name
Is the client named, or described as "a leading player in its market"? If nobody can look them up, your study is worth no more than a services page. Anonymising stays fine when the contract demands it, provided you give the sector, the size and the region.
- 02
Look for the dates
Find three: project start, project end, and the moment of measurement. A case study with no timeframe never says whether the result took three weeks or three years. That is the first thing a prospect wants to know.
- 03
Check every number
For each result you claim, confirm there is a starting value, an ending value and a named tool. A bare percentage should be completed or removed. If you can no longer find the original measurement in your files, take the number out.
- 04
Look for the method
Count the steps you actually describe. "We rebuilt their strategy" is not a method. A reader should be able to follow the order of operations. Fewer than three concrete steps and you have written a brochure.
- 05
Check consent and evidence
Do you have written permission to publish the client's name and numbers? Did you keep dated screenshots or photos from the project? Without consent, one complaint is enough to force the page offline.
How to activate or improve it
Three levels, from a first case study written in an afternoon to a library your competitors cannot copy.
Write one complete case study
- Pick the project you are proudest of and still have the numbers for.
- Get the client's consent by email, spelling out what you will publish: their name, their numbers, their quote.
- Follow four parts: starting situation, problem, what you did, measured result.
- Give at least two numbers with their starting value and the period they cover.
- Close with a client quote carrying their full name and job title.
Make the study checkable
- Add a before and after table with the date of each measurement and the tool used.
- Include screenshots or photos taken during the project, never stock images.
- Describe one obstacle and how you handled it: that is what makes the rest believable.
- Link to the client's website, and ask them to link back.
- Sign the study with your own name, not the company's: readers want to know who did the work.
Build a library of case studies
- Write one study per client type and per problem type, to cover your commercial queries.
- Revisit each study a year later: did the result hold? Publish the dated update.
- Document your measurement method in public: which tool, which baseline period, what you deliberately exclude.
- Mark the study up in JSON-LD (Article with an identified author), so engines connect the page to the person behind it.
- Pair each study with a public review from the same client: when both tell the same story, it becomes proof.
5 common mistakes to avoid
Sector specifics
Patient cases come with rules
A topic where a bad page can hurt the reader: section 3.4.1 expects real trustworthiness. A clinical case goes online only with written consent and with every identifying detail removed. Describe the protocol, the timeline and the follow-up; leave outcome promises out, since most health professions forbid them outright.
Performance figures are a minefield
A study built around past returns can put you on the wrong side of your regulator. Write about how you work instead: the opening assessment, how often you review, how you explain a trade-off. Keep performance numbers in the official documents built for them.
The job site is your case study
A roofer needs no table: the roof before, the roof after, the date, the neighbourhood and the price paid will do. Name the town and the type of building, because those are the words your customers type. One study per area beats a single "our work" page that blends them all together.
The highest-return format you have
In B2B, the case study is usually the most-read page before a first meeting. Write one per industry: a buyer wants to recognise themselves. Add the number of people involved and the rollout time, the two figures they look for before price.
Signals to activate in parallel
A case study never stands on its own. These signals decide whether yours holds up:
FAQ
How do you write a case study when the client refuses to be named?
Anonymise the name and nothing else. Give the sector, the company size, the region and the year: "a 12-person accounting firm in the Midwest, in 2025". Keep every number and the whole method. What makes a study uncheckable is not the missing name, it is the missing detail.
How many case studies should you publish?
One per client type and per problem you solve. Three deeply detailed studies beat fifteen ten-line summaries. A good test: a prospect should reach the study that looks like their own situation in under two clicks from your services page.
How long should a case study be?
Between 800 and 1,500 words in most trades. Under 500 words, either the method or the numbers are missing. Over 2,000, add a summary at the top with the three headline results, because half your readers will read only that.
Should a case study be gated behind a form?
Publish it openly. A gated page is not indexed, so it earns no visits and no citations in AI tools, and it counts for nothing with an evaluator who will never fill in your form. If you want leads, offer a longer PDF version alongside the public page.
What if the project results were only average?
Publish it anyway and say what got in the way. An honest study about a hard project helps a prospect more than a flawless win, because it shows how you behave when things stall. The one thing never to do is round a number in your own favour.
Anonymized case study
Before · an "our work" page
After · four complete, dated studies
Illustrative figures meant to show the typical progression of the signal. They do not come from a real client case and must be replaced with your own measurements before any external communication.