Gather results before drafting, in writing, with three things for every number: the figure, the measurement period, and the source it came from (a report, an export, an email, or a named person's estimate). Treat anything without a source as an attributed estimate. Measure operational outcomes (time, volume, steps, cost) rather than claims that need clinical or comparative evidence, and ask the customer for the before as well as the after. A single well-sourced result beats five vague ones.
Key takeaways
- Every number needs three things: the figure, the measurement period, and the source.
- Ask for the before, not just the after. A result without a baseline isn't a result.
- An estimate is a legitimate result if it's attributed and stays an estimate. It becomes a problem only when it's rounded into a statistic.
- Operational measures (time, volume, steps, staffing) are easier to source and easier to approve than outcome or comparative claims.
- Ask at project end, in writing, with a short list. Memory fades and goodwill decays.
When should you ask for results?
At the point where the customer can see the change and still remembers the before: typically six to twelve weeks after go-live for an operational change, or at the natural end of an engagement for an agency project. Earlier than that, the result isn't stable. Much later, the baseline is forgotten and the person who knew it may have moved on.
Ask in writing, with a short, specific list. A vague "do you have any results we could share?" gets a vague answer. A list of three things you'd like to cite, each with a box for the number, the period, and where it came from, gets usable data.
What should you ask for?
For every result you'd like to cite, three things.
- The figure, as the customer would state it. Not rounded, not improved.
- The measurement period. "Tray turnaround averaged 3.8 hours" means little. "Averaged 3.8 hours, measured weekly from January to March" is a result.
- The source. A dashboard export, a report, a spreadsheet, an email, or, if that's all there is, the name and role of the person estimating.
And always the before. "Delayed cases fell to four a month" is only meaningful next to "from an average of eleven over the prior two quarters." Ask for the baseline explicitly, with its own period and source; it is often the harder number to recover.
Which sources count as evidence?
In rough order of strength for review purposes.
| Source | How it's cited | Review risk |
|---|---|---|
| System export or report | "Per the department's weekly turnaround report, Q1." | Low. Reviewers can verify. |
| Customer's internal analysis | "Per the lab's own validation log." | Low to medium. Customer must agree it can be cited. |
| Written statement from a named person | "The SPD manager reports..." | Medium. Attributed, so honest; still a report, not a measurement. |
| Transcript or call notes | "In a call on March 3, the lab director estimated..." | Medium. Attribute with timestamp; get written confirmation of the wording. |
| Recollection, unrecorded | Not citable as a figure. | High. Write the qualitative change and leave the number out. |
Note that a customer's internal data often needs the customer's permission to be cited even in an anonymised story. Ask when you ask for the number.
How do you handle estimates and rough numbers?
Estimates are fine. Converting them into statistics is not. If the lab director says "probably about a third less hands-on time," the publishable sentence is: "The lab director estimates roughly a third less hands-on time per validation; the lab has not yet pulled timesheet data." It is attributed, it is hedged, and it tells the reader exactly how much weight to put on it.
What you may not do is write "33 percent less hands-on time." That sentence claims a measurement that does not exist. In a regulated industry it is the sentence that gets the case study pulled; in any industry it is the sentence a sceptical prospect will test by asking for the data.
Two practical habits. First, when a customer gives an estimate, ask whether the underlying data exists and whether they'd be willing to pull it. Often they will, and the estimate becomes a measurement. Second, if they won't, ask them to confirm the estimate in writing so that at least the attribution is solid.
Which outcomes are safe to measure?
Operational outcomes are easier to source and easier to approve than outcome claims that need clinical, scientific, or comparative evidence. In medical device and biotech marketing in particular, the difference between the two is the difference between an approvable story and a pulled one.
- Time: turnaround, validation time, time to proposal, hours of training to competence.
- Volume and throughput: samples per week, cases per month, sites live.
- Steps and effort: hands-on hours, manual steps removed, repeat cycles.
- Cost and staffing: overtime, shifts, headcount redeployed.
- Adoption: percentage of staff trained, departments using, time to full rollout.
What to avoid unless you have cleared evidence: anything that reads as a safety, efficacy, or clinical outcome claim, and anything comparative about a competitor. The medical device guide and biotech guide cover the specifics.
A results request checklist
Send this to the customer, adapted to the story.
- The one outcome you'd most want a prospect to know about.
- For that outcome: the before, the after, the measurement period for each, and the source.
- Up to two supporting outcomes, each with the same three details.
- Any figure that's an estimate, confirmed in writing as an estimate.
- Whether the source documents can be cited, and whether the story can be named.
- The name and role of the person who can approve quotes and results.
This is close to the proof section of the Case Study Works intake. When a customer completes it, the draft practically writes itself, and every number in it arrives with its source already attached. If the proof isn't there yet, the intake says so before the clock starts, and the approval workflow has nothing to stall on later.
Frequently asked questions
What if the customer only has rough numbers?
Use them as attributed estimates, in the customer's own words, and ask whether the underlying data exists. Never round an estimate into a precise figure. An honest, attributed estimate is a legitimate result.
How many results does a case study need?
One well-sourced headline result with a baseline, plus one or two supporting measures. Five vague numbers are weaker than one you can defend.
Do we need the customer's permission to cite their data?
Usually yes, even in an anonymised story, if the data came from their systems. Ask when you ask for the numbers, and get the answer in writing.
Can a case study have no numbers at all?
Yes. A qualitative story with a clear before and after, attributed observations, and an approved quote is publishable and often more credible than a story with numbers nobody can source.
What should we do if a result changes after publication?
Keep the approval record with the measurement period stated, so the result stands as true for that period. If a later figure contradicts it materially, update or retire the story rather than leave a stale claim live.
Upload the results you have. We'll show you what's sourced and what isn't.
The intake asks for every number with its period and source, and the draft marks anything that still needs one. First draft within two business days of a complete intake.
Written Case Study from $1,495. Complete Case Study Kit $3,495. Pay online, then complete the intake. No calls.