How to Get Real Numbers Out of a Customer for a Case Study

Ask for before, after, and a time window. Negotiate form when legal blocks the exact figure. Never invent a metric.

How to Get Real Numbers Out of a Customer for a Case Study
Image URL
Do not index
Do not index
AI summary
Ask customers early for a measurable before-and-after result, including the time period, unit, and data source. If exact figures cannot be shared, agree on an accurate range, relative change, or relevant operational proxy; never estimate or invent metrics. When no number is approved, keep the story specific with concrete examples, context, and limitations.
Title
How to Get Real Numbers Out of a Customer for a Case Study
Date
Description
How to ask for case study metrics, negotiate a publishable figure when exact numbers are blocked, and keep the story honest with zero invented stats.
Status
Current Column
Person
Writer
You have a customer who said yes. You have a draft that says "significant improvement." Sales wants a number. Legal will not invent one for you.
Here is how to ask for a metric, what to do when they cannot share it, which proxies are legitimate, and how to keep the story honest when there is no number at all.
If you want the full case-study system — pick, invite, collect, write, approve — start with our guide to customer case studies. The result-stage questions that tee this up live in case study interview questions. This page is the extraction playbook around those questions.

What you are actually trying to get

A usable case study metric is not "ROI improved."
It is one line a stranger can check:
Before → after → over what window → measured how.
"Average ticket resolution fell from 18 hours to 4 hours, April to June, from our Zendesk reports."
That line has four parts. Miss one and a sceptical buyer discounts the whole claim.
Pick one primary metric family before you send the invite. Not five. One.
Family
Examples of what to ask for
Time
Hours per task, cycle time, time-to-first-value
Cost
Tool spend replaced, overtime cut, outsourcing avoided
Volume
Deals closed, tickets handled, pages shipped
Quality
Error rate, rework, on-time rate
Revenue / pipeline
Pipeline influenced, win rate, average order — only if they own the number
I would reach for time or quality first. Revenue is the number everyone wants and the one legal blocks most often. Operational metrics usually clear faster and still convince the people who will use the product.

Where the number lives before you ask

Do not invent a dashboard the customer does not have.
Before the invite, ask yourself — or their CSM — one practical question: where would this number come from if they said yes?
  • Their analytics or CRM export
  • A helpdesk / billing / payroll report
  • Your own product analytics, if they already see that number inside your app
  • A spreadsheet they already send to their boss
If nobody can name the source, you are about to ask them to guess. Guessing is how soft adjectives get into drafts. Either pick a metric with a real source, or plan from the start to publish without a number.

The ask sequence

Do this in order. Skipping ahead is how you end up estimating for them.

1. Ask early, in writing

Do not wait until the polished draft. Put the metric question in the first question set.
A number negotiated after design and legal review is a number people dig in on. A number agreed while answers are still draft is a number people can still shape.
If you are collecting async — email, form, or a link — the customer can pull the figure from a dashboard instead of guessing on a call. That is the point.

2. Ask for three parts in one question

Copy-pasteable:
What changed after you started using [Product]? If you can share a number, please give me three parts: what it was before, what it is now, and over what period. Even a rough range helps.
That is better than "What results have you seen?" which returns adjectives.

3. Refuse the adjective

When they reply "much faster" or "a lot more pipeline," stay with it.
Rescue follow-ups that work:
  • "Can you put a number on that — even a range?"
  • "What was a normal week before, and what is a normal week now?"
  • "If you had to pick one metric that moved the most, which one?"
Do not accept a comparative with no base. "60% faster" without "faster than what" is marketing, and a finance reviewer will treat it that way.

4. Confirm the unit out loud

Read it back in their words before you write it into a draft:
"So I'm hearing: support tickets from about 40 a day to about 25 a day, over the first quarter on [Product], measured in your helpdesk. Is that right to publish?"
Catch percent vs percentage points. Hours per person vs hours across the team. Revenue vs profit. Wrong unit in a published case study is worse than no unit.

5. Negotiate the form, not the truth

When exact figures are blocked, you are not stuck. You are choosing a publishable form of the same fact.
Offer these in order, and get them to pick one:
  1. Relative change — "cut wait time by about a third"
  1. Range or rounded figure — "between 40 and 50 hours a week"
  1. Operational proxy — task time, error rate, steps removed (usually less sensitive than dollars)
  1. Named period without the dollar — "in the first 90 days after launch"
Write down which form they approved. That note is what you send legal later.
What you do not do: invent a midpoint, round in a direction that helps you, or borrow a number from another customer.

Email you can send when the first answer is soft

Use this when the async reply came back as adjectives:
Subject: Quick follow-up on the [metric] question

Hi [Name] — thanks again for the answers.

On results, you mentioned [their soft phrase]. For the case study, it helps readers if we can show before → after over a clear window.

Would any of these work for you to publish?

1) Exact figure from [system], if you're comfortable
2) A range (e.g. "roughly X–Y per week")
3) A relative change ("about a third less")
4) A different operational metric from the same workflow

If none of those clear, we can publish without a number and keep the before/after story specific. Totally fine — I'd rather be accurate than padded.

Which option do you prefer?
That email does three jobs: it offers forms, it gives them an exit, and it records that you did not invent anything when they say no.

When they cannot share a number

This is normal. Plan for it instead of treating it as a failed interview.
Try these fallbacks in order:
  1. Same metric, softer form (range / relative / proxy) — see above.
  1. A different metric in the same family — if revenue is blocked, ask for cycle time on the same workflow.
  1. A single concrete example — one deal, one week, one ticket type — labelled as an example, not a typical result.
  1. A private version for sales under NDA, and a public version without the figure — label both, record who may receive which.
  1. Publish with no number. Keep the story specific.
Never fill the gap yourself. An estimate the customer did not make is a fabrication. So is another customer's figure wearing this customer's name.
If the only thing they will approve is "we love it," you do not have a case study metric problem — you may have a case study vs testimonial decision. A strong quote with a named role can still earn its place; it is a different artefact.

Proxies that are legitimate (and ones that are not)

A proxy is legitimate when it is measured, theirs, and about the same job the case study claims.
Legitimate:
  • Time on a named task before vs after
  • Error or rework rate from their own reports
  • Steps removed from a workflow they can describe
  • On-time or SLA rate from a system of record
  • Volume they already report internally (tickets, deploys, demos)
Not legitimate:
  • An ROI you calculated from an assumed hourly rate they never approved
  • "Industry average" savings pasted onto their story
  • A best-week result written as if it were the usual week
  • Rounding "a little over 30%" up to "nearly 40%"
If you convert hours into money, say so in the draft, name the hourly cost they approved, and label the result as an estimate. If they will not approve the conversion, keep the hours.

How to publish with zero numbers

A case study without a metric is still a case study if the before and after are concrete.
What carries the proof instead:
  • Named role and company context — so a reader can self-identify
  • A specific broken Tuesday — what they were doing before, in their words
  • A specific working Tuesday — what changed in the workflow, not "the platform"
  • One limitation — who should skip this, or what did not improve
That last line matters more when you have no number. Honesty is the substitute for a chart.
For the quality bar on the quote itself, what makes a testimonial good still applies — specificity over adjectives, even when the artefact is longer than a testimonial.
Do not pad emptiness with a fake percentage. Prefer a thin, true story over a thick, invented one.

Doing this with Testimonial

The manual path above works with a Google Doc and a calendar. Where it breaks is the chase: scheduling the interview, then chasing the follow-up when the metric answer comes back soft.
Async case-study interviews are how we run that ask inside Testimonial.
You build a question set of 5–10 questions, save it as a template, and send each customer a personal link. They type or talk. Voice answers are transcribed; the original audio stays for playback. Answers land pre-attributed in your inbox. You can resend, revoke, or remind from the same place.
Two details that matter for metrics specifically:
  • Ask the three-part metric question in the template. They can look at a dashboard before they answer. You are not relying on memory mid-call.
  • Promoters can be auto-invited. On Ultimate +, NPS responses of 9 or 10 can trigger a case-study invite automatically — which is when people are most willing to dig for a real number.
Case Study collection sits on Ultimate + — $95/month per space, billed annually ($1,140/year). Check plans and pricing for the current table.
The limitation, stated plainly: Testimonial collects the answers. It does not invent a metric, clear your customer's legal team, or turn "significant improvement" into a percentage. If they will not approve a figure, you still have to write the honest version — or ship a quote instead. For the no-call mechanics end to end, see running a case study interview without a call.

Common failure modes

Three patterns that produce fake-looking metrics even when everyone meant well.
1. You asked too late. The draft is designed, the quote is locked, and then someone notices there is no number. At that point the customer is done being helpful. Ask in the first question set.
2. You asked for five metrics. They pick none. One primary family, one follow-up if thin, then stop.
3. You "helped" the arithmetic. They said "maybe around thirty percent." You wrote "nearly 40%." That is not editing. That is inventing. Write the weaker true number, or drop the number.
If sales pushes for a bigger figure than the customer approved, the answer is no. A case study that sales will not send is still better than a case study legal has to retract.

A short checklist before you hit publish

One primary metric, not five
Before, after, and time window present — or an explicit decision to publish without a number
Unit confirmed and read back
Published form matches what they approved (exact / range / relative / proxy)
No estimate you made on their behalf
Consent on file for the words and the figure
Limitation line included when the number is soft or absent
If you are still building the wider programme — ask, collect, display — the whole testimonial pipeline is the hub those pieces hang from.

Wrapping up…

You now have a sequence for case study metrics: ask early for before / after / window, rescue soft answers, negotiate the form when exact figures are blocked, and publish a specific Tuesday when there is no number at all.
Stuck on which single metric to ask for first? Reply and tell me what your product actually changes on a normal week — I will help you pick the question.

Written by

Damon Chen
Damon Chen

Founder of Testimonial