Backing Up the Numbers You Publish: Customers and Metrics
Evidence · 10 min read ·
How to define, measure and evidence figures like customer counts and active users so that every public number can survive a reader's question.
Numbers carry authority. "Four thousand customers" sounds like a fact, and a reader may place a good deal of weight on it. Yet few numbers on launch pages are as simple as they look. Four thousand what? Counted when? Including whom? A figure with no definition is closer to an impression than to a measurement.
This guide is about publishing numbers you can stand behind. It explains how to define a metric, how to count it, how to keep evidence and how to present it so that a careful reader's questions are already answered. It applies to customers, users, downloads, uptime, ratings and any other figure on a page.
Why numbers deserve care
A published number does three jobs. It signals scale, it signals momentum and it signals that the team is serious. If a number is inflated, it does the same jobs falsely. Readers who discover that will not only discount the number, they will discount everything else you have said.
There is also a practical risk. In many places, claims made to consumers must not be misleading, and numbers are a classic place where misleading claims arise: counting every sign-up as a customer, quoting the best month as typical, rounding up generously. An honest, documented number is the safest answer.
Start with a definition
A key performance indicator is a measurable value used to assess progress toward an objective. The word that matters is measurable: it must be defined well enough that two people counting would get the same answer.
For each number you publish, write a definition that states:
- What is counted. A paying account, a registered user, a download, a completed order.
- The unit. Accounts, people, organisations, installations.
- The period. All time, the last thirty days, a calendar month.
- Inclusions and exclusions. Do free accounts count? Test accounts? Staff accounts? Duplicates? Cancelled customers?
- The source. The system, report or query the number comes from.
- The date of measurement. When the figure was last taken.
Write it in a single sentence that a reader could understand: "Paying accounts with at least one active subscription on 1 March, excluding trials and internal test accounts."
The difficult word: customer
"Customer" is a slippery term. A customer base is the group of people or organisations that buy a firm's products. But what counts as buying?
Be clear about each possibility.
- Signed up: created an account. This is interest, not customership.
- Activated: did the main thing at least once.
- Active: used the product in the last period.
- Paying: currently paying.
- Ever paid: paid at least once, including those who left.
The numbers differ widely, and each is honest if labelled. The mistake is to publish the largest, "ever signed up", under the label that readers read as the smallest, "customers". If you want to use a large figure, use the right word for it: "accounts created" or "people who have tried it".
Count carefully
Counting should be repeatable.
- Use a single source of truth, such as your billing system or your product database, rather than copying figures between spreadsheets.
- Write the query or steps down, so that anyone can repeat them.
- De-duplicate. Remove duplicate accounts, test accounts and internal users.
- Check against a second source where you can: billing against product usage, for example. If they disagree by a lot, find out why before publishing.
- Freeze the figure with a date. Save the output or an export, as evidence, with the time it was taken.
- Have a second person review the count before it goes public.
Data integrity, the accuracy and consistency of data over its lifecycle, is protected by exactly these habits.
Round honestly
Rounding is fine, as long as it does not mislead.
- Round down, not up. "Over 3,900" is honest for 3,940. "Nearly 4,000" invites a hard look.
- Use thresholds you have crossed. "More than 1,000" is true if you have 1,043.
- Be consistent. If you round the same way each time, trends are readable.
- Do not combine different measures to reach a bigger figure, such as adding users and downloads.
- Avoid false precision. "4,017 happy customers" implies a satisfaction measure you have not made.
A modest, accurate number tends to be more convincing than an inflated one, because careful readers notice the difference.
Date everything that moves
Numbers age. Put the date next to any figure that changes.
- "Used by 1,200 teams, as of 1 March."
- "Average rating 4.6 across 310 reviews, March."
Then keep the promise by updating or removing it on a schedule. A stale number is a small falsehood that accrues slowly. If keeping a date current is too much work, use a phrase that remains true, such as "more than a thousand", and keep the evidence for the threshold.
Other kinds of metrics
The same principles apply across the usual figures.
Uptime. State the period, the method of measuring and what counts as downtime. "99.9 per cent over the last ninety days, measured by external checks every minute, excluding scheduled maintenance."
Speed. Give the test conditions: device, network, location and date. A result from a fast office connection says little about real users.
Ratings. State the source, the number of ratings and the date. Do not select which ratings to show.
Growth. Give the base and the period. "Doubled in three months" from ten to twenty is true and not the same as from ten thousand to twenty thousand.
Financial figures. State whether they are recognised revenue, bookings or estimates, and the period. Where possible, tie them to accounts that could be audited.
Savings and outcomes. Describe how they were measured, over whom and for how long. A claim that users "save five hours a week" should rest on a survey or measurement you can describe.
Keep an evidence file
For every published number, keep a small file.
- The definition.
- The export or screenshot, dated.
- The method used.
- The name of the person who produced and reviewed it.
- The pages where the number appears.
- The date for the next review.
An audit, in the general sense, is a systematic examination of records to check accuracy. You are preparing for the possibility, however remote, of being audited by a reader, an editor or a verification scheme. The file also saves time when you update the figure. The guide on keeping an audit trail for claims shows how to organise it.
Publishing the method
Readers are reassured when a number is accompanied by a short note on how it was counted. Place it near the figure, or behind a "how we count this" link.
A good note answers three questions: what is included, what is excluded and when it was taken. Keep it short. The aim is not to hand over your internal data, only to let a careful reader see that the number is not a conjuring trick.
When a number goes down
Eventually a number will fall: customers leave, a month is weaker, a metric is redefined. This is when honesty is tested.
- Update the page, rather than leaving the higher figure up.
- If a figure has become embarrassing, consider whether it is the right one to publish at all. Choose a measure that fairly represents the product, and say that you have changed it.
- If you change a definition, note it, and do not compare the new number with the old as if they were the same thing.
- If you find an earlier error, correct it openly, as described in the guide on handling lapses and errors.
A team that corrects its numbers in public is easier to trust than one whose numbers only ever rise.
Customer names and logos
Naming customers is a special case. A logo on a page implies that the company uses and approves the product.
- Get written permission for each logo or name.
- Use only current customers, or label past ones as such.
- Do not imply endorsement beyond what they agreed.
- Remove names promptly if asked.
- Be careful with partial relationships, such as one team in a large company using a free tier, which does not mean the company is a customer.
A worked example
A small company's home page says "Trusted by 5,000 teams". The founder is asked by an editor how that is counted. She checks and finds the figure is the number of accounts ever created, including trials that were never used, internal test accounts and duplicate sign-ups.
She and her colleague write definitions for four measures and run the counts from the billing system and the product database. Accounts ever created: 5,012. Accounts that completed setup: 2,870. Accounts active in the last thirty days: 1,430. Paying accounts: 612. Billing and the product database agree to within eight accounts for the paying figure, and they trace the difference to two accounts paid by invoice.
They change the page to read "Over 600 paying teams" with a small date and a link to a note: "Teams with an active paid subscription on 1 March, excluding trials and test accounts." They file the exports with the definitions and set a monthly review. The number is much smaller than the original. It is also one the founder can defend in a sentence, and the editor, who had asked the question, includes the product in her next selection, citing the clear definition.
Questions teams ask
Is it dishonest to publish a big number with the right label? Not if the label is accurate, such as "accounts created". Be sure readers will understand it.
Do I have to publish revenue? No. Publish only what you choose to, but anything you do publish should be true.
Can I use estimates? Yes, if labelled as estimates, with the basis stated.
What if I cannot measure something? Say so, or leave it out. A missing figure is better than an invented one.
Summary
A published number is a promise that it can be shown. Define what is counted, over what period and with what exclusions, count from a single source, check against a second and have someone review. Use the right word for each measure, round down rather than up and date any figure that moves. Keep an evidence file, publish a short method note and update or remove numbers when they change. A smaller, honest figure that you can defend beats a bigger one you cannot.
Questions and answers
- Why do published numbers need evidence?
- Because readers rely on them to judge a product, and an unsupported figure can mislead and damage trust if challenged.
- What is a metric definition?
- A precise statement of what is counted, over what period and with what exclusions, written down so the number means the same each time.
- Is it acceptable to round numbers?
- Yes, if you round honestly and consistently. Rounding down is safer than rounding up.
- Should I publish the methodology?
- A short note on how a figure is counted answers most questions and makes the number more credible.
- What if a number falls?
- Update it, or remove it and explain. Leaving a higher, stale figure up is the riskier choice.