The Screenshot Test: Auditing Every Number in Your Demo Before a Stranger Does

A founder I know once showed a journalist a dashboard with a beautiful upward curve. The curve had no axis labels, no date range and no note explaining that it was modelled from a pilot of eleven users. The journalist didn't ask. They screenshotted it, dropped it into a piece, and within an hour someone in the comments had asked what the sample size was. That question is much harder to answer after publication than before.

Demos are not neutral. The moment a chart leaves your laptop — in a pitch deck, a launch page, a sales call recording, a journalist's browser or an early adopter's phone — it starts behaving like evidence. Under UK advertising rules, that matters. The ASA's CAP Code expects marketing claims to be substantiated and not misleading. Consumer protection law takes a dim view of misleading commercial practices. Even in a B2B pitch, a screenshot can travel further than the slide it came from.

This is not about having perfect data. It is about being honest about the data you have.

Where demo data hides in plain sight

Start by listing every place a number appears. Not just the big hero chart. The customer count in the top-right corner. The logo wall. The 'trusted by' strip. The before-and-after panel. The map with pins. The waitlist counter. The 'live' activity feed. The testimonial with a job title. The chart in the pitch deck that says 'market growth' but actually shows your own projected growth. The dashboard tile that says 'revenue' but is pulling from a sandbox account.

Founders often know which numbers are real. The problem is that nobody else does. A demo environment is designed to look alive, so it is easy to forget that a synthetic customer called 'Acme Test Ltd' looks exactly like a claim.

Three labels, used consistently

A simple labelling system sorts most of this out. Every data point in a public-facing or journalist-facing demo should be one of three things: real, illustrative or projected.

Real means it comes from actual usage, with a defined source and a date range. If it is real customer data, you need permission to show it and a plan for anonymisation. Do not assume a logo wall is fine because the logo is public — check the contract and the relationship.

Illustrative means it is made up to show how the product works. Label it visibly: 'Illustrative data — not customer data.' Put that label in the chart frame, not in a hover tooltip that disappears the moment someone screenshots the page.

Projected means it is a forecast. Say so, and say what it is based on: 'Projected from 40 pilot users, Sept–Oct 2026.' A projection without a basis is just a vibes chart with a y-axis.

The screenshot frame is the real canvas

Designers tend to label charts for the full-screen experience. Reviewers tend to crop. If your label lives in a footer, a tooltip or a settings panel, it will not survive the screenshot that ends up in a Slack channel or a news story.

Put the label where the data is. In the chart title. In the legend. In a small caption directly beneath the number. Use plain English, not a footnote. 'Demo data' is better than 'sandbox environment'. 'Sample: 11 pilot users' is better than 'n=11'.

Then do the boring test: screenshot every screen yourself, crop it the way a stranger would, and see whether the claim still makes sense. If the screenshot is misleading without the surrounding context, the screen is the problem — not the person who cropped it.

Customer counts need definitions, not bravado

'10,000 users' is one of the most abused phrases in UK startup marketing. Registered users? Monthly active? Waitlist sign-ups? Newsletter subscribers? People who once clicked a LinkedIn ad? Each is a different claim, and each needs a different label.

Pick a definition and stick to it. Add an as-of date. Say whether it is UK-only or global. If it is cumulative, say so. If it includes free accounts, say so. If the number is a waitlist, call it a waitlist — not a customer base. Early adopters are unusually good at spotting the gap between 'signed up' and 'uses it every week'. Journalists are better.

Before-and-after panels are where credibility leaks

A before-and-after chart is a story with a moral: our product made this better. That moral only holds if the baseline, timeframe and metric are the same on both sides. A common own goal is comparing a quiet January against a busy March and calling it growth. Another is comparing a pilot cohort's best week against the full customer base's average week.

Write the comparison in one sentence before you design the slide: 'Same metric, same customer segment, 12-week period, measured by [method].' If you cannot write that sentence, you have a design problem, not a data problem.

Keep the demo tenant clean

If your product has a demo environment, treat it like a shop window. Use synthetic names that are obviously synthetic. Watermark exported screenshots with 'DEMO'. Do not mix real customer records into a sales demo unless you have a clear legal basis and a very good reason. A stray real email address in a screenshot is a data protection incident with a nice gradient.

Run a 30-minute pre-launch audit

Book half an hour. Open every deck, launch page and demo flow. For each number, write down: what it claims, where it comes from, and whether the label would survive a crop. Then choose one of three actions: label it, source it, or remove it.

Keep a one-page data note for the claims you expect to be asked about. Not a legal dossier — just the definition, the date range and the source. When a journalist asks 'where does this number come from?', you want to send a link, not a paragraph of hedging.

The goal is not to make your demo look less impressive. It is to make it survive being looked at closely. In 2026, the screenshot is the first review. Make sure it is a fair one.