The real danger of grabbing a sample is that its fake numbers come along with its layout. The rule that survives all three: if a figure is quoted without a sample size and a date, it is not a measurement, however confidently it is dressed up.
You searched for a sample because you want something to copy. Fair. But a lot of what you'll find copies the wrong things, so here is a good one first — then the short list of design rules that make it good, the numbers that actually belong on a clinic's version, and the one uncomfortable finding that decides whether any of it is worth the screen space.
A good one, on a single screen
Here is a clinic's weekly dashboard reduced to what earns its place. Five numbers, each with a target or a trend beside it, and two small charts — bars and a line, not a gauge or a pie in sight. You can read the whole thing in about five seconds, which is the entire point.
Sample: a clinic's weekly KPI dashboard
Notice what isn't there. No speedometer for "practice health". No pie chart splitting revenue into eight wedges you can't compare. No number reading $142,387.42 when $142k is all you can act on. Those aren't stylistic quibbles — they're the specific things dashboard experts spent twenty years telling people to stop doing.
What makes it good (and most samples miss)
There's a real body of expert guidance here, and it's remarkably consistent. Three ideas do almost all the work.
1. It fits on one screen
Stephen Few, who effectively wrote the book on this, puts — in his white paper Common Pitfalls in Dashboard Design, the document linked below, which carries a vendor addendum I should have flagged — "exceeding the boundaries of a single screen" at the very top of his list of common dashboard mistakes. The moment a dashboard needs scrolling or a second tab, it stops being a dashboard and becomes a report. The value of the format is that everything important is in view at once; lose that and you've lost the reason to build it.
2. It's built for glancing, not exploring
The Nielsen Norman Group defines a dashboard as "a single-page view that imparts at-a-glance information on which users can act quickly" — information meant to be "consumed fast, with a minimum of interaction or cognitive processing." That one sentence rules out a lot. It's why the sample above uses bars and a line: the eye reads length and position more accurately than any other visual cue, which is why position beats both. Length and angle actually share a tier in Cleveland and McGill; the stronger objection to a pie is that it also asks you to compare areas, and area sits a tier below — NN/G notes that angle "communicates quantitative information poorly," which is why a gauge is harder to read than a bar. A gauge that needs a second look has already failed the "at-a-glance" test. There is one case where a gauge does earn its place, and gauge charts in Excel works through it.
3. Every number has context
Few's second mistake, right after the single-screen rule, is "supplying inadequate context for the data." A number on its own can't be judged. Is a 96.2% collection rate good? You can only tell because the target sits beside it. Is 41 days in A/R fine? Only against the 30–40 range. This is the single most common failure in the sample dashboards people copy: big confident numbers, floating, with nothing to compare them to. A figure with no target, no trend and no benchmark isn't information — it's decoration that happens to be numeric.
What actually belongs on a clinic's version
Design gets you a good-looking dashboard. The right numbers make it a useful one. For a medical or dental practice, five earn a permanent place — and the honest part is that only some of them have a real benchmark to show beside the figure.
The five: net collection rate (of the money you were entitled to, how much you got), days in A/R (how long it takes to arrive), no-show rate (capacity you paid for and didn't use), provider utilisation (how full the schedule was), and patient retention (whether they come back). Revenue belongs on the screen too — but as the scoreboard, not one of the levers. There's a whole piece on why those five, and why not revenue, and a longer 12-KPI guide for the quarterly view.
These two have a professional target you can put a reference line on. The AAFP states the adjusted collection rate "should be 95%, at minimum," with the average at 95–99% and top performers at 99%+, and that days in A/R "should stay below 50 days at minimum; however, 30 to 40 days is preferable."
There's a real figure here, but it's a range, not a single number. MGMA's single-specialty aggregate put the medical no-show rate around 6.8% (2023); peer-reviewed primary-care studies run higher and wider — a systematic review found a mean nearer 15%, with individual studies ranging from about 3% to 48% depending on setting and population. Show your own number against your own trend, and treat any single "industry no-show rate" with suspicion.
These are the two everyone wants a benchmark for, and neither has a credible one. The famous "80–89% optimal utilisation" traces to an article MGMA hosts that was written by a scheduling vendor, footnoted to that vendor’s own whitepaper, dated 2020, and about exam rooms rather than providers. Patient-retention figures — "60–70% is normal", "85%+ for top practices" — come, as far as I can find, from software-vendor blogs with no dataset behind them.
How many numbers? Fewer than you've been told to fear
Someone will tell you a dashboard should have "7±2" items, citing psychology. They're misremembering. George Miller's 1956 paper — the "magical number seven" — was about how many distinct tones or line-lengths a person can tell apart, and how many items fit in immediate memory. It was never a rule about tiles on a screen — Miller was openly sceptical of reading anything grand into the number (he called seven "a pernicious, Pythagorean coincidence"), and later designers have warned against using "7±2" to justify interface limits. Working-memory research puts real capacity closer to three or four chunks anyway.
None of it really applies, because a dashboard is a recognition task, not a recall task — the numbers are on the screen; you're not holding them in your head. So the honest guidance isn't a number, it's a constraint: everything fits on one screen and reads at a glance. In practice that lands somewhere around five to nine headline figures — but treat that as a design convention, not a law of the mind. If a tenth number genuinely changes what you'd do on Monday, keep it. If it doesn't, it's costing you the glance.
The part every sample skips: a dashboard that just exists changes nothing
Here's the finding that should sit under every "sample dashboard" you copy, and never does.
The landmark meta-analysis of feedback — Kluger and DeNisi, Psychological Bulletin, 1996 — pooled 607 effect sizes across 23,663 observations. On average, showing people data about their performance helped (d = 0.41). But in more than a third of cases, it made performance worse. Not neutral — worse. Handing someone a number is an intervention, and interventions can backfire.
That's why the layout is the easy part. A good dashboard is designed to be used, and using it is the bit that isn't on the screen: each red number gets an owner and a date, and the review compares to a target rather than just staring. Build the prettiest dashboard in the world, hang it on the wall and never turn it into a decision, and you're not in neutral territory: across that meta-analysis, showing people their numbers made things worse about a third of the time, in a meta-analysis of feedback studies from 1905 to 1995, mostly individual task feedback rather than dashboards. The fix — an owner and a date for every red number — is my recommendation, not a finding; but the risk it guards against is.
The benchmarks you'll copy by accident
The real danger of grabbing a sample is that you inherit its fake numbers along with its layout. Three you'll meet constantly, none of which means what it claims:
- "80–89% optimal utilisation." Vendor-sponsored, no sample, no method, no stated origin for the range.
- "60–70% patient retention is normal." Software-vendor blogs, all the way down. No dataset anywhere.
- "Missed appointments cost US healthcare $150 billion a year." Traces to a single vendor's estimate, endlessly re-quoted, never tied to a study.
And the cliché that licenses all of it — "what gets measured gets managed" — is worth retiring too. Peter Drucker never said it. The line descends from a 1956 paper by V. F. Ridgway titled "Dysfunctional Consequences of Performance Measurements" — which was a warning that measuring the wrong things, or too many things, distorts behaviour. The phrase people use to justify a crowded dashboard originally argued for a careful one.
How to build the one above
The sample needs no special platform — every number in it comes from exports your practice management system already produces. Three routes, depending on how much you want to build: a KPI dashboard in Excel if you want it done today, Power BI if you want it to refresh itself, or a ready-made template if you'd rather not start from a blank screen. If you're shopping the third route, what Power BI dashboard examples hide is the checklist for telling a real one from a screenshot, and the same check for free Google Sheets dashboard templates is one formula long.
Frequently asked questions
A small set of numbers you can act on, each shown with the context needed to judge it — a target, a trend, or a benchmark. For a clinic that's roughly five: net collection rate, days in A/R, no-show rate, provider or chair utilisation, and patient retention. The rule isn't which metrics so much as how they're shown: one screen, glanceable in a few seconds, every number sitting next to something that says whether it's good or bad. A bare figure with nothing to compare it to is decoration, not information.
Three things the design experts keep returning to. It fits on one screen — Stephen Few names exceeding a single screen as the number-one dashboard mistake. It's built for glancing, not exploring: the Nielsen Norman Group defines a dashboard as a single-page view consumed fast with minimal cognitive processing, which is why bars and lines beat pies and gauges (the eye reads length and position far more accurately than angle). And every number carries context — a target or a trend — so it can be judged in a second.
There's no magic number, and the famous one is a myth. "7±2" comes from George Miller's 1956 paper about telling apart tones and lengths — not tiles on a screen, and the paper gives no support to counting items on a display. A dashboard is a recognition task: the numbers are on screen, you're not recalling them, so the limit barely applies. The real constraint is spatial — everything fits on one screen and reads at a glance. In practice that's usually five to nine headline KPIs, but treat that as a convention, not a law.
One screen, about five numbers, each against its own target or trend: net collection rate (with the AAFP's ≥95% target beside it), days in A/R (against 30–40 days), no-show rate, provider utilisation, and patient retention. Revenue sits on the page as a scoreboard but isn't one of the five levers. Crucially, the metrics with real professional targets show a reference line; the ones with no credible benchmark (retention, utilisation) show your own trend, not an invented "industry standard." That honesty is what separates a usable example from a pretty one.
Not automatically. The landmark meta-analysis of feedback (Kluger & DeNisi, 1996; 607 effect sizes across 23,663 observations) found feedback helped on average but made performance worse in over a third of cases. Showing people numbers is an intervention that can backfire. A dashboard is a tool for a review, not the review itself; on its own it does nothing. What makes it pay off is the boring part around it — a target, an owner for each red number, an action plan — not the chart design.
WRITTEN BY
Olha · clinic data analyst
I build the reporting our managers open every morning at a multi-branch medical clinic — and package it so other practices don't have to start from scratch.
The sample dashboard uses invented numbers to show structure, not to state facts; it's labelled as illustrative. Every real figure — the AAFP targets, the MGMA and peer-reviewed no-show ranges, the Kluger & DeNisi result, the Miller and Ridgway attributions — was checked at its primary source, and the draft was then run through an adversarial fact-check pass. Where a widely-quoted benchmark turned out to have no method behind it, that's said in the sentence rather than buried. Lucid Vitals is not affiliated with the Nielsen Norman Group, Stephen Few / Perceptual Edge, the AAFP, MGMA or Microsoft.