Test plan template for Healthcare

The test plan with three columns a provider needs: which service line the test runs against, which claim the variant makes, and the clinical review that cleared both arms of it.

Everything else is the general test plan, unchanged. One hypothesis with a because in it, one variable, a named control, a minimum run, a read date set in advance and a decision rule for every outcome.

Use it when a change to wording needs clearing before it can be tried, and when the number that matters is an appointment somebody attended rather than an enquiry somebody sent.

Use it again as a record of what each service has been allowed to claim, and what was narrowed. In a practice that record is harder to reconstruct than any performance number.

A health test begins with a sentence, not with a design. The claim the variant intends to make goes to clinical review first, and only a cleared claim becomes creative.

That order is what the claim under test column enforces. Write the claim plainly — a waiting time, a process, a credential, something about what a first appointment involves — and have somebody qualified read it against what the service can actually support. An hour of review before the build replaces a round of rework after it, and the rework always lands after the test was meant to start.

Both arms need a reference. The control is usually already cleared, but a test that moves a claim to a different position, or removes the context around it, can change what the control is saying too.

Most claims in this category come back narrowed rather than refused outright: a timeframe qualified, an outcome softened, a comparison removed.

Those edits are findings, and they belong in the row with the date and the reviewer. Over a year they become the only reliable record of what each service may say and in what words, which is information a practice otherwise keeps in one person's memory.

It also stops the annual cycle where somebody proposes an outcome claim, it is narrowed to nothing, the test runs on a weakened version and loses, and the whole idea is written off for the wrong reason.

The claims most likely to survive review are about process, and they are also the ones most likely to move the number.

How long the wait is. Whether a referral is needed. What the first appointment involves. What it costs. Whether parking exists. Those are checkable statements, they are usually cleared without difficulty, and they answer the question that is actually stopping somebody from booking — which is almost never whether the service works, and almost always whether it will be difficult.

Outcome claims are what most teams reach for first and what most reviews reduce to nothing. A test plan that concentrates on process will get more rows into market and will move attendance more often.

The specific trap here is that the outcome depends on something the advertising does not control.

A variant that produced more enquiries during a week when the appointment book was full will show fewer attended appointments than a control that ran when there was room. The ad was better and the result is worse, and nothing in the reporting explains why.

So capacity is written into the notes across the run, and a test should not be started in a period where availability is expected to change sharply — a practitioner on leave, a new clinician starting, a seasonal peak. Where capacity does move mid-run, record it and treat the row as inconclusive rather than adopting a finding that is really about the book.

Enquiries are easy to move and weakly connected to what fills a clinic. A more urgent creative will produce more messages, a proportion of which never turn into anything, and a plan that reads on enquiries will adopt it.

Where the practice management system records attendance, read on attendance, and use the same words in the row that the system uses so the two documents can sit beside each other. The read date then has to allow the appointments to happen, which usually means four weeks rather than one.

For services that work mainly through referral, the equivalent metric is referrals received, and the same logic applies: count the thing that becomes a patient.

Most practices cannot read a small difference. A service producing a handful of appointments a week will never distinguish a modest improvement from an ordinary quiet fortnight.

The response is to test things large enough to see: the whole message, the audience type, the landing destination, whether the ad names a price or a wait at all. Testing button colours or minor copy edits consumes the traffic a real test needed and produces rows nobody should believe.

Where a difference genuinely cannot be read, say so in the result. Not enough appointments on either side is an honest outcome and it is better than a confident one built on three.

The last invisible variable is reception. If the enquiries from the variant are answered faster, or by a different person, or with a different script, the test is measuring the handling.

This happens most often when a test is visible internally and somebody is keen for it to work. Write into the row that both arms are handled identically, and tell the front desk that the test exists only in the ad.

Every row has a hypothesis with a because in it. Every row names the claim the variant makes, in plain words, and both arms carry a clinical review reference with a date and a reviewer. The primary metric is an attended appointment where the system records one. Capacity across the run is written down, and a row where it moved is marked inconclusive. Enquiries from both arms are handled identically. And narrowed claims are recorded as rows too, because that history is the half of this document that compounds.