A small test still has to include the work.

The part that has to work

In Fare, my book on personal effectiveness, I describe a prototype as a small working version of a project. It should give you information you cannot get by continuing to think about the project. The word working matters here.

For example, the book suggests that someone considering a bar spend time working with an owner who can explain how it runs. Getting close to the work can reveal things that a course for obtaining the licence will not. The point is to find out what you would actually be taking on.

Now take an existing business considering a new service. The example that follows is hypothetical: a company sells equipment and customers ask for faster help when it stops working. The founder is considering a paid priority-support service.

Putting a button on the website and counting enquiries would establish whether people are interested. Useful, but the difficult part is still untouched. Can the company provide that support at a price customers will accept, without making its existing commitments harder to keep?

I would want the test to include handling the work. Agree a limited paid trial with a customer, with the promised level of support clear beforehand. The person doing it needs time allocated for it. Calling it a trial does not make that person's other work disappear.

Suppose the trial brings a request that only the founder knows how to answer. The founder deals with it quickly, the customer is pleased, and the service appears to have worked. But if the founder's time is left out of the assessment, the test has concealed a cost that will return with the next customer. It has also left a question unanswered: who could provide this service if more customers bought it?

So the trial has to follow a request far enough to show how it was resolved and whose time it used. There may be very little to record. What matters is that the account includes the work that would otherwise be easy to overlook.

What the test changes

For this example, suppose customers accept the trial price, but delivery repeatedly requires the founder's intervention and delays existing work. That is a reason to hold off on selling the service more widely. More enquiries would increase the problem the trial has just exposed.

The next attempt would need a different delivery arrangement. Perhaps the person providing support needs specific training. Perhaps the promise needs to be narrower. Choose the change that addresses what actually happened, then test that arrangement before committing more capacity.

And if the company cannot spare enough time to run even the limited trial properly, I would leave the project unopened for now. A half-run trial would not establish that the idea is bad. It would establish very little.

This is why I ask what information will only become available once the work is attempted. In this example, the customer had already expressed interest. The missing information was what it would cost the company to keep the promise. That is what the test needed to find out.