Starting to measure whether any of this works

On July 8, about a week into building ParrotTalk, I did something that felt faintly ridiculous at the time. I installed proper measurement on a site that almost nobody was visiting.

It felt like putting a speedometer on a bicycle I was still pushing up a hill. But I had a suspicion, and it turned out to be right: the habit of measuring is much easier to build when the numbers are too small to hurt your feelings.

Doing it once, properly

The technical part took an afternoon. Rather than paste a tracking snippet into each page and let the copies drift apart over time, I put the loading logic in one small shared file and included that file across the site. One place to change, one place to audit, one place to switch off. Every page that needs it points at the same code.

That sounds like a minor tidiness point. It became one of the better decisions of the early weeks. Months later, when I needed measurement to respect the cookie banner properly, and later still when I needed my own visits to stop polluting the picture, both were single edits in a single file rather than an archaeology expedition across a dozen pages. Doing the boring thing once buys you the right to change your mind cheaply later.

The other rule I set that day: nothing gets sent unless you have accepted cookies. The measurement never loads itself uninvited. If you say no, there is nothing to say no to twice. Given that the site asks you for no account, no email and no payment, it would have been a strange kind of hypocrisy to be greedy about the one thing left.

Humble beginnings

Then I opened the reports, and there was almost nothing there. A handful of visits, most of them mine. The kind of numbers where you can name the people individually.

I want to be honest about that rather than skip past it, because founder journals have a bad habit of starting at the point where the graph turns upward. Mine started at a flat line near zero, and staying honest about that was the entire exercise.

There is a strong pull, when the numbers are that small, to arrange them into a nicer shape. Two visits become "early traction". A single returning visitor becomes "retention". You are not exactly lying, you are just choosing generous words, and generous words are how you end up building the wrong thing for six months. Small numbers are extremely easy to flatter, which is precisely why they are such good training.

What I decided to watch

So I wrote down what I would actually pay attention to, and it was short.

• Does anyone start a test at all.
• Do they finish the section they started, or stop halfway.
• Do they come back a second time.

Not page views. Not time on site. Those go up when people are lost, and I had no interest in a metric that rewards confusion. The three above are the ones a practice tool actually lives or dies by, and they are uncomfortable in the right way: each one can only be improved by making the product genuinely better, never by making the copy shoutier.

The one I braced for was the second. Somebody starting a test tells you the front page did its job. Somebody finishing tells you the test itself was worth their evening. The gap between those two is where all the honest, unflattering information lives, and it is the only number I have ever found that reliably tells you what to build next.

Why measure this early

Two reasons, and neither is about growth.

The first is that you cannot recover a number you did not collect. If I had waited until the traffic was interesting, I would have had no way of knowing what "before" looked like. Every improvement afterwards would have been a matter of opinion and vibes.

The second is the habit itself. Learning to look at a number you do not like, while the numbers are small enough to be painless, means that by the time they matter you already know how to read them without editing them in your head. I would much rather learn to face a disappointing chart when the chart represents four people than four thousand.

The candidate-facing version of the same idea already lived in the product: the dashboard shows you your own bands over time, including the ones that went down. I did not build a version that quietly hides your worse attempts, and I was not about to run my own business on softer terms than I offer you.

It is still a modest set of numbers. But it has been an honest one from the first week, and that turns out to be worth far more than a big one you have learned to misread.

Everything measured here starts with a single sitting of the free Quick Mock Test.