Back to Kriterion

Why it needs time before it says anything

What Kriterion can tell you in week one, what it cannot, and exactly what has to happen before the number means something.

Nobody is any good at a job in their first week. An app that gives you a confident answer on day one has not learned anything about you. It has looked you up in a table.

What has to happen first

Kriterion works out your burn rate from what you ate against what your weight did. Both halves of that need to exist before there is anything to work with, so the app waits rather than guessing early and correcting loudly later.

Before thisYou need
Any estimate from your own data exists at all Two weigh-ins within a rolling four weeks. Weighing in less than twice a month leaves nothing to fit a trend to
A margin is shown beside it Three weigh-ins across your history, so there is enough spread to say how noisy your own weight actually is
A week counts towards the calculation Four logged days in it, covering most of the week's span. A week where you logged Monday and then went quiet is not a week the app will pretend to read

None of that is a countdown you are waiting out. It is a description of when there is enough to say something honest.

The formula does not vanish. It fades.

Day one, a supply teacher goes off the register, because the register is all they have. By half term they know the class and they do not need it.

Kriterion starts with the textbook equation for the same reason, then lets it fade as your own data arrives. Early on the number leans on the formula. Later it leans on you. There is no moment where it switches over and everything jumps; the weight simply shifts, which is why the figure moves gradually rather than lurching.

What the range is doing

The margin beside the number is not decoration and it is not a legal hedge. It tightens every time a week's worth of real data lands, and it widens during any stretch where nothing does.

That includes an ordinary gap. Go quiet for a fortnight and the range grows, because the app genuinely is less sure than it was. This is the opposite of how most software behaves: normally the number just sits there, looking equally confident whether it was calculated this morning or last spring.

It will not always go down. The range narrows when data arrives and widens when it does not, so a fortnight of good logging tightens it and a fortnight off does not. We would rather say that than promise a line that only goes one way.

What you get in the meantime

Everything except the measured number. Logging, your whole history, workouts, the timer, cycle tracking, readiness, weight trends, the coach that explains its call and will adapt your session if you let it. None of that waits on anything.

So the first fortnight is not a trial period with the good bit locked. It is the ordinary app, while it gets to know you.

Why this is the honest trade

An app that hands you a confident figure on day one is not being more useful than one that waits. It is being less truthful about the same uncertainty. The number it gives you is exactly as much of a guess either way; the difference is whether it says so.

The version that waits is slower, and you should know that before you start. It is also the only version that can ever tell you it was wrong.

The rest of the argument. How the measuring actually works, what the free tier includes, or join the list and we will tell you once when it ships.