Getting real signal from analytics while your numbers are still tiny

Forty views. Two buyers. You look at it on Monday, then again at lunchtime, then again before bed, and by Wednesday you have constructed an entire theory about your pricing. Small numbers are dangerous precisely because they are readable — you can hold every single data point in your head, which makes it feel as though you understand them.

Why small numbers mislead

Two buyers out of forty views is 5%. One buyer out of forty is 2.5%. Three is 7.5%. Nothing about your page, your price or your product changed between those three worlds — one person's Tuesday afternoon changed. When your denominator is small, every individual human being is worth several percentage points, and a percentage that moves that much is not measuring your business, it is measuring randomness.

The second trap is looking at the same small number repeatedly. Checking analytics four times a day does not add information, it adds emotion. You end up reacting to the gap between checks rather than to anything that happened, and the reaction is usually to change something, which destroys the only thing that would have made the week readable: consistency.

The honest position is that at this size you cannot answer subtle questions. You cannot tell whether £29 beats £34, or whether the second headline is better than the first. What you can answer is blunt: is anyone arriving, are they arriving at the right page, and does anything at all sell. Those three are enough to be getting on with.

Reading a small week badly, or well

Almost everyone reads a first month the same way: they look for a trend line and try to extrapolate it. There is no trend line in four data points. There is, however, a comparison worth making, and a set of counts worth writing down.

A method that works at small scale

Open analytics in your dashboard once a week, at the same time, and follow the same order every time. The order matters more than the numbers, because it stops you from starting with whatever looks most alarming.

1. 1Write down the raw counts per product: views, buyers, revenue. Counts, not percentages.
2. 2Ask the arrival question first. Under about 50 views on a product in a week, you do not have a conversion problem, you have a traffic problem, and no amount of page editing will fix it.
3. 3If views are healthy and buyers are zero across two weeks, look at the page: the first screen, the price, and whether it is obvious what someone receives.
4. 4Compare products against each other in the same week. Same traffic, same you, same weather — the differences between them are the most honest signal you have.
5. 5Note the traffic source next to the counts, so a spike is attached to its cause rather than floating free.
6. 6Change one thing, write down the date, and do not touch it again for two weeks or 200 views, whichever comes first.

Two products, one week

Say a week gives you this: the template had 40 views and 2 sales, the mini-course had 38 views and 0. It is tempting to conclude that the course is priced wrong. Hold on. Two versus zero on forty views is well within the range of ordinary luck, and you would see that result reasonably often even if both pages performed identically.

Now run the same week again, and again, and suppose the pattern holds: 6 sales to 4 across three weeks and 120 views each. That is still not proof, but it is now worth acting on, because the direction has been stable while nothing else changed. The other thing worth noticing is smaller and often ignored: 38 views on the mini-course means barely anyone is finding it, which is a distribution problem you can fix this week, unlike a conversion problem you cannot yet measure. Adding that product as a card on your link-in-bio page and mentioning it in one post is a bigger lever than any edit to the price.

The mistake most people make

The mistake is quitting a product on evidence that could not possibly support the decision. Someone launches, gets 60 views and one sale over two weeks, decides the market does not want it, and deletes the page. Sixty views is roughly the number of people who would walk past a market stall in ten minutes. Nobody would close a stall on that. Before you retire anything, check that at least a few hundred of the right people have actually seen it, and that you gave the page more than a fortnight to be found.

At small scale you are not measuring performance. You are measuring whether you have distribution yet.

Small numbers deserve to be treated gently: written down weekly, compared sideways rather than backwards, and left alone in between. If you are still setting things up, the features overview shows what gets measured and where. Do that for two months and you will have something that no amount of refreshing produces — a short record of what you did and what followed, which is the only thing that ever turns tiny numbers into a decision you can trust.