First 100 days mistakes: reflexes to avoid in a slow month

Most early mistakes are not ignorance. They are reflexes: a reasonable response to a situation, applied to a situation that only looks like the one it was learned for. A slow month in the first hundred days is where those reflexes arrive in a crowd.

So this page collects the mistakes that pressure produces, each paired with its replacement. The replacement is usually slower and almost always cheaper, and a few of them keep you clear of problems that no dashboard would ever show you.

What to take away

Four reflexes that spend money in a slow month

Four reflexes spend money or attention and teach you nothing. Each answers a number with an action before the number has been understood. In every row the replacement is the same move: look, count or wait first.

Mistakes of timing

Slow weeks make early readings feel like verdicts. Reach is visible immediately and demand is not, so most timing mistakes are conclusions delivered on evidence that has not finished arriving.

Reading retention before a cycle has passed. Retention is a claim about what happens after enough time, and in week three not enough time has happened. Every early retention number is dominated by people who have not had the chance to leave.

Declaring the launch a failure in week one. Reach is visible immediately. Demand is not. A quiet first week is almost always a distribution result, and a distribution result wants a different response entirely, one that starts with launch planning mistakes rather than a product rewrite.

Asking a question the evidence cannot answer yet. Note the lag and schedule the question for the week it becomes answerable. This is the single habit that separates a hundred days of learning from a hundred days of arguing.

Mistakes of counting

Pressure also bends arithmetic. A disappointing week makes any larger-looking figure attractive, and small samples can be made to look like almost anything, so the counting rules matter most when they are least welcome.

Reporting rates without counts. Two out of nine displays as a percentage identically to two hundred out of nine hundred. Always the fraction, and always the check: if one event had gone the other way, does the conclusion change. Problem discovery metrics show what small denominators can carry.

Projecting customer value from a curve you have not watched. It produces a large number that makes every other decision look justified. Anything past the weeks you have observed is an assumption; write it as one in the same sentence, using your own arithmetic rather than a borrowed ratio. Market validation metrics show how to keep such figures separate from observed ones.

Mistakes of conduct

These cost more than the numbers do, and they happen in the weeks when the numbers are disappointing. Each one trades something that is hard to rebuild for a better-looking month.

Adding claims to the page under pressure. Time saved, results delivered, people served. The FTC expects an advertising claim to have a reasonable basis before it runs, and early is exactly when unsupported claims get added because they feel like a way out of a slow week.

Turning early customers into testimonials they did not mean to give. Enthusiasm in a support email is not an endorsement, an incentive changes what has to be disclosed, and an edited quote is not what the person said. The FTC's endorsement guidance covers what a testimonial has to reflect and what has to be disclosed alongside it.

Making cancellation difficult when the curve looks bad. It improves the number and destroys the evidence, and where billing recurs it runs into the FTC's negative option rule on consent and cancellation. A customer who cannot cancel becomes a customer who disputes the charge and tells other people why.

Over-promising delivery to close a slow month. A date you cannot hold is a refund and a lost reference, and the second costs more.

Mistakes about people

The last pressure is social. A slow month makes you want reassurance, and reassurance is easy to collect from the wrong questions and the wrong readers.

Asking about the future. What people say they would do at some later price, in some later situation, is a prediction from someone with no stake. Ask what they have done. Customer interviews cover the same failure mode.

Deciding alone what the feedback means. Two people reading the same question log independently and then comparing is a twenty-minute exercise that catches most misreadings.

The one habit that prevents most of them

Write the week down before you act on it. Counts with denominators, what surprised you, and one line on what you changed.

A week that is written cannot be reinterpreted later, which removes the raw material of nearly every mistake on this page. It also makes the four-week direction visible, and a direction that holds for four weeks is the only signal available at these sizes. What to do with it is the subject of the first hundred days.

Common questions

How do I stop myself explaining every movement?

Write the movement down with no explanation, and put the explanation in the following week's entry. Half of them will not need one by then, and the ones that do will have a second data point beside them.

What if the loudest customer is also the biggest?

Then their request is a commercial decision, not a product signal, and it should be written down as one. The confusion happens when a commercial decision gets recorded as evidence about the market.

How many cancellations before a pattern is real?

Fewer than you think for generating a hypothesis, more than you have for confirming one. Three people saying the same unexpected thing is worth investigating. It is not yet a finding.