Your best week is a bad baseline
Record weeks are mostly luck running hot, and every week measured against them looks like decline.
A newsletter goes out on a Tuesday and something clicks. The subject line lands, two people forward it, a small community picks it up, and the week ends with forty signups instead of the usual twelve. On Monday the founder writes "40/week" at the top of the planning doc as the new bar. Three weeks later signups are back around fourteen, and the mood in that doc has turned from celebration to worry. Nothing broke. The bar was simply drawn at the one place it was never going to stay.
This pattern is so common that statisticians gave it a name a long time ago: regression to the mean. When a number is made of a steady underlying rate plus a lot of luck, the most extreme weeks are the ones where the luck ran hottest. The next week is very likely to be less extreme, not because you did something wrong, but because the luck does not repeat on demand. If you pick your best week as the baseline, you have picked the week most likely to make everything after it look like decline.
Why the best week is the worst reference
Invent a small business to see the shape. Say your real, steady rate is about fifteen signups a week, and any given week wobbles by five or six in either direction depending on who happened to see what. Over a quarter of thirteen weeks, one of them will almost certainly land in the mid-twenties just by chance. Add one lucky share on top and forty is not strange at all.
Now look at what each choice of baseline does. Compare every week to forty and twelve out of thirteen weeks are "down" by more than half. Compare to fifteen and most weeks are within a few signups of normal, with one pleasant outlier you can study. Same data, same business, two completely different stories about whether things are working. The story you tell yourself decides whether you keep going, change the page, or quietly stop a channel that was doing its job.
a record is a data point, not a promise.
The same thing happens in the other direction, and it is just as costly. A campaign has a miserable first week, you tweak the headline, and the second week is better. It is very tempting to credit the tweak. Often the second week was simply less unlucky. We covered the downward version of this in reading a dip without panicking; the upward version is quieter, because nobody investigates good news.
Which numbers make a fair baseline
The honest baseline is boring on purpose. Use a median of recent weeks rather than an average, because a single spike drags an average up and barely moves a median. Six to eight weeks is usually enough for a small business: long enough that one good or bad week does not dominate, short enough that it still reflects what you are doing now rather than what you were doing last season.
Keep the spike out of the baseline, but do not throw it away. Write down what was different that week: which email, which post, which referral, which day. A spike is the cheapest research you will ever get, because it tells you what an unusually good week looks like from the inside. The question is not "why can't we hit forty again?" but "what part of that week can we repeat on purpose?" Sometimes the answer is a channel worth a second look. Sometimes it is a one-off mention that no amount of effort will recreate, and that is fine to admit.
Goals, budgets and the trap of the lucky launch
Where this bites hardest is in plans built on the best number. A paid campaign that returned four dollars for every dollar in its first week becomes the justification for tripling the budget, and then it returns two. Part of that is the audience getting more expensive as you scale, which we wrote about in doubling the budget is a new campaign. Part of it is simply that the first week was the lucky draw, and you planned as if luck were a setting.
A safer habit is to set goals against the median plus a modest stretch. If your typical week is fifteen, aim for eighteen. That target is ambitious enough to require real work and close enough that hitting it tells you something. A goal pinned to your best week mostly tells you how lucky that week was.
When you look at a dashboard, it helps to have both numbers side by side: this week, and the typical week over the last couple of months. You do not need anything fancy for this: a spreadsheet with one row per week and a median cell at the bottom takes ten minutes on a Friday, and it will change how most of your Mondays feel.
A rule of thumb
Before you call any week a trend, ask whether it would still look special next to the median of the eight weeks before it. If one week is far above or far below that line, treat it as a story to understand, not a new normal to defend. If three or four weeks in a row sit on the same side, now you are looking at a real change, and it deserves a real decision.