Small Shop Notes

A Fifteen-Minute Weekly Review for a Small Shop: Four Numbers and One Decision

A Fifteen-Minute Weekly Review for a Small Shop: Four Numbers and One Decision

About Alex: a pen name. There is no shop behind it, and no results are claimed. What follows is arithmetic and opinion, and I'll mark which is which.

In short

Pick one weekly slot, copy four numbers into a log (visits, orders, profit per order, and a note on what you changed), and end by writing one decision for next week: keep, change one thing, or wait. With only a few orders a week, a swing of one or two orders is mostly noise, so "wait" is a respectable decision. (My judgement; the arithmetic is below.)

What you get from reading it:

  • A log sheet you can copy into a notebook or spreadsheet
  • A six-week example, with every figure recomputed
  • Rules for reading small counts, and a plan for a week with no orders

Picture a seller with a few orders a week. On Monday she opens the stats and sees two orders, and her mood drops. On Thursday she opens them again, sees four, and decides the new photo is a triumph. On Saturday it is one order, and now the photo is a failure.

Nothing about the shop changed between those three looks. What changed was the number, and the mood followed it.

That is the problem this guide tackles. It isn't a shortage of data; it's that the looking is random and nothing gets written down, so there is no earlier week to compare against. A fixed slot fixes both.

Who this is for. A seller with a few orders a week, on a marketplace or on your own site. If you have no orders at all yet, jump to the section on a week with none. If you want a one-off deep look at the whole shop, Why Your Shop Isn't Growing does that; this guide is the weekly habit that feeds it.

What it costs. Nothing and no new software. A notebook works, and so does a spreadsheet. You need about 15 minutes, once a week.

The 15-minute slot, minute by minute

You'll end up with a repeatable routine, so the weekly look stops being a mood.

Choose a fixed day and time and treat it like a short appointment. Which day is up to you. In my judgement, a good one is a day when the week's orders are mostly in, so you are not reviewing half a week. Outside the slot, I'd try to leave the stats page alone. That is advice for your calm, not a law of nature.

1. Copy (5 min)
Visits, orders, profit per order, change note
→
2. Compare (4 min)
This week against the last four, as a range
→
3. Decide (4 min)
Keep, change one thing, or wait
→
4. Close (2 min)
Write the reason, shut the tab
The 15 minutes: 5 + 4 + 4 + 2. The timings are a suggestion, not a measured optimum.

Four steps, and the log sheet in the next section is where they land. If the first few weeks overrun, that is fine; copying gets quicker once you know where each number lives on your dashboard.

The log sheet: four numbers and a decision

You'll end up with a sheet you can copy by hand or paste into a spreadsheet.

Each number answers a different question.

Here is the sheet. Copy it into a notebook; each week is one block.

WEEKLY LOG        shop: ___________
Week of: ____ to ____
1 Visits:           ____  (state: __)
2 Orders:           ____  (state: __)
3 Profit per order: $ ____
4 What I changed: __________________
  (or: nothing)
--------------------------------------
COMPARE with the last 4 weeks
 visits ____   orders ____
 rate (orders / visits): ____ %
 Is the gap bigger than a few orders?
  yes / no / too small to tell
--------------------------------------
DECISION for next week (circle one)
 KEEP   CHANGE ONE THING   WAIT
Because: __________________________
Re-check on (date): ____

"State" works as in Why Your Shop Isn't Growing: read means you saw the number, 0 only when the dashboard really shows zero, not measured when nothing counts it, unknown when a counter exists but shows nothing, and last known for a figure copied from an earlier week. Mixing those up turns "my dashboard doesn't count this" into "this is zero".

No calculator needed to start: if this week's orders are within two of the average of the last four weeks, write "too small to tell".

Once there are a few weeks, a one-line-per-week table keeps the history readable. The next section shows one.

A worked six weeks for an imaginary shop

You'll end up seeing what the log looks like once it has some history.

Meet Leo. Leo is imaginary, and so is his shop of printable recipe cards. Every number below is invented to show the arithmetic, not to suggest what any real shop gets. I recomputed each figure with Python.

Leo's average order value is $12.00 and his fees and costs come to $5.50 an order, so his profit per order starts at $6.50. In week 5 he raises the price by $1.50. For simplicity I hold his costs at $5.50 (in a real shop, fees might move a little with the price), so profit per order becomes $8.00.

Leo (imaginary), six weeks
Wk  Vis Ord  $/ord  Changed
1    80   2   6.50  baseline
2    95   3   6.50  new 1st image
3    70   1   6.50  nothing
4    88   2   6.50  nothing
5    91   3   8.00  price +1.50
6    84   2   8.00  nothing

Totals over the six weeks: 508 visits and 13 orders, a rate of 13 ÷ 508 = 2.56%. Profit across the six weeks: weeks 1 to 4 had 8 orders at $6.50, which is $52.00; weeks 5 and 6 had 5 orders at $8.00, which is $40.00; together $92.00.

Here is what Leo wrote as his decisions. Each is written at the end of the week, for the week ahead.

Leo (imaginary): decisions
Wk  Decision  Because
1   CHANGE    baseline; try new image
              (it costs nothing)
2   WAIT      3 vs 2 orders: too small
3   WAIT      1 order: inside chance
4   CHANGE    image: no sign either
              way; try a price rise
5   WAIT      3 orders; give it weeks
6   WAIT      price is 2 wks old;
              recheck wk 8

Read the decisions against the numbers. After the image change, weeks 2 to 4 together show 6 orders from 253 visits, 2.4%. Week 1 showed 2 orders from 80 visits, 2.5%. That is no visible difference, so Leo keeps the image for now (it costs nothing) and moves to the next test, one thing at a time.

Leo bends his own four-week rule in weeks 1 and 4 on purpose: the image and the price are cheap and easy to undo, so he tries them while the numbers settle. The rule is about when to trust the order rate, not about whether to try something.

Then the money. Weekly profit averaged $13.00 across weeks 1 to 4 and $20.00 across weeks 5 and 6. Is that the price rise working? Partly, and the arithmetic can separate the parts. At 2.0 orders a week, the extra $1.50 per order adds 2.0 × $1.50 = $3.00 a week.

The other $4.00 comes from orders going from 2.0 to 2.5 a week (half an order at the new $8.00 profit), a gap that two weeks cannot tell from chance. So Leo credits the price with the first part and waits on the second.

Small counts: what not to conclude

You'll end up with a few rules for deciding when a change is real.

With a few orders a week, each order is a big share of the total. A shift from 2 to 3 is "up 50%", and also just one more order. The percentage sounds dramatic; the count is tiny.

A useful picture: even if nothing about your shop changes, the number of orders per week moves around by chance. Each visitor is like a coin flip with a low chance of landing on "buy".

Here is a tiny piece of arithmetic on Leo's numbers. Suppose his true visit-to-order rate really is his six-week figure, 2.56%, and a week brings 85 visits (about his weekly average). Under a simple binomial model, which treats every visit as an independent chance and so is a simplification, the chances come out like this.

So a one-order week and a four-order week both fit a shop whose conditions did not change. That is the point: swings like the ones in Leo's log do not by themselves mean anything happened. (This is arithmetic from a model, not a measurement of any real shop. Real visitors are not identical coins.)

A second tool is a range. A Wilson interval gives a plausible range for a rate, given a count. For Leo's weeks:

Those ranges overlap almost entirely. With this little data, the price-rise weeks and the earlier weeks cannot be told apart on order rate. The ranges are 95% Wilson intervals, computed with Python. If you want to try one yourself, a search for "Wilson score interval" will find calculators.

From that, here are the rules I'd use. They are my judgement, not a statistical law, and a seller with a few thousand visits a week can relax them.

  1. A change of one or two orders in a week is not a result. Write it down, and don't act on it alone.
  2. Compare pooled weeks, not single ones. I'd want at least four weeks on each side of a change before leaning on the order rate, and I'd still check whether the ranges separate, as a rough check and not a test.
  3. One change at a time. If you change the image and the price in the same week, no later number can say which did anything.
  4. Trust arithmetic more than counts. Profit per order is a calculation from your price and costs, so with costs unchanged it moves the moment the price does. What a price rise does to the number of orders is the noisy part.

About rule 2: with about two orders a week, even four weeks on each side (about 8 orders a side) only separates a change of about two and a half times or more: 8 orders against 20 or more. Doubling to 16 is still within chance in a simple exact test. Smaller changes will not show in the order rate, so judge those on cost and ease, not on the log.

A caution that goes the other way: "too small to tell" is not "nothing happened". It means the log has not yet gathered enough to say. Keep logging, and the pooled weeks will say more each month.

Choosing keep, change one thing, or wait

You'll end up with a decision you can write in one line.

The three choices suit different situations, and the table pairs each with a trigger. The triggers are my judgement.

Choose When Example line
KEEP The numbers look fine for your goal, or a change has run four weeks or more and you are leaving it in place "Keep the price; re-check in two weeks"
CHANGE ONE THING No change is running and one number has looked weak for several weeks "Swap the first image"
WAIT A change is under four weeks old, or the counts are a few orders "3 vs 2 orders, too small"

If you aren't sure which number is weak, the stage table in Why Your Shop Isn't Growing names four places a shop can leak. For a page-level idea for the one thing, Audit Your Product Page in 30 Minutes gives ten checks.

Write the reason in the same line. "Wait because 3 vs 2 orders is too small" is a decision. "Wait" alone is a shrug.

Also set a re-check date. A decision with no date drifts; with a date, you know when the answer will be worth reading.

A week with no orders

You'll end up with a plan that doesn't depend on the number going up.

A zero-order week feels like a verdict. Here is how I'd read it, step by step.

  1. Write the 0 honestly. If the dashboard shows 0 orders, write 0 with state read. If you cannot see the figure, write unknown. Check that the date range on your dashboard matches the week you mean.
  2. Look at visits next. If visits are down as well, this week's question is about arrival, not about the page. Orders can't follow visitors who didn't come.
  3. If visits look normal, don't panic. At Leo's rate of 2.56%, a week with 85 visits and no orders has a chance of about 11% under the simple model. That is not rare, so one such week is weak evidence.
  4. Profit per order stays at the last known figure. Don't write 0 there; with no orders there is nothing to average. Copy the last value and mark it last known.
  5. Pick a slot-sized task that doesn't depend on a number. For instance, read your product page as a stranger would, or write the next sentence of the description. Then log it under "What I changed".

If several weeks in a row bring no orders, the pooled picture starts to speak. Ten weeks with 850 visits and no orders would be hard to put down to chance at a 2.56% rate: that is about 3 in 10 billion in the simple model, and still about 3 in a million if the true rate were as low as 1.5%.

A run like that points back to the page or the traffic. The stage table in Why Your Shop Isn't Growing is the place to start.

A different case: if visits are zero too, the log may be telling you the shop is not being found yet. That is the reach stage, and the log still helps, because it shows week by week whether any change moves visits at all.

Your next step

You'll end up with the first block filled in today.

Copy the log sheet and fill in the first block, even if it holds last week's numbers. Put your slot in the calendar. At the end of week four, add up the visits and orders of the four weeks, compute the rate, and write your first decision with a reason.

If you'd like a bigger snapshot of your shop to start from, read Why Your Shop Isn't Growing. If you also log email sign-ups, a guide to reading your first 30 days of email list numbers (not in this set) would do the same for sign-ups; this one is the weekly shop routine and stays with shop visits and orders.

Sources and checks

What this guide does not claim