Our monthly report is one page and it contains four numbers: visitors, conversion rate, average order value and repeat purchase rate. Under each one, what moved, what we did, and what happens next. That is the whole document.
The reason is not minimalism for its own sake. It is that a report should support a decision, and a long report supports none. When forty metrics arrive at once, the reader looks for the ones that went up, and everybody involved gradually stops being able to tell a good month from a lucky one.
Four numbers work because they multiply into revenue. If revenue moved, one of them moved, and the report shows which. There is no room to hide a bad month behind a chart of impressions, and no need to celebrate a metric that has no connection to money.
It also keeps us honest about our own work. If we spent the month on something, it should show up in one of the four. If it did not, we say so.
Visitors, with the sources separated
The first number is how many people arrived, split by where they came from and what they cost.
We show paid, organic, email, social and direct separately, because a total is not actionable. A month where paid traffic grew and everything else shrank is a different business from a month where the reverse happened, even when the totals match.
Alongside the count we show what the paid visitors cost and how many orders they produced, so cost per visitor sits next to cost per order. A cheap visitor who never buys is expensive, and only the second number reveals it.
We also show returning visitor share, because a store that is genuinely growing has a rising proportion of people who come back on their own. That is the number that says whether the demand you are building is yours or rented.
What we deliberately leave out is impressions, reach and anything else that counts people who did not arrive. Those numbers can be useful when diagnosing a specific campaign. They do not belong in a document about whether the business grew.
If we spent the month on something, it should show up in one of the four. If it did not, we say so.
Conversion rate, by device and by step
The second number is the share of visitors who bought, and we show it split by device and by funnel step rather than as a single figure.
The single figure moves when your traffic mix moves, so on its own it is as likely to mislead as to inform. A month with a successful social campaign will usually show a lower blended conversion rate and more orders, and a report that shows only the blend makes that look like a failure.
The step view is where the work is. Add-to-cart rate, cart to checkout, checkout completion. When one of those moved we say which, and connect it to whatever we changed that month.
Mobile and desktop are always separate lines. Most stores get most of their traffic from phones and most of their revenue from a mix, and the gap between the two devices is one of the most reliable indicators of where the next fix is.
When a change we made was on a specific set of pages, we report the rate for those pages, not for the store. Diluting a real result into a store-wide average is the easiest way to conclude that nothing works.
Average order value, next to items per order
The third number never appears alone. It always sits beside items per order and beside gross margin per order, because on its own it can rise for reasons nobody should be pleased about.
If the average went up and items per order went up with it, customers changed their behaviour and the merchandising is working. If the average went up while items per order stayed flat, the mix changed or a discount ended.
Margin per order is the check on all of it. A larger order bought with a discount that gives back more than it gains is not a win, and this line is where that shows up.
We also report the count of orders sitting just below the free shipping threshold, when a store has one, because it is the most direct evidence of whether the line is in the right place.
Again, we split by channel. Different sources bring different baskets, and reporting a single store-wide average makes a change in traffic mix look like a change in merchandising.
Repeat purchases, read as a trend
The fourth number is the share of customers who bought more than once, and it is the only one we do not expect to move in a month.
So we report it as a line over the last twelve months rather than as a figure with an arrow. Anything else invites overreaction to noise.
Beside it goes time between first and second order, which sets the timing of every flow we build, and revenue from each flow separately: welcome, post-purchase, replenishment, win-back. Separating them means we can say which automation is earning and which is decoration.
We show new customers and returning customers as two counts, not one ratio, because the ratio moves in confusing ways during a strong acquisition month. Two counts cannot mislead you the same way.
And we report list health honestly. Growth in subscribers is only worth reporting next to whether they open anything. A list that grows while engagement falls is a liability being recorded as an asset.
What the page says besides the numbers
Three short sections sit under the four, and they are the part people actually read.
What we did. A plain list of the work that shipped, with the lever each piece of work was aimed at. Not activity for its own sake, and not hours.
What it did. The honest connection between the work and the numbers, including the times when we cannot draw one. Some months a change is too small to separate from noise, and saying so is more useful than a story. If something did not work, it goes here in the same plain language as the things that did.
What is next. The one or two things starting in the coming month, and why those and not something else. This is where the plan gets adjusted in public, so nobody is surprised by where the effort went.
We write it the day it is due, with the same structure every month, so the reports stack into a record you can read end to end. A year of these tells you more about a store than any dashboard, because it contains what was expected as well as what happened.