What is a good ecommerce dashboard?

Online revenue is visitors multiplied by conversion rate multiplied by average order multiplied by how often customers come back. A dashboard that leads with those four tells you not just how the month went but which part of the machine moved.

Put them across the top, each with the current period, the previous period, and the same period a year ago. The year ago comparison matters because most stores are seasonal and a fall against last month can be a completely normal March.

Underneath, show revenue itself and the number of orders, so nobody has to infer them.

The reason this beats a revenue chart is diagnosis. Revenue down with visitors flat and conversion down is a site problem. Revenue down with conversion flat and visitors down is a traffic problem. Revenue flat with visitors up and conversion down is a traffic quality problem, which is a different fix again. One screen, and you already know which team is working on it this week.

Add one profitability line so growth is never celebrated in isolation: total marketing cost as a share of revenue, or contribution after cost of goods and marketing. A store can improve all four numbers and still be worse off if it bought them with discounts, and the dashboard should make that visible.

Keep the top section to a single screen on a laptop. If it needs scrolling, it is a report, not a dashboard.

Beneath the headline, each of the four numbers gets two or three supporting metrics. Not more, because the point is to explain a change, not to store everything.

Under visitors: sessions by source, cost per visitor from paid, and the share of visitors who are returning. Those three separate a paid problem from an organic one and show whether the audience is being rebuilt each month or accumulating.

Under conversion: add to cart rate, cart to checkout rate, and checkout completion, split by device. Those three locate the leak on the product page, at the shipping stage, or at payment, and the device split is where most mobile problems reveal themselves.

Under average order: items per order and the share of orders that sit just below the free shipping threshold. Those two tell you whether bundling is working and whether the threshold is in the right place.

Under repeat purchases: the share of customers who have bought more than once, the time between first and second order, and revenue from automated email and text flows.

That is around a dozen numbers in total, which is about the limit of what a person will actually read every week.

Where you can, show each as a small trend line rather than a single figure, because direction is more informative than a value. And keep the same layout every week, so people learn where to look instead of reading the labels.

The fastest way to make a dashboard useless is to let two systems answer the same question. Decide the source for each number and put it next to the number.

Revenue and orders come from the store admin, because that is the ledger. Traffic, sources and on site behaviour come from analytics. Spend comes from the ad platforms. Email and text performance comes from the sending platform. Nothing appears twice from two places.

Then write the definitions where people can see them. Does revenue include tax and shipping. Are refunds deducted, and in which period. Does conversion rate use sessions or users. Is a returning customer defined by email address or by account. Are subscription renewals orders. Every one of those is a legitimate choice, and the harm comes from leaving it unstated so two people report different figures.

Date handling deserves the same treatment. Fix the time zone across every source, state whether the period is calendar month or a rolling four weeks, and be consistent. Rolling periods are better for spotting change, calendar months are better for comparing with accounts.

Add a last updated stamp and a note on which numbers are still settling. Attribution and returns both move for days after a period closes, so a figure read on the first of the month is not final.

Finally, keep the dashboard where the accounts and data live in the store's own name, so the reporting does not disappear if the person building it changes.

Different decisions run on different clocks, so build up to three views rather than one enormous screen.

The daily view is operational and short: yesterday's orders, spend, any campaign that has stopped, stock on best sellers, and anything broken. It exists to catch problems, not to judge performance, and nobody should be changing strategy from it.

The weekly view is the four numbers and their diagnostics, compared with the previous weeks. This is where most decisions are made: what to test, where the leak is, which campaign to adjust.

The monthly view adds profitability, cohort behaviour, channel mix, and progress against whatever the quarter's plan said. It is the one that goes to whoever is not looking at the business daily.

Then be firm about what to leave off. Impressions and reach, unless a specific decision depends on them. Social followers. Bounce rate on its own. Any metric that cannot be acted on. And ad platform revenue reported as though it were company revenue, because every platform claims orders the others also claim, and adding them together produces a number larger than the business.

A good test for any tile: if this moved twenty percent, what would we do differently. If the answer is nothing, remove it. A dashboard with a dozen numbers people read every week is worth more than fifty nobody opens.