Ecommerce metrics to track, and the ones to ignore

Your analytics page has forty numbers on it. Name the one that tells you what to do on Monday morning.

Most store owners cannot, and that is not a failure of attention. Almost none of those forty numbers is actionable at the volume a small store runs at, and nothing on the page admits it.

So here is the uncomfortable part first, because it is the most useful thing here. If your store takes fewer than roughly a hundred orders a month, nearly every rate on that dashboard is noise. Conversion rate, bounce rate, add-to-cart rate, average session duration: all ratios computed on samples far too small to tell a real change from a quiet fortnight. At that stage you should be counting things, not measuring rates.

The rest of this is which things to count at which stage, and which numbers to stop opening.

Why the rates do not mean anything yet

This is arithmetic, not an opinion, and you can redo it on your own figures in a minute.

Say 200 visitors last month and 4 orders: a 2.0% conversion rate. If exactly one more person had bought, it would read 2.5% — a 25% improvement in your headline metric produced by one individual changing their mind about a candle.

This month, 300 visitors and 9 orders: 3.0%. That looks like a 50% gain and a reason to repeat whatever you did. Move two orders either way, an entirely ordinary amount of variation, and the same month reads 2.3% or 3.7%.

The rough rule is that you need on the order of a hundred conversions on each side of a comparison before a twenty or thirty percent difference is distinguishable from chance, and thousands to see a 10% difference. Nine orders a month is not within an order of magnitude of that.

None of which means skip analytics. Install it on day one, because history you did not collect cannot be recovered. It means do not decide from it yet.

The three things that decide which metrics matter

How many orders you get. Under about a hundred a month, count events; above it, rates start to carry signal.

Whether you are paying for traffic. The moment money goes into ads a second set of numbers becomes mandatory, because you can now lose money faster than you learn.

Whether people buy more than once. If your product repeats, repeat rate is the most valuable number you own and takes months to see. If it genuinely does not, every acquisition must pay for itself on the first order, which moves every threshold below.

Stage one: under about 100 orders a month

Count things that happened. Each of these is a whole number, so it cannot lie to you with a ratio.

• Orders this week, and revenue this week. Written down, same day each week.
• Where each order came from. From asking, not from attribution software: one optional question at checkout. At nine orders a month you can simply know.
• Which products sold, and which never have. The second list is the useful one and nobody looks at it.
• Repeat orders. A count, not a rate: customers who have ordered twice. Zero is information too.
• Contribution per order. Revenue minus goods, shipping and payment fees. The shipping math works this through.
• Cash in the bank, and cash tied up in stock. The two numbers that end small stores.

Your real metric at this stage is traffic, and specifically whether it is growing at all. Everything else is downstream of whether anyone is arriving.

Stage two: steady orders, not yet buying traffic

Now rates start to be worth reading, slowly, and over months rather than weeks.

• Average order value. Stable enough to act on, and the input to a free shipping threshold.
• Conversion rate by channel, quarterly. Not weekly. Differences between channels are large, so they become visible far sooner than small changes within one.
• Repeat purchase rate at 90 days. Of the customers who first bought three months ago, how many bought again. A real cohort, not a lifetime average.
• Refund and return rate by product. One product returning far more than the others is a listing problem, and it shows up here before anywhere else.
• Email list size, and the revenue it produces. The one audience you are not renting.

Stage three: you are spending on acquisition

The arithmetic gets strict, because now a wrong number costs money every day rather than every quarter.

Illustrative figures throughout — substitute yours. Average order $38, contribution per order $18.75.

• Customer acquisition cost, per channel. Spend divided by new customers from that channel. If a channel cost $440 and produced 20 new customers, CAC is $22.
• CAC against first-order contribution. $22 of CAC against $18.75 of contribution means the first order loses you $3.25. That is not automatically wrong, but it is only acceptable if people come back.
• Expected contribution including repeats. If a quarter of customers buy once more, expected contribution is 18.75 + (0.25 x 18.75) = 23.44 against $22 of CAC. Marginal. A tenth of that repeat rate and the channel is simply unprofitable.
• Payback period. Months until a cohort's cumulative contribution exceeds what you paid to acquire it. This decides whether you can afford to grow.
• Blended CAC. All marketing spend divided by all new customers, including the free ones. Per-channel figures flatter themselves; blended does not.

The numbers to stop opening

Each of these costs you the only genuinely scarce resource you have, which is attention.

The test for any metric is a single question: what would I do differently if this number moved by a fifth? If you cannot name the action, it is not a metric, it is decoration.

Where the obvious answer is wrong

"Conversion rate is the key metric." Not at 200 visitors a month. Doubling a 2% conversion rate gets you four extra orders and is very hard; doubling traffic gets you four extra orders and is merely work. Optimize what you can actually move.

"Our best channel is the one with the most orders." Only until you divide by cost. A channel delivering half the orders at a third of the CAC is better, and the orders column will never show it.

"Wait for more data before deciding." Below a hundred orders a month, more data is not coming soon. The alternative to a statistical answer is not paralysis — it is talking to six customers, which takes an afternoon.

"Lifetime value justifies the acquisition cost." Possibly, once cohorts are old enough to show a repeat curve. Three months in, LTV is your first-order value times optimism. Use repeat rates you have actually watched.

How often to look

Weekly, five minutes: orders, revenue, cash, anything on fire. Monthly: AOV, returns, repeat counts, and spend against contribution if you advertise. Quarterly: conversion by channel, cohort repeat rates, products that never sell. Checking a conversion rate daily is not diligence — it is reading a random number generator and having feelings about it.

What analytics will never tell you

It will not tell you why. None of the numbers above explains why someone left, which is usually what you most need. Six customer conversations — including the people who asked a question and never bought — beat any dashboard at your scale.

It will not tell you what to do next. A metric narrows the question; the decision stays judgment made on incomplete information, which is the normal condition of running a store.

And it will not make a small sample large. No tool or AI can extract a reliable conversion rate from nine orders. Anything reporting a confident trend on that data is reporting noise with a nice chart around it.

Where mAutomate fits

mAutomate keeps sales, inventory and analytics in the same dashboard as the storefront and the marketing, so performance sits next to whatever produced it. Advanced analytics is part of the Pro plan; the plans page lists where each piece sits.

The AI surfaces and recommends, and nothing goes live until you approve it — which matters here, because a recommendation drawn from forty orders deserves your skepticism before it deserves your budget. Choosing which ecommerce metrics to track is mostly choosing what to ignore, and no software makes that choice for you.