Five metrics that matter, and the fifty that don't

The usual response to a business decision that went badly is to start tracking more things, on the reasonable theory that the missing input was information. What follows is a dashboard with thirty figures on it, reviewed for the first two weeks and then not at all. This is not a failure of discipline. Thirty numbers cannot be held in mind at once, cannot be compared against each other, and cannot each have someone responsible for acting on them, so what actually happens is that attention settles on whichever one moved most dramatically — which is usually the one with the smallest denominator and the least meaning.

A short list works for the opposite reason. Five numbers can be remembered, so you notice when one is unusual without being told. Five numbers can each have a stated target, so "unusual" has a definition. And five numbers can be argued about properly in half an hour, which is the actual mechanism by which measurement changes anything. The discipline worth acquiring is not collecting more; it is deciding what you will ignore, deliberately, and writing that decision down so it survives the next enthusiastic afternoon.

Before adding any figure to a list you intend to keep, answer one question: what would you do differently if this number moved substantially in either direction? If there is no answer, the metric is decoration. It may be interesting, it may be genuinely true, and tracking it still costs you attention that a metric with a decision attached could have used.

This test disqualifies most of what gets tracked. Follower counts, page views, impressions, email open rates and app downloads are all real measurements that rarely change a decision for a small business, because there is no lever attached to them — you cannot do anything specific about impressions, and the actions people take in response are usually things they would have done anyway. The technical term for this class is a vanity metric, and the defining feature is not that it is false but that it moves in the same direction as good news without telling you what caused it. Its close relative is the metric that only ever goes up: cumulative totals, lifetime customers served, all-time revenue. A number that cannot fall cannot warn you about anything.

Revenue is on the list because it must be, with one condition: track it per period rather than cumulatively, and against the same period last year rather than last month, so seasonality does not masquerade as a trend. Gross margin — revenue minus the direct costs of delivery, as a percentage of revenue — is the number that tells you whether revenue is worth having. It comes from your sales records and your purchase records, and it is worth calculating per product as well as overall, because a blended figure conceals the case where the fastest-growing line is the worst-margin one.

Customer acquisition cost is everything spent on winning customers in a period divided by the number of new customers gained in it. Two disciplines make it honest. Count new customers, not orders, since repeat orders were not bought by that spending. And resist the per-channel breakdown until you trust the total, because the total is arithmetic while the split depends on assumptions about which touchpoint deserves credit — assumptions that are frequently wrong and always invisible in the resulting figure. Compare acquisition cost against gross margin per customer; if it exceeds the margin on a first purchase, the business is relying on customers returning, which makes the next metric load-bearing rather than optional.

Cash runway converts your bank balance into a deadline: the current balance divided by average monthly cash outflow, expressed in months. It is the most clarifying number available to a small business because it answers the question every other metric dances around — how long you have. Calculate it from actual outflow over the last three months rather than a budget, since budgets describe intentions and the bank statement describes what happened. Watch the direction more than the level; a runway shortening steadily while revenue grows is the specific signature of growth consuming cash, and it is much easier to act on early.

Repeat purchase rate is the share of customers who buy more than once, measured over a window long enough to suit your buying cycle. It is the metric most often missing and most often decisive, because it determines whether acquisition spending is an investment or an expense. It also degrades gracefully: even a crude version, counted by hand from a sales list, tells you something real. The window matters and should be stated, because "repeat rate" measured over three months and over two years are different numbers, and quoting one while thinking of the other is a common way to be confidently wrong about the health of a business.

Metrics change behaviour through a rhythm, not a report. Fix a time — the same half hour weekly — and put the five figures in one place where each new entry sits directly beneath the last. Beside each, a target or an expected range, so a number can be described as off rather than merely noted. Beside that, one line of plain writing on anything unusual and what you intend to do about it. The written line matters more than it looks: it converts noticing into a decision, and it means that in three months you can still reconstruct why a month was strange.

Two things keep this alive. Someone has to be accountable for each metric, even if that is you for all five, because a number nobody owns is a number nobody acts on. And the review has to be allowed to conclude that nothing needs doing, otherwise it degenerates into inventing activity to justify the meeting. What you are building is a memory of your own business precise enough to argue with. Owners who sustain it describe the same benefit, and it is not better decisions in the moment — it is being surprised less often, because a trend visible for four weeks was acted on in week two.

They report what happened; they do not explain it. A gross margin that fell three points does not say whether a supplier raised prices, whether discounting became habitual, or whether the product mix shifted, and distinguishing between those means going back to the underlying transactions. Every one of these numbers is also an average over a period, which means it hides its own distribution: an acquisition cost of ₹900 is compatible with most customers costing ₹300 and a few costing ₹5,000, and those are different businesses requiring different responses.

They are also lagging. Each one describes a period that has closed, produced by decisions made earlier, which is the argument for watching direction rather than reacting to levels and for pairing them with a forward view of cash. And they cannot tell you whether you are measuring the right five — that depends on what your business actually turns on, and the honest process is to review the list itself once or twice a year and ask whether any of them has stopped driving a decision. A metric that has not changed anyone's mind in a year has become decoration, however good a metric it was when you chose it.