Sales pipeline metrics that matter (and the ones that do not)
A pipeline metric earns its place only if you can name the decision it would change. Apply that test and most dashboards lose about half their tiles. The six below survive it: coverage against target, win rate by owner, average cycle length, slippage rate, forecast bias, and the concentration of uncertainty. The rest are mostly activity counts wearing a suit.
The six worth tracking
1. Coverage against target. Gross open pipeline divided by the number you need. Decides whether to generate more pipeline or work what you have. It is a pipeline measure, not a forecast — a comfortable coverage ratio says nothing about your odds of hitting the number.
2. Win rate, by owner. Not company-wide, which averages away everything useful. Per person it decides coaching, territory and how much to discount each rep's calls.
3. Sales cycle length. Median, not mean — a single eighteen-month enterprise deal will drag a mean somewhere useless. Decides how far ahead you need pipeline, and when a deal has gone quiet enough to be dead.
4. Slippage rate. What share of deals forecast for a month actually closed in it. This is the metric most teams are missing and it is the best early warning there is: slippage rises before revenue falls.
5. Forecast bias. The average signed error between forecast and actual. Decides how much to adjust this month's number. A team that is consistently 20% high can simply subtract 20%, which is the cheapest accuracy improvement available to anyone.
6. Concentration of uncertainty. How much of the quarter's doubt sits in the top few deals. Compute p × (1 − p) × value² per deal and look at the top three as a share of the total. Decides what the review spends its time on, and whether your forecast is an average or a bet.
Of those, the two almost nobody tracks are slippage and bias — and they are the two that tell you something about the future rather than describing the present.
Four that mislead
- Total pipeline value. Goes up when deals are added and when deals go stale. A number that rises for both good and bad reasons cannot be read.
- Activity counts. Calls, emails, meetings booked. They measure effort, correlate weakly with outcomes, and reliably produce the behaviour they measure rather than the one you wanted.
- Average deal size. In a lumpy pipeline this is a fiction — the mean sits between the big deals and the small ones and describes neither. Look at the distribution.
- Stage conversion rates, for forecasting. Fine for diagnosing where deals stall. Poor as a probability, because stage reflects your process rather than the buyer's intent.
Two rules about shape
Prefer medians to means for anything where one deal can dominate: cycle length, deal size, time in stage. The mean of a skewed distribution is a number no deal resembles.
Prefer trends to levels. A 3.2× coverage ratio means little in isolation. The same figure having fallen from 4.1× over two months means something specific and actionable.
What a useful dashboard looks like
Six numbers, each with its previous value beside it, and one sentence per number saying what you would do if it moved. If you cannot write that sentence, remove the tile. The measurement has a cost — someone maintains the data behind it — and a metric nobody acts on is pure overhead that also makes the real signals harder to see.