Average Sales Cycle Describes Winners Only

Why this matters

The average sales cycle is the number shops reach for when they need a forecast or a follow-up horizon, and it can answer neither question, because it is computed only on the leads that were won. Deals that never closed have no cycle at all and contribute nothing. The figure therefore describes the shape of your wins, and a shop that uses it to decide how long to keep chasing a live opportunity will abandon a fifth of its own future jobs on schedule, every quarter, and never see it happen.

What it counts, and the two populations it cannot see

Mean days from lead created to lead won, across won leads only. That is the whole construction, and it is a textbook case of the survivor-bias shape a sibling card owns: an average over completed cases that excludes the still-open ones, which are the difficult ones.

A quarter, 140 leads worked, read at the end of it:

The reported sales cycle is 11.4 days, computed on 46 of the 140. The other 94 are invisible to it, and between them they hold every fact about how long a decision can take.

Here is the win distribution behind that mean:

Forty-six wins. The median sits between the 23rd and 24th values, both in the 4-to-7 band, so the median win is around day 6 while the mean is 11.4 days, pulled up by the ten wins past day 14.

The horizon it gets used for, and what that costs

The owner sets a follow-up rule: our cycle is about eleven days, so call it two weeks, and after two weeks a lead is a lost cause.

Count what that rule abandons. Wins landing after day 14: 6 plus 3 plus 1, or 10 of the 46 wins, 21.7 percent of them. The rule throws away, by design, the leads that produced more than a fifth of the quarter's jobs.

It is worse than it looks, because the rule is self-confirming. Leads dropped at day 14 never win, so next quarter's win distribution has a shorter tail, so the mean cycle falls, so the horizon shortens again. Two or three quarters of that and the shop genuinely does have an eleven-day business, having manufactured it.

The mean was never a horizon. It is the centre of the wins, and a horizon has to come from the far end of something.

Where the 11.4 days actually went

Before deciding the cycle is too long, find out whose time it is. Break the mean into its segments:

Use means for this and not medians. Means of the parts add to the mean of the whole, provided every win in the population carries a recorded date for every segment; medians of the parts do not add to anything. Where some wins closed without an estimate or without a recorded estimate date, compute all three segment means over only the wins carrying all three dates, and print how many wins that is. A shop that decomposes a median cycle into median segments will produce a set of numbers that do not sum and then spend an afternoon looking for the missing days.

So 3.0 of the 11.4 days, 26.3 percent of the mean cycle, is time before the customer has a number in their hands, and that portion belongs to the shop. The remaining 8.4 days, 73.7 percent, is the customer deciding. Cutting the estimate turnaround in half would take about 1.3 days off the mean cycle, which is real but is not where the length is.

One segment needs a warning attached before anyone acts on its size. The 0.4 days to first contact looks negligible here and it is not, because this decomposition is computed over wins only, so it measures how fast the shop answered the leads it went on to win. Speed of first response does not shorten a cycle. It decides whether a lead becomes a win at all, which shows up in the win count and never in the duration of the wins. Judging response speed by its share of the cycle is reading the wrong number entirely, and a sibling card covers what that window is actually worth.

Losses take longer than wins

The loss timing is the number the horizon should have been built from, and most shops have it and never look at it. Here it is 19 days against the wins' 11.4.

The direction is not an accident. A yes arrives as an event: the customer calls, signs, or books. A no usually arrives as silence, and silence has to age before anybody is willing to write it down. Expect time-to-loss to run longer than time-to-win in almost any shop, and treat the reverse as a signal that losses are being recorded promptly and wins are not.

One caution attaches to the loss figure in the same breath, because it is easy to corrupt. If the office marks dead leads in a monthly sweep rather than when the decision is known, time-to-loss measures the gap to the next sweep, not the gap to the decision. The tell is a loss distribution with a spike at a uniform interval or on the same day each month. Where that is what you see, the loss timing is not usable yet, and the fix is to mark the outcome at the contact that established it.

The open list, and how old it is

The 42 still-open leads are the population nobody reports on, and they are where the forecast actually lives.

Forty-two leads. 17 of them, 40.5 percent of the open count, are older than 60 days. On the won side, 1 of 46 wins, 2.2 percent, took longer than 60 days.

Those two shares sit over different populations and are not comparable as numbers. Read together they say something simple: most of the open list is sitting past the age at which this shop's wins have historically arrived. That is not a probability for any single lead, and it should not be quoted as one. It is a statement about where the wins came from, and it is enough to act on.

Forecasting with it without lying

The cycle can do one honest job: given wins that are going to happen, it tells you roughly when they will land. It cannot tell you which leads those are, and the two questions get merged constantly.

This shop won 46 of the 140 leads it worked, 32.9 percent, and that is a floor on the eventual rate rather than the rate itself, because 42 of those 140, 30.0 percent, were still open when it was computed and some of them will win. Applied to the 16 open leads still inside day 19 it gives about 5 wins expected from the fresh group, deliberately on the low side, and with a median win at day 6 most of them should resolve inside a fortnight. That is a defensible short forecast.

What you cannot do is extend it down the age bands using the win distribution. The fact that 6 of 46 wins landed in days 15 to 30 is a share of wins, not a conversion rate for an open lead of that age, and treating one as the other is the most common way this number produces a fictional forecast. The two are shares of different populations, and the second one is not on the page at all.

Getting the real figure means building a cohort: take leads created in a quarter that closed long enough ago to be fully resolved, find the ones still open at day 20, and compute what share of those eventually won. That is a real rate and most shops have never computed it. It is usually far lower than the win distribution suggests, which is the point.

Two clocks that get confused

The sales cycle runs from lead created to lead won. The estimate-to-job lag, which a sibling card covers, runs from estimate raised to job created. They are measured between different events over different populations and they answer different questions.

A lead that sat 40 days before anyone quoted it and then converted 5 days after the quote went out carries a 45-day sales cycle and a 5-day estimate-to-job lag. Use the first to set a quote validity period and you write a 45-day window on work customers decide in 5. Use the second to set a follow-up horizon and you drop leads at day 5 that had another five weeks in them.

Name which clock a number came from every time you write it down. Both are useful. Neither substitutes for the other.

The age ladder

Build the ladder from the shop's own two timings rather than a borrowed rule, and apply it to the open list before anyone reads a pipeline figure.

The day-60 cut is about three times the 19-day loss mean, rounded up to the nearest band edge in the shop's own table. Rounding up is the conservative direction here: it keeps leads live longer, so the ladder never discards a lead because of a rounding convenience.

Running it once removes 40.5 percent of the open list from the live pipeline. Those leads are not deleted, and the long-cycle list is not a graveyard: a lead parked there with a reason and a date is exactly what a seasonal follow-up campaign is built from. What it is not is a live opportunity, and it should stop being counted as one.

References

• See related: Days From Estimate to Job and What a Long Tail Means - the other clock, and how to read a duration distribution
• See related: Pipeline Value Is Not a Forecast Until You Weight It - where this age ladder gets applied
• See related: Estimate Conversion Rate and the Cohort Problem - how to build the aged cohort the real open-lead rate needs
• See related: Lead Conversion Rate Excludes the Leads You Never Touched - the same funnel measured by count rather than by time
• See related: First Response Time and the Window That Decides the Job - why the 0.4 days at the front of this cycle matters far more than its share of the cycle suggests