How is average sales cycle length calculated?

Total days from a defined start event to a defined end event, summed across a set of closed deals, divided by the number of deals.

Why is average sales cycle length misleading?

It blends distinct deal populations — fast deals with a champion and budget already in place, slow deals going through a full committee — into one number that may not represent any actual deal.

Does it matter where you start and stop measuring?

Yes. In this article's worked example, measuring from opportunity-created versus first touch on the identical deal produced a 27% difference in the reported cycle length.

What should teams track instead of one average?

Cycle length segmented by deal type, and stage-level duration trends so a change in the total can be traced to a specific stage. The median is worth reporting alongside the average, but it only corrects for skew and outliers — where deals split into two genuine populations, the median lands in the same misleading gap the average does.

How do you find out which stage is causing a longer cycle?

Break the average into stage-level durations for the current and prior period and compare them. In the article's worked example, an entire 15-day increase in the average was concentrated in the negotiation stage.

Why your average sales cycle length is hiding the real problem

Guide · sales · 5 min read · last verified 2026-07-21

Reviewed before publication Editorial board Independent commercial review

In shortA single average sales cycle number blends fast and slow deals into a figure that describes neither, and hides which stage is actually driving the change.

What the average actually computes

Average sales cycle length is total days-to-close summed across a set of deals, divided by the number of deals. It's easy to calculate, easy to put on a dashboard, and it describes almost no individual deal in the set.

Worked example: a 60-day average that describes zero real deals

Take 10 closed-won deals. Five closed in 20 days — fast-moving deals with an internal champion and a pre-existing budget line. Five closed in 100 days — deals that went through a full committee evaluation and procurement review.

Average = (5 × 20 + 5 × 100) / 10 = (100 + 500) / 10 = 60 days.

Not one of the ten deals took 60 days. The average sits in a gap between two real populations, and if you use it to predict when a deal currently in pipeline will close, you'll be wrong for both groups — you'll chase the fast deals too slowly and the slow deals too quickly.

And the median doesn't rescue it. The usual reflex when an average misbehaves is to reach for the median instead. Do it here. Sort the ten deals: 20, 20, 20, 20, 20, 100, 100, 100, 100, 100. With an even count of 10, the median is the mean of the 5th and 6th values: (20 + 100) / 2 = 60 days. Identical to the average we just discredited, and still describing none of the ten deals. Switching to the median here buys you nothing, because the median fixes a different problem than the one on the table.

Two failure modes get run together constantly, and they have different fixes:

Skew and outliers. Most deals cluster in one place and a few extreme values drag the mean away from that cluster. The median fixes this, because it cares about the rank of the extreme values, not their magnitude — one 400-day deal moves the mean a long way and the median barely at all.
Bimodality. The deals sit in two genuine clusters with a gap between them. Neither the mean nor the median fixes this, because there is no single center to find — any one number you compute has to land somewhere, and the only place left is the empty gap between the two populations. The fixes are segmentation (split the populations and report each) or a histogram (show the distribution and let the two humps be visible).

The 10-deal set above is the second kind: balanced, no outliers, two real populations. That is precisely why the median lands on 60 as well. If you take one thing from this piece, take that: "use the median instead" is good advice for a long tail and useless advice for a split book.

This is a bimodal distribution problem, and B2B sales cycles are bimodal more often than the single-number dashboard suggests: new-logo deals versus expansion deals, deals with a pre-existing champion versus cold-sourced deals, deals under $25,000 that skip procurement versus deals over $100,000 that don't. Blend any of those populations into one average and the number stops describing anything.

The definition problem: cycle length measured from what, to what

Before the bimodal issue even comes up, teams often aren't measuring the same thing. "Sales cycle length" needs a start event and an end event, and both are contested:

Start: first outbound touch? First reply? MQL? Opportunity created in the CRM?
End: contract signed? First invoice? Closed-won stamped in the CRM (which can lag the signature by days)?

Worked example: a deal gets first contacted on day 0, becomes an MQL on day 5, gets logged as an Opportunity on day 20, and closes on day 95.

Measured from opportunity-created to close: 95 − 20 = 75 days.
Measured from first touch to close: 95 − 0 = 95 days.

That's a 20-day difference, or 26.7% relative to the shorter measurement (20 / 75). Two teams can report "our sales cycle" for the identical deal and land 27% apart, with neither number being wrong — they're just answering different questions. If the definition isn't written down and applied consistently, quarter-over-quarter comparisons of "cycle length" are comparing apples measured with different rulers.

Where the real signal lives: stage-level duration, not the average

The headline number also hides where time is actually being spent — which is the part that's actionable. A single average going from 60 to 75 days tells you sales cycles got longer. It doesn't tell you where, and "sales cycles are getting longer" produces vague responses: more urgency training, more end-of-quarter discounting, generic pressure on reps to move faster. None of that fixes a problem you haven't located.

A second worked example: the 15-day increase that was entirely a negotiation problem

Break the same headline number into stage durations.

Q1 average cycle = 60 days: Discovery 10 + Demo 10 + Proposal 15 + Negotiation 15 + Contracting 10 = 60.

Q2 average cycle = 75 days: Discovery 10 + Demo 10 + Proposal 15 + Negotiation 30 + Contracting 10 = 75.

Every stage held steady except Negotiation, which doubled from 15 to 30 days — accounting for the entire 15-day increase in the headline average. That's not a "reps are slower" problem or a "buyers are less interested" problem. It's a negotiation-stage problem specifically, and it points at a short list of likely causes: pricing approval taking longer to route internally, legal turnaround on redlines, or a new procurement step that wasn't there last quarter. You can go find out which. You can't go find out what's wrong with "sales cycles," because that phrase doesn't point anywhere.

What to track instead of one number

Cycle length by segment (new logo vs. expansion, deal size band, source) instead of one blended figure.
Median alongside the average — but know its scope. The median is the right tool for skew and outliers, where a few very long or very short deals drag the mean off the cluster. It is not a fix for bimodality: as the worked example above shows, a set split evenly between 20-day and 100-day deals has a median of 60, the same as its mean. Report it, but don't expect it to find a center that isn't there.
Stage-level duration trends, quarter over quarter, so a change in the total can be traced to a specific stage.
A written, fixed definition of start and end events, applied the same way every quarter so the number is comparable over time.
Distribution, not just central tendency — a histogram of cycle lengths will show you the bimodal split an average erases.

The takeaway

A single average sales cycle number is easy to report and hard to act on. It blends populations that behave differently, depends on a start/end definition that's rarely made explicit, and hides the stage-level detail that would actually tell you what to fix. If the number moves, the useful question isn't "why did our sales cycle change" — it's "which stage did," and that question only has an answer if you're tracking stages, not just the total.