Pipeline Metrics and Coverage

Five definitions

Win rate. Deals won divided by deals closed (won plus lost) in a period. Exclude deals still open. Decide once whether disqualified deals count as losses and keep the rule, because changing it silently changes the number. Matthew Dixon's 2022 analysis of about 2.5 million sales conversations put the average win rate at 26%, with high performers near 60%; your own number is the only one that matters, but that range is a useful sanity check.

Stage conversion. For each stage, the share of deals that entered it and later reached the next stage. Lesson 1.2 used it to detect inflated stages; here it tells you where deals leak.

Cycle time. Median days from a deal's creation, or from its entry to your first qualified stage, to its close, computed separately for won and lost deals. Lost deals that take as long as won ones usually mean the team is slow to disqualify.

Coverage. Open pipeline value expected to close in a period divided by the target for that period.

Velocity. Revenue the pipeline produces per unit of time, defined in the last section.

Coverage: derive it, do not borrow it

Sales teams often say they need three times their target in pipeline. That rule of thumb has no single traceable source, and the right multiple depends entirely on your win rate. If you win one deal in four (a 25% win rate on closed deals) and your deals are similar in size, you need roughly four times your target in qualified pipeline that will close in the period, before allowing for slippage. If you win one in two, two times is enough. Coverage of three times with a 20% win rate is a forecast miss waiting to happen.

Two refinements make coverage honest. Count only pipeline whose expected close date falls in the period, and only deals past your qualification stage; early-stage deals have lower win rates and should be weighted or counted separately. And watch coverage for the following period, not just the current one. By the time this quarter's coverage is low, it is usually too late to create deals that can close in it. Next quarter's coverage is where you still have time to act.

Velocity and Little's Law

A common way to combine the metrics is pipeline velocity: the number of qualified opportunities, times the win rate, times the average deal size, divided by the cycle length. It is arithmetic rather than research: the expected value of the deals in flight, spread over the time they take to close. Its usefulness is that it shows all four levers at once. A 10% improvement in any one of them raises velocity by about the same amount, so the question becomes which lever is cheapest to move for your team.

Little's Law, from John Little's 1961 proof and his 2011 review, adds a constraint: the number of deals in the pipeline equals the rate at which you create them times the average time they spend in it. You cannot carry more deals without either creating them faster or keeping them longer, and keeping them longer lowers velocity. Teams that add deals to look healthy usually just lengthen their cycle.

The human side caps all of this. Salesforce's 2026 State of Sales survey (vendor data) found the average seller spends only 40% of their time selling. Moving that number is often the cheapest lever of all, and it is the one AI tools most plausibly help with, provided the time saved goes into buyer conversations rather than more volume.

Where the numbers come from in Kavanah

get_pipeline_summary gives, per stage, deal count, total amount and revenue, probability-weighted versions of both, the open ('capturable') pipeline and won totals. That is the raw material for coverage: filter to deals with an expected close date in the period and past your qualification stage.

Win rate, stage conversion and cycle time come from deal histories. Kavanah records stage changes on each deal's activity timeline, but it does not compute these rates for you; ask the agent to list deals closed in the period with their stage, creation and close dates, and compute the rates in a sheet, or have the agent do the arithmetic and show its working. Do it the same way every period.

Two built-in signals help with the leading edge. list_stale_deals finds deals past their stage's silence window, which is where cycle time quietly grows. And for network cadences, the contact runway check (get_contact_runway, with a threshold you set) warns when a campaign is about to run out of people to start conversations with, which is the earliest possible warning that next quarter's coverage will be thin.

Put the five numbers on one page, reviewed weekly, with the trend over the last six periods. A metric reviewed without its trend invites overreaction to noise.