Where your sales funnel leaks. Usually inside a stage, not between two of them.

The short answer. A conversion rate is a snapshot, not a clock.

Ask a sales leader where their funnel leaks and most point straight at the stage with the lowest conversion rate on the dashboard. That number is real, but it is also incomplete, because it only counts deals that have already finished moving through a stage, either forward or out. It says nothing about the deals currently stuck inside that stage, quietly not moving in either direction, which is where a meaningful share of lost revenue actually sits.

Ebsta and Pavilion's 2025 GTM Benchmarks report, built in association with Pavilion from an analysis of $48 billion in pipeline data and a survey of around 2,000 revenue leaders, puts a sharp number on this. When a late-stage deal slips beyond two months past its expected close date, the win rate on that deal drops by 113%. Early-stage slippage barely moves the number by comparison. The same report found that low-performing teams' deals are 217% more likely to slip at a late stage than high performers', which means the leak concentrates exactly where it does the most damage, close to the revenue a forecast is already counting.

That is the gap a stage-conversion table cannot show. Two pipelines can report an identical conversion rate out of proposal stage, say 40%, while one moves every deal through in three weeks and the other lets the same deals sit for ten. Both show "40% conversion" on the dashboard. Only one of them is actually leaking revenue, and the conversion rate alone cannot tell you which.

How it works in practice. The metric that catches a leak a conversion rate misses.

Average days in stage, measured per stage and benchmarked against that stage's own typical cycle, is the number that exposes a stalled deal while it is still open, rather than only after it eventually closes or dies. A deal sitting in proposal for sixty days against a typical cycle of eighteen is a visible leak the moment it crosses that threshold, long before it shows up as a lost opportunity in next quarter's conversion report.

Segmenting that number by deal size matters almost as much as tracking it at all. A small, simple deal that takes forty-five days in negotiation is a real problem; a complex, multi-stakeholder enterprise deal taking the same forty-five days might be entirely normal. Benchmarking every deal against one blended "typical cycle" number hides the segment where the leak is actually abnormal inside a number that looks fine on average.

Pairing days-in-stage with a stalled-deal flag closes the loop: any open deal that has sat in its current stage longer than that stage's benchmark, with no logged next step in the last two weeks, gets surfaced automatically in the weekly pipeline review rather than waiting to be noticed by chance. The flag does not say the deal is dead. It says someone needs to look at it now, while there is still time to act, rather than next quarter when it finally gets marked lost.

Setting the benchmark itself takes a bit of care. A new team with no historical close data has nothing reliable to measure "normal" against yet, and a borrowed industry figure, First Page Sage or a similar published range, is a reasonable starting point for the first couple of quarters. Once six months or so of closed-won deals exist in the CRM, the benchmark should switch to the business's own historical median days-in-stage, because a borrowed external figure reflects somebody else's buyers, sales motion and deal sizes, not this pipeline's. A benchmark that never gets replaced with the business's own data stays permanently approximate, which is fine as a starting point and a problem if it is still the only reference a year later.

Where a deal sits matters as much as how long it has sat there. A deal flagged for slow movement in an early stage, say qualification, is worth a note and little more, since Ebsta's data shows early slippage barely moves the eventual win rate. The same number of days flagged at proposal or negotiation deserves immediate attention, precisely because that is where the 113% drop in win rate actually concentrates. Treating every stalled flag with equal urgency wastes attention on the stages where slow movement is survivable and risks under-reacting to the one or two stages where it genuinely is not.

What good looks like. A worked scenario on a healthy-looking funnel that wasn't.

A UK B2B services team I worked with, illustrative rather than a named client, had a stage-conversion table that looked solid across the board: lead to qualified at a healthy rate, qualified to proposal in line with industry ranges, proposal to closed-won sitting right where benchmarks said it should. Revenue had plateaued for two quarters regardless, and nobody on the leadership team could point to a specific cause from the dashboard they were looking at every week.

The conversion rates were not wrong, they were just the wrong question. Once we pulled average days in proposal stage and segmented it by deal size, a clear pattern appeared: mid-size deals, the segment generating the most revenue, were sitting in proposal for an average of fifty-one days against a historical typical cycle of twenty-two. They were not being lost, which is why the conversion rate still looked fine eventually, they were simply rotting in place for an extra month before either closing or quietly dying much later than they should have.

The fix was not a new proposal template or a pricing change. It was a stalled-deal review added to the existing weekly pipeline meeting, specifically flagging any mid-size deal past thirty days in proposal with no logged buyer contact in the previous ten days. Within two quarters, average days in proposal for that segment dropped from fifty-one to twenty-eight, and closed revenue from that segment rose even though the underlying conversion rate barely moved, because deals that would previously have died slowly, after consuming a rep's attention and a forecast's optimism for months, now either closed or got marked lost within weeks, freeing that attention for deals that could actually close.

The decision rule that came out of it: a deal review should ask how long a deal has been in its current stage before it asks what its probability of closing is. A probability estimate on a deal that has silently stalled for two months is just a guess dressed up as a number, and the time question is the one that catches the problem before the probability question has to.

Pitfalls to avoid. Where sales funnel leak diagnosis goes wrong.

The first pitfall is diagnosing the leak purely from conversion rates, which, as above, show what eventually happened to deals that have already finished moving, not what is currently stuck. A funnel can report healthy conversion at every stage while actively leaking revenue through deals that simply haven't finished stalling yet.

The second is benchmarking days-in-stage against one blended average across every deal size and complexity. That average flattens exactly the variation that matters, making a genuinely stuck mid-size deal look unremarkable next to a complex enterprise deal that is behaving entirely normally for its segment.

The third is treating a stalled-deal flag as a reason to immediately mark the deal lost. The flag is a prompt to check in and find out what is actually happening, a changed decision-maker, a delayed budget cycle, a competing priority, not a verdict. Marking every flagged deal lost automatically just moves the inaccuracy from the forecast to the loss report instead of fixing it.

The fourth is only reviewing days-in-stage at the monthly or quarterly forecast meeting. By the time a stalled deal surfaces in a quarterly review, it has often been rotting for most of the quarter already. The stalled-deal check belongs in the weekly pipeline review, where there is still enough runway left in the stage's typical cycle to actually intervene.

The fifth, specific to UK B2B teams selling into longer procurement cycles, is confusing a genuinely long, normal late-stage legal or procurement review with a stalled deal. Not every deal sitting still for sixty days is leaking: some are working through a buyer's own internal process on a timeline the seller doesn't control. The distinction is whether there is a known next step and a date attached to it. A deal with a confirmed procurement review due back on a specific date is progressing on schedule. A deal with no logged activity and no known reason for the silence is the one actually leaking.

The sixth is fixing the stage once and assuming the leak stays fixed. A proposal stage that tightens from fifty-one days to twenty-eight after a review process goes in can drift back out over a few quarters as reps get busy, new hires skip the new habit nobody explained to them properly, and the exceptions that used to get caught in the weekly review start slipping through again. Days-in-stage needs to stay on the dashboard permanently, not just for the quarter immediately after a fix, because the same quiet drift that created the original leak is always available to create it again.