EN "Proposal sent," "demo delivered," "followed up" — these stage names share one flaw: the action is entirely on your side, so you can advance the deal whenever you feel like it. Anything you can do unilaterally is equally likely to appear in a deal that closes and a deal that dies, which means it carries no information about the outcome. Only one kind of criterion moves a deal forward: an action the buyer paid for — with time, with disclosed information, or by pulling a third person in. So audit your stage names one at a time and ask whether the step is something you complete or something they must do, then rewrite it: "proposal sent" becomes "they booked the next meeting and named who attends." When you're unsure, use the fallback question — if I do nothing at all after this step, will they still come to me? If the answer is no, the deal has not moved, however good the meeting felt.
EN "We closed twenty deals this month and won twelve, so our win rate is sixty percent" is almost always false, and the fault is in the denominator: only deals that closed get counted, while the largest single category of outcome never closes at all. Those deals stall, drag, and quietly age, because nobody enjoys clicking "lost." Take one month's intake instead and follow that whole cohort to its final destination — a hundred leads that produced twelve wins, twenty-eight losses and sixty deals still sitting there gives a real win rate of twelve percent, not sixty. Four numbers only mean anything computed this way: cohort win rate, stage-to-stage conversion, cycle length calculated separately for wins and losses, and the share that ended in no decision. That last figure usually shocks people the first time, and it is where your improvement actually lives.
EN Multiply each deal by a probability, add them up, and you get something that looks like an answer. The problem isn't that it's imprecise — it's that the number corresponds to no possible reality. Deals are binary: the money arrives in full or not at all. An expected value only converges over many repetitions, and eight deals in a quarter is a high-variance gamble rather than a forecast. With eight coin-flip deals, "four wins" is the single most likely outcome and still happens only twenty-seven percent of the time; roughly three quarters in ten land outside the three-to-five band, which is when you find yourself explaining a number that makes no sense. Worse, you wrote those percentages yourself, and self-reported confidence runs systematically high. Replace the single number with three tiers — commit, best case, pipeline — putting a deal in commit only when the buyer has done something costly. Then write your forecasts down and score them at quarter end. Calibration comes from feedback, not from effort.
EN "We have 2.6 million in pipeline" carries no information, because 2.6 million can equal zero just as easily as 26 million can. Queueing theory settles it: in steady state, work in progress equals arrival rate times average time in system. So a growing pipeline has exactly two possible sources — more coming in, or things moving slower — and the second is far more common. That is congestion, not strength. Three pipelines with identical totals can need completely unrelated fixes: too many leads that never convert means the problem is qualification, not lead volume, while a fat middle means deals are sitting sixty days after the proposal because nobody with a budget is in the room. So run four questions weekly: which stage leaks hardest, which stage sits longest, how much of the pipeline is aged past twice its stage average, and whether coverage clears target divided by your real cohort win rate. Total value is the output of those numbers, not a metric of its own.