Weighted pipeline: what the number tells you, and what it hides

Weighted pipeline is deal value multiplied by probability, summed across your open deals. That is the whole formula. It is an average, and an average is a statement about many comparable things. A sales pipeline is usually neither many nor comparable, which is where the trouble starts.

None of what follows is an argument that the number is wrong. It is arithmetic; it computes what it says it computes. The argument is that it answers a different question from the one you are asking it, and that the gap between those two questions is where quarters are missed.

What the number actually is

Two inputs go in. The value of the deal, which you probably know. And a probability, which in most CRMs is attached to the stage rather than to the deal: every opportunity in Negotiation is 70%, every one in Proposal is 40%. That figure is either the historical close rate of deals that have sat in that stage, or — more often than anyone admits — a number an administrator typed in when the system was configured and nobody has revisited since.

Multiply, sum, and you have a forecast. There is no model underneath it. Nothing in it knows that this buyer has gone quiet, that the champion left, or that the last four deals of this shape took two quarters instead of one.

An average of what, exactly

Averages earn their keep when you have many things and each is small relative to the whole. Insurers price policies this way and it works, because they write millions of them and no single claim moves the book. The average is reliable precisely because nothing in it matters individually.

Now take an ordinary quarter: eleven open deals, one of them worth more than the three smallest put together. Here is such a pipeline, run through ten thousand simulated quarters.

The weighted number says 84% of target, which in most forecast calls is a sentence that ends the conversation. The simulation says the quarter lands anywhere between 305.000 € and 885.000 €, and that the actual odds of reaching the target are 41%.

Both numbers are correct. They are answers to different questions. One is where the middle of the distribution sits. The other is how wide the distribution is, which is the part that decides whether you make the number.

Three things it cannot tell you

Too few deals to average. With eleven opportunities, the law of large numbers has nothing to work with. The spread above is not noise around a true value; it is the honest shape of eleven coin flips of different sizes.

Sizes that are not comparable. Ten small deals and one large one are not eleven deals. The uncertainty a deal contributes goes as p × (1 − p) × value², so doubling the size of a deal quadruples the doubt it carries. In this pipeline, 92% of the total uncertainty comes from just three deals:

• Blackwell expansion — 79% of the spread
• Garrow new logo — 9% of the spread
• Everley rollout — 4% of the spread

Weighting hides that completely. Those three are the only deals in the quarter whose movement changes the answer, and the weighted total gives you no way to find them.

No way to state the risk. You can give your CEO a number. You cannot give them your confidence in it, or the bad case, or the probability of the bad case, because a single figure has nowhere to put any of that. So the risk gets carried in your tone of voice, and tone of voice is not something anyone can plan against.

Why it survives anyway

Because it is one number, and a board meeting wants one number. Because every CRM computes it by default. And because when it is wrong, it is wrong in a way nobody measures: almost no company writes down what it forecast at the start of a month and compares it to what closed, so the error never gets a name.

That last one is the most fixable thing on this list and the least often done. Twelve rows in a spreadsheet — what you said, what happened — will tell you more about your forecasting than any new report. When I did it on my own numbers I found I was high by a similar margin nearly every month, which is a bias rather than bad luck, and you correct a bias by subtracting it.

What to do instead

Keep the weighted number if your board wants it. It is a reasonable estimate of the middle. Just stop asking it questions it cannot answer, and add two things beside it.

• A range, and the odds against your target. Not because the range is more accurate, but because it is honest about what is not known, and because a quarter at 41% is a different conversation from a quarter at 84% of coverage.
• The handful of deals driving the spread. Value times uncertainty, biggest first. In most quarters it is three or four names, and they are what the review should be about.

Neither needs software. A spreadsheet with deal, value, your own odds and a column for what actually happened will get you most of the way, and the record it builds is worth more than the maths.