Pipeline Value Is Not a Forecast Until You Weight It

Why this matters

A pipeline figure gets quoted to bankers, used to justify a hire, and used to decide whether a slow month is a problem. It is almost always the plainest possible sum: add up the estimated value of every lead that is neither won nor lost. That number is not a forecast, it is an inventory of hope, and on a real shop's records it can run several times the work that will actually be booked. Hiring against it is how a shop ends up carrying a crew through a quarter that never arrives.

What it is, and the three things it is not

The construction: sum of the estimated values across every lead not yet marked won or lost.

• It is not aged. A lead nobody has touched since last spring counts at full value.
• It is not weighted. A lead at first contact counts the same as one with a signature pending.
• It is not revisited. The estimated value is a guess somebody typed when the lead was created, usually before anyone saw the site.

And underneath all three, it is not period-scoped. A pipeline is a stock, not a flow. It has no window, so comparing it against a quarterly booking target is comparing a level against a rate, and the two are not the same kind of number. A sibling card covers that mismatch in general terms; here it is the fourth correction.

The pipeline as it reads

A shop's open list on the day the figure is pulled. Index the shop's typical service ticket at 1.0 unit and read everything in those units.

Two hundred and fifty units against a shop whose recent quarters book somewhere in the sixties. The owner reads that as four quarters of work in hand.

Correction one: age it

Apply the age ladder from the shop's own win history. Here that ladder closes anything past day 60, because the shop's wins have effectively run out by then.

Thirty-four of the 78 leads, 43.6 percent of the open count, are past the age at which this shop wins anything, and they carry 78.0 of the 250.0 units, 31.2 percent of the headline value.

Notice the shape of that table, because it is the normal shape and it is worth expecting: stale leads concentrate in the early stages. Twenty-four of the 34 are sitting in New, untouched, which is what an untouched lead does, and none are in Verbal yes, because a lead with a signature pending gets chased. The ageing correction therefore lands hardest on exactly the stage whose leads are least likely to close, stripping 33.0 of the 78.0 stale units out of New alone. Across all three corrections the largest single reduction is Quoted's, at 67.5 units, because that is where the live value is.

Correction two: weight it

A lead with a verbal yes and a lead nobody has called are not the same asset. Weight each stage by the share of leads reaching that stage that this shop's own history says eventually win.

The entrant column is the count each weight was derived from, and it deliberately does not total, because a lead that reached Quoted entered New as well. A weight printed without that count beside it is a number nobody can check, and the floor those counts have to clear is set at the end of this card.

Weighting removes another 100.28 units, which is 40.1 percent of the original headline, and it does more work than ageing did. That surprises most owners, who expect the dead wood to be the story.

Correction three: reality-check the entered value

The last correction is the smallest and the one owners argue about longest.

On this shop's records, the value entered at lead creation runs about 1.25 times what the eventual quote comes in at. That is not dishonesty, it is the arithmetic of optimism at intake: a customer describing a problem over the phone describes the largest version of it, and whoever types the lead types that. Once a quote exists the number is real, so only the two pre-quote stages need correcting.

• New: 2.28 divided by 1.25 is 1.824
• Contacted: 8.14 divided by 1.25 is 6.512
• Quoted and Verbal yes are unchanged at 28.50 and 32.80

Corrected total: 1.824 plus 6.512 plus 28.50 plus 32.80, or 69.64 units.

Against the 250.0 headline, that is a factor of 3.6.

And the decomposition, which is the part to carry away:

The order matters and the table has to say so. Running the value correction last charges it only against what survived the first two, so its 0.8 percent is the marginal contribution in this sequence, not its size in isolation. Correct the entered values first, on the raw 113.0 units sitting in New and Contacted, and the same adjustment removes 22.6 units, 9.0 percent of the headline, before anything else touches them. The reason to decompose at all is that the ranking of causes is not obvious in advance, and here the correction the shop had been arguing about is the one that matters least in any ordering.

The fourth problem: a pipeline has no period

Even 69.64 units is not a quarterly forecast. It is what the open list is eventually worth, and eventually is doing real work in that sentence.

To reach a period figure you need each stage's timing as well as its odds. On this shop's history, Quoted and Verbal yes resolve in weeks while New and Contacted mostly do not, so the portion reasonably expected inside the current quarter is 28.50 plus 32.80, or 61.30 units, and the remaining 8.34 belongs to later quarters.

So the headline of 250.0 is 4.1 times the in-quarter figure of 61.30, and the in-quarter figure is 24.5 percent of the headline. That is the comparison to make before anyone hires against a pipeline number, because 61.30 against a shop that books in the sixties is an ordinary quarter, not four quarters of work in hand.

The one thing the figure is genuinely good at

None of this means the pipeline number is useless. It means the level is unreliable and the movement is not.

The corrected 69.64 units says little on its own, because nobody knows what the right level is for this shop. The same figure computed the same way a month from now says a great deal: if it falls while bookings hold, the shop is consuming pipeline faster than it is replacing it, and the quarter after next is the one in trouble. That is a real early warning and the raw headline cannot give it, because the raw headline rises whenever nobody cleans up, which is exactly when things are going wrong.

The trap sits in the words "computed the same way". If you tighten the age cut, or re-derive the stage weights, or start correcting entered values, the new figure is not comparable with the old one. Either recompute the prior period on the new rules before you compare, or say plainly in the same sentence that the comparator is on the old basis. A corrected figure benchmarked against an uncorrected one overstates the change by the size of the correction, which here would be most of the movement.

Keep a note of the basis beside each month's number: the age cut in days, the date the weights were derived, and the window they came from. It is one line and it is what makes a year of these figures a series rather than a pile.

Where your stage weights come from

Borrowed weights are worse than no weights, because they carry somebody else's mix and somebody else's stage definitions while looking rigorous. Derive your own.

Use entrants, not residents. Take every lead that entered a stage in a past window, and compute the share of those that eventually won. Do not compute it over the leads currently sitting in that stage: those are the ones that have not resolved, which is a different and much worse population. A lead that entered Quoted, moved to Verbal yes and won is a Quoted entrant that won, and it belongs in the numerator of the Quoted weight.

Use a window that has closed and aged. Compute it on entrants from a window that closed at least one full decision cycle ago, for the reason a sibling card sets out at length: a window that is still open holds leads whose outcome is still ahead of them, and counting those as losses drags every weight down. Weights derived from an open window are systematically too low, so the forecast built on them is systematically pessimistic, which is a different failure from the one this card started with and just as expensive.

Clear a count floor per stage. At least 30 entrants in the aged window before a stage weight is used as a number, and roll windows together until you get there. Then print the entrant count beside the weight, which is what the third column of the weighting table is for: the 5 leads sitting in Verbal yes today are not the evidence behind its 80 percent, the 46 entrants are. Those counts are why that window was rolled out to four quarters, because Verbal yes was the only weight that came anywhere near failing the floor. Re-derive it first whenever the window moves, since it is carrying the most on the least - its 32.80 is 53.5 percent of the 61.30 in-quarter figure, on fewer records than any other weight in the table.

Then check the weights are doing any work. Two degenerate cases make this concrete. If all your stage weights come out near 100 percent, you are reporting the raw pipeline with extra steps. If they all come out roughly equal, the weighted figure is a fixed multiple of the raw figure and carries exactly the same information, which means your stages are not distinguishing between leads and you have stage names rather than stages. Weights that spread the way the ones above do, from 12 percent to 80 percent, are the sign that moving a lead between stages means something.

Finally, run the age sweep before the pipeline figure is read, not after. Monthly is enough for most shops. A stale lead that survives into the reported number has to be argued out of it in a meeting, and it never is.

Write the figure with all three labels attached or do not write it: aged as of a date, weighted on a named trailing window, and with the pre-quote stages flagged as carrying entered estimates rather than real quotes.

References

• See related: Average Sales Cycle Describes Winners Only - the age ladder applied here, and where its day cut comes from
• See related: Estimate Conversion Rate and the Cohort Problem - why stage weights must be computed on an aged, closed window
• See related: Lead Conversion Rate Excludes the Leads You Never Touched - the untouched leads that dominate the stale column
• See related: Average Estimate Value Counts Drafts and Conversion Does Not - what an entered value is worth before a quote exists