Forecasting bottom-up: key accounts, renewals, new business

A sales forecast says how much the team will sign in a period. The common shortcut is a weighted pipeline: each deal's value times its probability, summed. Leo's sheet does exactly that and arrives at over ten million for the quarter.

The arithmetic is fine. The inputs aren't: each probability is the owner's feeling. A seller's gut number mixes hope, effort and memory, and one optimistic owner can double the total on their own.

Weighting works when the probability is earned: taken from how often deals that reached the same stage were actually won. So Teams builds the forecast bottom-up, in three parts: key accounts judged one by one, the base of renewals and expansions, and the many small new deals weighted by stage history. Each part gets the method that suits it.

Big deals distort a forecast the most: one key account in the wrong quarter can be most of the error. So they don't get a percentage. Each gets a judgement with evidence: can this deal sign this quarter, given the buyer's own calendar?

Teams' three key accounts show how. Veloria wants a paid pilot first and a start early next year. Cobalt is at the proposal stage, but its contract with its current school runs until March, and a switching buyer usually can't sign until that contract allows it. Orchid is still in discovery. Whatever the sheet says, none of them signs before January.

Moving a key account isn't marking it lost and isn't pessimism. The deal is alive and keeps being worked; it goes into the quarter it can really sign, with a written reason next to it, so anyone can see it's the buyer's calendar.

For the many small new deals, a percentage helps, as long as it comes from history: of the deals that reached a stage, how many were won? That is the stage conversion.

How you combine past quarters matters. At the proposal stage Teams won 2 of 3, then 1 of 5, 1 of 1 and 1 of 4. The average of the rates, (66.67 + 20 + 100 + 25) / 4 = 52.92 %, lets a quarter with one deal count as much as a quarter with five. The pooled rate adds the deals first: 5 won of 13 reached, 38.46 %.

Pooling weights each quarter by how many deals it actually had, which is what you want when you apply the rate to many deals. Here the difference is fourteen points; on a few million of proposals, that's real money.

With pooled stage rates, each live new deal contributes to the expected forecast in one step: its value times the pooled conversion of the stage it is in.

A live proposal-stage deal worth 960,000 UAH adds 960,000 × 38.46 % = 369,216 UAH. It won't sign for 369,216, of course: it will be won in full or lost. But across many deals at the same stage, roughly 38 in every 100 have closed, so the weighted values add up to a fair estimate of the whole group.

That is also why the method suits the many small deals and not the key accounts. Applied to one huge deal, a percentage gives a number that will certainly be wrong; applied to twenty small ones, the errors cancel out. And the rate applies only to deals that are actually alive.

Pooled rates come from deals that kept moving: they reached a stage and then were won or lost. A deal that has gone silent isn't part of that pattern any more, so giving it the full stage rate overstates the forecast. A stuck deal at full stage odds is the oldest lie in a pipeline.

So before weighting, filter. A deal flagged under Teams' rule (more than 30 days without buyer activity, or a close date moved twice) adds nothing to expected this week. It goes to the honest question first: «Is this still a priority this quarter?» If the buyer answers and the deal moves, it comes back with its stage rate; if not, it goes to lost with a reason.

The same filter keeps stuck deals out of commit and best case.

The second part of the forecast is the base: renewals and expansions of current clients. It is tempting to count it in full because these clients already know the product. Instead, judge each renewal like a key account: is the contract moving, and is the client responding?

The Lviv design studio renews in December, and the contract is already with their accountant: that renewal is base. Their wish to add a support team is new scope, so it is still a proposal and is forecast like one. A renewal whose client hasn't replied in five weeks isn't base at all; it's a churn risk, and the right move is a call this week.

Watch slippage too: a close date that keeps moving is a deal telling you it isn't ready. Teams flags a deal after two moves.

One forecast number hides the range. Teams reports three, each built by a rule:

• Commit: what you would bet on. Live renewals ready to sign plus new deals at the contract stage with recent buyer action, at full value.
• Best case: commit plus live proposal-stage deals that could close if things go well, also at full value.
• Expected: the base plus every other live deal times its pooled stage conversion.

Commit and best case are categories counted at full value; expected is a weighted number. Together they tell Leo what is nearly certain, what is possible and what is likely.

The owners' gut forecast, every deal times its owner's probability with stuck deals included, isn't a fourth number. It is what the three replace.

Before the forecast goes to Leo, check that the three numbers make sense together. Commit is the floor. Best case adds every live proposal at full value. Expected weights the live deals by history, so it should normally land between commit and best case.

If expected comes out above best case, something went in wrong: usually a stuck deal kept its stage odds, a key account was weighted instead of judged, or a deal sits in two parts at once. Fix the input, not the total.

Then explain the drop from the sheet deal by deal: which key accounts moved to which quarter and why, which deals were flagged, and why the rates are pooled. A forecast that is honest in October saves an awkward conversation in December.

How it fits together: stages, qualification and the bottom-up forecast are now all in place. The case lesson applies them for real: thirty open deals, the stage history and Leo's ten million.