How to Run a Forecast Retrospective After a Missed Quarter

What is a forecast retrospective?

A forecast retrospective is a structured review of why the called number differed from the closed number, run on the forecasting process rather than on the people. The output is a change to how you forecast, not a list of deals that went sideways.

Most teams skip it. The quarter closes, leadership issues a summary, and the next quarter starts with the same model and the same assumptions. The miss gets attributed to two lost deals and the process defect that produced it survives intact.

Run the session once per quarter regardless of outcome. A quarter you beat by 15 percent contains as much information as a quarter you missed by 15 percent, because both mean the forecast failed to describe what happened.

What data do you pull before the meeting?

Decompose the variance into sources before anyone talks, because the conversation follows whichever number appears first.

Build the pre-read with three parts.

First, the variance decomposition. Split the quarter's revenue into the paths it could have arrived by:

Second, the metric movement. Win rate, average closed-won deal size, cycle length from qualified to closed, and slip count, each compared against the trailing four quarters.

Third, called versus closed by segment and by manager across four quarters. One quarter of variance is an anecdote. Four quarters of variance in the same direction is a design flaw.

How do you separate a model problem from an execution problem?

Check whether the inputs moved. If deal size, win rate, or cycle length shifted and the forecast did not, the model failed. If none of them moved and you still missed, execution failed.

Forecasts miss most often because something in the business or the market changed while the model kept running on old assumptions. Four patterns cover most of it:

- A new competitor enters and creates pricing pressure. Average deal size falls while volume holds. - Interest rates rise. Private equity slows capital deployment, valuations compress, buyers cut cost to protect earnings, and win rates fall. - Macro uncertainty rises. Buyers make fewer decisions and cycles stretch from qualified to closed. - You change territories. Reps are distracted, pipeline still looks healthy, and execution drops underneath it.

Each of those produces the same headline result: pipeline stagnates, deals close for less money, and win rates decline. The retrospective's job is to name which one you were in. A model that cannot pick up those changes inside a quarter will keep producing a confident number that describes a business you no longer have.

What do you do about the claim that the data is bad?

Reject it as the primary explanation. Every revenue team believes their CRM data is uniquely bad and that this is why they cannot forecast. It is a near-universal belief and it is wrong.

Garbage in does not have to mean garbage out. As long as data entry is consistent, a model learns the pattern in how your team records deals and predicts accurately through it. A team that always logs close dates two weeks early is highly predictable. A team that logs them accurately half the time and optimistically the other half is not.

So change the question in the retrospective. Instead of asking whether the data is clean, ask whether it is consistent. Where two managers apply different stage definitions, where one segment uses close date as a review date, where amount is updated at close rather than at proposal, you have a consistency defect worth fixing. General complaints about data quality produce a cleanup project that never ends and never improves forecast accuracy.

What questions should the meeting actually ask?

Six, in this order, with the pre-read already circulated.

1. Which of the three revenue sources missed, and by how much? 2. Did the model's day-one call for that source differ from the rep-submitted call? 3. Which input metric moved first, and in which week did it move? 4. When did we have enough signal to act, and what did we do that week? 5. Which deals slipped, and had the close date already moved once before? 6. What would have had to be true in week three for us to have made the number?

Question four is the one that changes behavior. Most teams discover the miss in the final two weeks, which is too late to matter. Getting the forecast right in the last week of the quarter helps no one, because by then the quarter has already happened. The value sits in knowing the shape of the quarter on day one, early enough to build pipeline or reallocate capacity.

Question five links the retrospective to your best early warning. A close date change is the strongest slippage signal in the CRM, and a deal that slips from one quarter to the next is less likely to close even when it stays in commit. Count how many of the missed deals had already moved their date once, and you usually find the miss was visible weeks before it was acknowledged.

What should the retrospective produce?

One change to the forecasting process, one change to the data standard, and a published variance number that carries into the next quarter.

Cap it at two changes. Retrospectives that generate a ten-item improvement list produce no improvement, because nothing gets owned.

Set the accuracy bar explicitly while you are there. Manually built forecasts on new and expansion business typically land near 90 percent accuracy, and holding that number costs a large block of analyst time each cycle because the model is rebuilt rather than updated. A model trained on your own historical performance can hold a tighter band without manual adjustment and stay stable from day one through day ninety of the quarter. Decide which of those two you are running, then hold the next retrospective to the standard you picked. For the mechanics underneath the number, start with how to create a sales forecast and the definition of win rate you are measuring against.