How to Check Why Your AI Sales Forecast Changed Week Over Week
Your commit dropped $1.2M since last Monday, and the first question on the forecast call is the one nobody can answer.
The Monday call where the number moved
Picture a typical forecast call. Last week commit was $8.4M. This week it reads $7.2M. The CRO asks what happened, and the room starts reconstructing it from memory: one rep thinks a deal pushed, a manager remembers a pricing conversation, RevOps pulls up a snapshot report that shows the total but not the reasons.
Forty minutes later the team has a story. Nobody is sure it's the right one. The forecast changed, and the change itself was never recorded anywhere a person could inspect it.
The snapshot report stores each week's total. It doesn't store why the total changed, so every Monday the team rebuilds the reasons by hand, and a deal that quietly lost its economic buyer looks the same as one that slipped a week for a legal review.
📊 67% of sales operations leaders agree that creating accurate forecasts is harder today than it was three years ago. — Gartner |
What can move a forecast between two Mondays
Every change in the total traces back to a short list of causes. Some of them are evidence. Some of them are someone changing their mind. You need to know which is which before you can decide what to do about it.
What changed | How it shows up | Evidence or opinion? |
|---|---|---|
Close date pushed | Deal leaves this quarter's commit | Depends on whether the buyer said it |
Amount changed | Same deal, smaller number | Evidence if scope changed on a call; opinion if the rep hedged |
Stage moved back | Probability drops | Usually evidence, often a missed exit criterion |
Forecast category changed | Commit to best case, or the reverse | Opinion, until something in the deal backs it |
New deal entered or one was lost | Total jumps or falls | Evidence, if a buyer conversation confirms it |
Run your own last two weeks through this table. The rows where you can't name the source are the ones the forecast call should spend its time on.
Separate the evidence that moved from the opinion that moved
A forecast is a stack of judgments. When a rep moves a deal from commit to best case, that's a call they made, and it may be the right one. The problem is that the CRM stores it in the same field, with the same weight, as a close date the economic buyer confirmed on a recorded call.
So when the number moves, ask one question of each changed deal: did something happen in the deal, or did someone's confidence change? A pushed close date with a buyer email behind it is a fact. A pushed close date with no new activity in three weeks is a feeling, and it tells you more about the deal than the date does.
📊 Only 4% of deals close on the date the forecast predicted. — Revenue Operations Alliance (ROA), September 2026 |
When the AI changes its mind and the deal didn't
AI forecasting adds a cause the table above doesn't list: the model itself. A forecast that asks a general model to read call summaries and score the deal can return a different score from the same evidence on a different day. When that happens, part of your week-over-week change is noise from the model, and no amount of deal review will find it.
Ask any AI forecasting vendor one question: if nothing in the deal changed, will the score stay the same? If the answer depends on the prompt, the run or the summary, then some of the movement on your Monday call is the tool. Spotlight.ai answers each qualification question inside a closed answer space and computes the deal score in code from stored values, so the same evidence gives the same answer every time. If a score moves, something in the deal moved.
Reading a deal that dropped out of commit
The close date slipped
Check whether the buyer named the new date or the rep picked it. A deal pushed for the third time has usually lost its compelling event, and the close date is the last field to admit it.
A stakeholder went quiet
Look at who attended the last two meetings. If the champion is still engaged but the economic buyer hasn't appeared since discovery, the deal didn't change this week. It was weaker than the forecast said, and this week someone noticed.
The amount shrank
A smaller number on the same deal has two very different causes. The buyer cut scope on a call, which you can find and quote. Or the rep trimmed the amount to make the deal feel safer, which leaves no trace in the deal at all. Both look identical in a snapshot report.
The deal sat still
Time on stage is the quietest signal in the pipeline. A deal can keep its stage, amount and close date for a month while nothing happens in it. Compare days in stage to your own historical average for that stage, and treat anything well past it as a forecast risk before the rep has to call it one.
What an inspectable forecast gives you
An inspectable forecast is a bottom-up forecast where every number can be traced back to the evidence behind it. Click into a change and you land on the deal and the buyer conversation that moved it. The week-over-week question becomes a lookup.
Spotlight.ai builds this on Spotlight's Calibrated-Decision AI-Model (CDAM). It reads the buyer conversations and email threads on every deal, and records each qualification answer with where it came from: confirmed by the buyer in conversation, logged by a rep, or calculated from other signals. The Spotlight Score then adjusts each deal's stage probability on evidence, including MEDDPICC status, stakeholder engagement, slippage and time on stage. When a deal's score drops, you can see which input dropped.
That matters for reps as much as for managers. A rep who moves a deal out of commit can point to the evidence that justified the call, so the forecast review is about the deal and what the buyer said.
Yonatan Olsha, Global Head of Sales at Tulip, put it this way: "I live in Spotlight for my forecast and deal inspection. This is what's giving me the evidence I need to truly forecast and be proactive on my pipeline risk."
The same evidence serves the rest of the revenue team. RevOps gets complete deal data and the slippage patterns behind the number. Value teams get a business case built from the buyer's own words. Customer success forecasts renewals with the same rigor that won the deal.
Running the week-over-week review in 30 minutes
Start from the delta, not the total. List every deal whose amount, close date, stage or category changed since the last snapshot.
Sort the changes by dollar impact. The five largest movements usually explain most of the difference.
Tag each change with its source. Buyer said it, rep judged it, or the system calculated it. Untagged changes go to the top of the agenda.
Check activity before accepting a push. No buyer conversation since the last review means the new close date is a guess.
Look for repeat slippage. A deal pushed for the second or third time belongs in a separate conversation about whether it is real.
Write down what you decided and why. Next week's review starts from this list, so the reasoning survives the call.
Start with last week's delta
Pull your last two forecast snapshots and list the deals that changed between them. If you can explain every line with a buyer conversation, an email or a recorded decision, your forecast is already inspectable. If half the lines come down to someone's judgment with nothing behind it, that's the half to fix first.
Spotlight.ai runs that trace on every deal, every week. It has done the same work across $9B+ of pipeline autonomously qualified and executed, and it can run it on last quarter's deals so you can compare its trace against what your team concluded at the time.
FAQs About Forecast Changes Week Over Week
Why does my sales forecast change every week?
Deals change amount, close date, stage or forecast category between reviews, and new deals enter or drop out. Some of those changes come from buyer activity and some come from reps adjusting their own confidence.
How do I find out which deals moved my forecast?
Compare two forecast snapshots deal by deal and list every change in amount, close date, stage or category. Sort by dollar impact, since a handful of deals usually explain most of the difference.
What is an inspectable sales forecast?
A bottom-up forecast where every number can be traced back to the deal and the evidence behind it, such as a buyer conversation, an email or a recorded decision. It lets you answer why the number moved without reconstructing the story from memory.
Can AI explain why a forecast changed?
Only if it records where each input came from. An AI forecast that produces a number without the evidence behind each deal can tell you the new total but not the reason for the change.
How do you tell evidence from opinion in a forecast?
Ask whether something happened in the deal or someone's confidence changed. A close date the buyer confirmed is evidence; a close date moved with no new activity is a judgment.
How does time on stage affect the forecast?
A deal that sits in a stage well past your historical average for that stage is at higher risk even if nothing else changed. Spotlight.ai feeds time on stage into the Spotlight Score, so it changes the deal's probability and the forecast.
Comments