Forecast Management: Deal Probabilities, Weekly Reviews, and Accuracy Checks
Shusaku Yosa
Forecast management maintains assumptions, owners and change reasons, then evaluates predictions against later results. Adding deal value multiplied by win probability produces a weighted sum of expected bookings, not a weighted average. Fix the measure, customer unit and horizon first.
This guide models bookings within a target quarter. Recognized revenue and cash receipts belong in separate schedules. All deals and amounts below are fictional.
Build the deal forecast
Deal | Amount | Probability of final win within the quarter | Expected bookings | Check |
|---|---|---|---|---|
A | ¥2,000,000 | 60% | ¥1,200,000 | Decision date and terms |
B | ¥1,000,000 | 30% | ¥300,000 | Evaluation and next agreement |
C | ¥3,000,000 | 20% | ¥600,000 | Budget and delivery timing |
Total | ¥6,000,000 | Do not take a simple average | ¥2,100,000 | Dependencies between deals |
The expected ¥2,100,000 does not mean a contract for exactly that amount will occur. A few large wins or losses can produce a very different realized total. Track shared risks, such as several deals depending on the same budget approval.
The probability of reaching the next stage differs from the probability of eventually winning. Eventual win probability also differs from winning within the forecast quarter. Calibrate stage-based assumptions against the relevant horizon.
Calibrate with comparable historical cohorts
Select opportunities that were at the same stage on a past as-of date, using comparable product, size and customer segments. Count how many were won by the end of the relevant horizon, including losses and carryovers in the original population.
If eight of 20 comparable proposal-stage opportunities closed within the target period, the observed rate is 40%. Twenty records still leave substantial uncertainty, and this period’s conditions may differ. Combine cohort evidence with documented deal-specific facts rather than presenting a highly precise probability unsupported by the sample.
Make the weekly review about changes
Area | Question | Retain |
|---|---|---|
Value | Did scope or discounting change? | Previous and new amount with evidence |
Timing | What must happen before an in-period signature? | Date, obstacle and next review |
Probability | What new fact changes the assessment? | Owner’s reason and evidence |
Duplication | Is the same agreement counted elsewhere? | Deal IDs and parent/child relationship |
Action | Who will do what by when? | Owner, deadline and support needed |
Review the change from the previous forecast instead of reading every record aloud. Increasing a probability is not progress by itself. Confirm actual milestones such as the decision-maker’s agreement or resolution of contract terms.
Separate slippage from loss
When a strong opportunity moves to the next quarter, reduce its in-period probability or move its expected timing. Do not retain the same value in both periods. A lost deal and a timing delay require different corrective actions, so classify them separately in the error review.
When a deal becomes confirmed actual bookings, remove it from the open-pipeline estimate. Reconcile new contracts, expansions and multiple owners’ records at the agreed account and agreement level.
Evaluate saved forecast versions
Forecasting: Principles and Practice describes out-of-sample evaluation and mean absolute error, or MAE. Compare forecasts made at the same horizon, such as a month-start estimate for the end of that month. Do not add information learned later to the original prediction.
Consider three fictional periods. Here signed error is defined as forecast minus actual.
Period | Forecast | Actual | Signed error | Absolute error |
|---|---|---|---|---|
1 | ¥10 million | ¥9 million | +¥1 million | ¥1 million |
2 | ¥12 million | ¥13 million | −¥1 million | ¥1 million |
3 | ¥11 million | ¥8 million | +¥3 million | ¥3 million |
MAE is (1 + 1 + 3) ÷ 3 = approximately ¥1.667 million. Mean signed error is (1 − 1 + 3) ÷ 3 = ¥1 million, indicating overforecasting in this example. Use both: positive and negative errors can cancel in the signed average. Percentage errors also become unstable when actual values are near zero.
Retain an auditable operating record
Save the as-of date, target period, metric definition, source-data version, change reason, owner, approval and actual-close date. Store a revision as a new version rather than rewriting the old forecast after a win becomes known.
Each month, identify whether volume, deal size, win rate or timing explains most of the error. The purpose is to improve the next estimate and the actions supporting it, not merely to report a precision score.
Related practical guides
- What Is a Business Forecast? Forecasts vs. Budgets and Targets
- CRM Integration for Marketing Campaigns: Field Mapping, Handoff, and Testing
- Marketing Budget Allocation: Scenario Planning and Reallocation Decisions
Keep execution and budgets connected
Xtrategy supports project management, monthly budgets and actuals, and customer and deal records. Review its features to decide how your owners, costs and review dates fit the workflow, then get started.