Probability-weighted revenue forecast

For each stage, the probability is won / (won + lost) among deals that passed through the stage and closed in the 12 closed months before the snapshot. One window, always the same; with less than 12 months of history, use what there is and declare it. Passed through the stage: in HubSpot, hs_v2_date_entered_<stageId> filled on the deal, which counts even if the deal later moved to another pipeline (the field keeps only the latest entry, so it tells you whether it passed, not how many times). If the team does not mark deals lost and leaves them open indefinitely, the win rate among closed deals is inflated: for measurement, treat as lost any open deal with no activity or stage change for more than a declared N days. The CRM's default probability (the stage's metadata.probability or hs_deal_stage_probability on the deal) is only a declared provisional scale until there is history.

Active customer, partner, onboarding and post-sale pipelines also have stages with high probability, sometimes 80% to 100%. Added to the forecast, they become new revenue that does not exist. Write down the list of acquisition pipelines by pipeline ID, not by name, which changes. If a won deal is moved to another pipeline (for example, a closed sale lands in an activation or self-service pipeline), that destination pipeline goes on the list for wins and for conversion, or the final stage looks like it never wins.

A won deal enters at full amount in the month it was won and leaves the weighted part; a lost deal leaves at zero. Never apply the stage probability, a score or a temperature to a deal that is already closed.

The forecast month is the month of the expected close date. A deal with no expected date or with a date already past does not enter the current month: it goes to a separate overdue line the manager must review. Paused and frozen deals stay out: declare which field marks the pause (a temperature such as frozen, a status or a pause stage). Deals the rep removed from the forecast through the CRM's forecast category (in HubSpot, hs_manual_forecast_category = OMIT) also stay out and count as went out in the weekly snapshot.

If the team uses a qualification on the card (temperature, stars), it can adjust the probability by a fixed, written scale, and the scale states whether it replaces or multiplies the measured probability. Replace: probability = the scale value for the qualification x (1 - the stage's share of losses), where the stage's share of losses = deals lost in the month that were in that stage / all deals lost in the month. In a real case: top 80%, good 60% and average 40%. Multiply: probability = measured probability x factor, with factor 1 for the top qualification and below 1 for the others, or the same conversion is counted twice. A card with no qualification takes the lowest scale, never 100%.

Revenue is a deal metric. Never turn product lines into extra deals to add up. In recurring sales, declare whether the forecast is monthly value (MRR) or total contract value and use the right field; setup and one off services go on their own line.

Convert deals in other currencies to the reporting currency with a declared rate (fixed budget rate or monthly average) and keep the criterion week after week.

Save the forecast every week. If the CRM has a versioned deal history (an SCD Type 2 table with each version's valid from and valid to), the snapshot for any date since that history began can be rebuilt, without depending on weekly saves. Freeze each stage's probability in the snapshot: recalculating conversion every week makes untouched deals show up as amount changed. Comparing two snapshots, each deal falls into one category: came in, won, slipped (the expected date left the period), went out (lost, paused or recategorized) or amount changed. Apply the same period filter to both snapshots, or deals that change period become invisible.

If the team declares commit categories (will close, possible), show the commit next to the weighted forecast, never added together. The weighted number is history's reading; the commit is the rep's reading. The distance between them is information.

Deals created and won within the same month are never in the open pipeline of the snapshot at the start of the month, and in short cycle or self-service sales they can be most of the revenue. Add a run-rate line to expected revenue: the average, over the last 3 closed months, of revenue from deals created and won in the same month. Show that line apart from the weighted amount, so you know how much of the number comes from the pipeline and how much from pace, and measure accuracy with both.