Weighted Pipeline: Formula, Deal Stage Probability and Calibration

Every deal in Contract sent is weighted at 90% in HubSpot's default sales pipeline, a template value rather than a rate measured from the portal's own closed deals. A weighted pipeline built on the defaults, or on a manager's estimates, multiplies every deal in a stage by the same unmeasured number and repeats the error every quarter. Once the probabilities are measured and the pipeline edited, the board view moves to the new figures, while the forecast tool keeps the old ones on every deal whose stored probability HubSpot has stopped maintaining. On a dashboard the two failures look identical.

This article explains how HubSpot computes a weighted pipeline and how to set stage probabilities from evidence. Its first half covers the formula, the two places HubSpot stores a probability, the weighted figure that multiplies each, and the symptoms of the two disagreeing. The second half shows why a guessed probability biases a forecast, calibrates a sample pipeline, and sets out the calibration procedure, its costs and the limits of the method.

A weighted pipeline is the sum, across open deals, of each amount multiplied by its probability of closing won. A deal stage probability, or stage probability, is the percentage set on a deal stage in pipeline settings. Deal probability (hs_deal_stage_probability) is a separate number stored on each deal, and the weighted amount (hs_projected_amount) is a deal's amount multiplied by its Deal probability. A stage-to-close rate is the share of deals that entered a stage and later closed won, among those that have since closed.

The Weighted Pipeline Formula

The weighted pipeline formula is a sum of products: Σ (amount × probability) over every open deal in scope. The result is an expected value: if each probability equals its deal's true chance of closing won, the weighted pipeline is the average outcome over many repetitions of the same pipeline, and no single quarter need land on it. The formula has no term for time, which leaves the forecast period to the close date the forecast tool filters on. It also counts only deals that already exist, so a weighted pipeline is one input to revenue forecasting rather than a method on its own.

Stage Probability and Deal Probability in HubSpot

Stage probabilities are set in Settings under Data Management > Objects > Deals, on the Pipelines tab: each stage has a Deal probability dropdown, and Add option saves a custom value (HubSpot's pipeline documentation). The default sales pipeline sets its five open stages at 20%, 40%, 60%, 80% and 90%, with Closed won at 100% and Closed lost at 0%.

The forecast tool's weighted pipeline does not multiply that setting: its default deal amount, Weighted amount, multiplies each deal's own Deal probability. HubSpot updates it when a user moves the deal to a new stage, per the default deal properties documentation. The same page counts edits on the record, bulk edits, imports and writes by data sync integrations or the API as manual updates, and after one, HubSpot stops updating the value from open stages until the deal closes. The documentation does not say whether editing a stage's probability rewrites the deals already sitting in that stage.

The deal object definition publishes the weighted amount's formula, hs_deal_stage_probability * amount, with zero returned when no probability is present.

Weighted Figures and the Probability Each Uses

The stage-setting figures and the Deal probability figures agree only while every open deal's Deal probability equals its stage's setting. Each deal where the two differ separates the totals by its amount times the difference, and neither surface's documentation describes a reconciliation. A recalibration therefore moves the board view at once and is guaranteed to reach the forecast only on deals that change stage afterwards and that no manual update has written.

Deal Probability Symptoms and Their Causes

The board view and the forecast tool disagree for the same deals, or a recalibration moved one and not the other. Some open deals carry a Deal probability that differs from their stage's setting, left by a manual update or never rewritten by the settings edit. A deal view of one stage with a Deal probability column lists them, and steps 6 and 7 of the calibration procedure below realign and verify them.

A deal counts at full value in the forecast tool's Open pipeline and adds nothing to its Weighted pipeline. Its Deal probability is empty, and the published formula returns zero without one, so an integration or API call that clears the value cuts the deal's contribution to zero in every figure built on Weighted amount, until a workflow restores it. The board view still weights it at the stage setting.

A deal reports as closed lost in an open stage. The deal object definition has Is closed lost testing the deal's own Deal probability for zero or less while Is Deal Closed? tests the stage's setting, so a deal set to 0% by hand counts as lost in one report and open in another. Moving it to Closed lost resolves both; the deal score documentation names the stage as the source of truth for a closed deal.

Guessed Probabilities and Forecast Bias

A stage probability is a probability forecast applied to every deal in the stage, and forecast verification judges such a forecast by reliability: among the cases given 60%, the event should occur about 60% of the time. A 1977 study of subjective probability forecasts by Murphy and Winkler found that weather forecasters could meet that test, so judgement is not the defect. A probability is unverified until it has been compared with outcomes, and neither a default nor a guess has been.

Random deal outcomes partly cancel across a pipeline, so the noise they add shrinks relative to expected won revenue as deals are added. A wrong stage probability adds bias, which does not cancel, because the same number multiplies every deal in the stage: a stage set at 90% that closes at 69% overstates every deal in it by the same ratio, whether it holds three deals or three hundred.

Stage-to-Close Rates on Sample Data

The portal below is sample data and describes no client: one pipeline on HubSpot's default stages and probabilities, 28 open deals worth $1,030,000 closing in the fourth quarter of 2026, and the deals that entered each stage between 1 October 2025 and 31 March 2026, read on 7 October 2026. Deals still open are reported but excluded from the rates.

Counting deals, as these rates do, estimates a deal's chance of closing won; Amount won divided by Amount closed estimates the share of a stage's dollars that close. Where win rate varies with deal size, only the amount rate keeps the weighted total unbiased, though unequal amounts make it less precise; the sample assumes no such variation.

The standard error of a proportion, √(p(1 − p) / n), measures each count rate's precision. Contract sent rests on 29 closed deals, so its standard error is about 8.6 points and two standard errors span roughly 52% to 86%. That range excludes HubSpot's 90% and cannot separate 65% from 75%, so the setting becomes 70% rather than a 69.0% that claims precision the sample lacks. Rounded the same way, the five calibrated settings are 15%, 25%, 35%, 50% and 70%.

Both weighted columns follow the board view's arithmetic, and the forecast tool departs from it on one deal. A $45,000 Contract sent deal was imported with a probability of 25%, which stopped HubSpot updating its Deal probability: the board view counts it at $40,500 and the forecast tool at $11,250. After recalibration the board view counts it at $31,500, while the forecast tool stays at $11,250 until a workflow rewrites the value: a $20,250 disagreement on one deal.

The defaults overstate the calibrated figure by $181,250, or 51%, with every stage erring in the same direction. Assuming independent outcomes and equal amounts within each stage, the standard deviation of won revenue at the calibrated probabilities, √Σ n·a²·p(1 − p), is about $88,000, so the default-weighted figure sits 2.1 standard deviations high.

At ten open deals in the same mix, two standard deviations span 83% of the expected value either way, so one quarter's miss says little about the probabilities; calibration rests on a cohort of closed deals pooled by stage.

Calibration Procedure

1. Confirm that stage calculated properties are on for every stage: on the same Pipelines tab, open the pipeline and read the stage calculated properties column. They are off by default for new pipelines and stages, need Professional or Enterprise, and backfill from existing data once switched on (stage calculated properties).
2. Pick a cohort window that ended long enough ago for its deals to resolve, such as six months ending six months before the measurement date, and record the dates.
3. Create a single object report on deals, filtered to the pipeline, to the stage's Date entered property within the window, and to Is Deal Closed? equal to true, broken down by Is Closed Won, counting deals and summing Amount. Repeat for each stage, noting the deals still open.
4. Divide won by won plus lost, in dollars where win rate varies with deal size, then compute the standard error and round to the nearest 5 points. Where two standard errors either side still contain the current setting, leave it: the evidence does not support a change.
5. Enter each calibrated value in the stage's Deal probability dropdown.
6. Build a deal-based workflow that enrolls each stage's open deals, with a branch per stage, and uses the Edit record action to set Deal probability to the calibrated value, including on deals HubSpot no longer updates.
7. Multiply one stage's open total by its new setting, the board view's documented computation, and compare the product with a report summing Weighted amount for the same deals. The two match to the dollar once the rewrite has reached every deal, and a gap means some deals still carry another probability.
8. Move a deal that the step 6 workflow rewrote to the next stage and confirm that its Deal probability changes to that stage's setting. The documentation's list of manual updates neither names nor excludes workflow writes, and if the test fails, the workflow must re-enroll on every Deal stage change and maintain the value itself.

The Business Case for Calibration

What calibration buys is a weighted pipeline whose remaining error is outcome noise, which shrinks as the pipeline grows, plus the rates' sampling error, which shrinks as the cohort grows, provided the open deals resemble the cohort in size and age.

What it costs is the Professional tier that stage calculated properties, single object reports and workflows need, a report per stage, a workflow per pipeline, and remeasurement, because close rates move with pricing, product and team. The forecast tool needs Sales Hub or Service Hub Professional or Enterprise, with a Sales or Service seat for each user other than Super Admins. Lower calibrated probabilities cut the weighted pipeline the day they apply, a drop to explain before anyone reads it as a lost quarter.

The case is strongest with enough closed deals per stage to estimate a rate and one sales motion per pipeline. It is weakest with a few dozen deals a year, where sampling error can exceed the correction, and where one pipeline mixes motions with different close rates, such as new business and renewals, which separate pipelines serve better.

Boundaries of the Method

This covers stage-based weighting in HubSpot. Forecast categories, forecast submissions and the forecast tool's configuration belong to HubSpot forecasting, and the AI deal score, which the Weighted amount formula does not use, is out of scope. Settings, navigation and formulas are as documented in October 2026, and HubSpot's documentation governs wherever it has since changed.

A rate measured at stage entry describes a deal on arrival, while the weighted amount applies it equally to a deal four months into the stage; weighting by age takes a workflow that adjusts Deal probability from Time in current stage, or a calculation property reported in place of Weighted amount. Excluding a cohort's open deals raises its rate if they later lose: Appointment scheduled falls from 13.6% to 12.2% if all 18 do. The range in the second figure assumes independent outcomes, so it understates the spread when deals share a cause such as a market shock.

In Summary

HubSpot stores a deal's probability twice, as a stage setting that the board view multiplies and as a Deal probability that the forecast tool multiplies, and the two diverge on every deal whose value a manual update has written. Either way the weighted pipeline is Σ (amount × probability) over open deals, an expected value with no term for time.

Guessed probabilities, HubSpot's defaults included, bias a forecast because one wrong number multiplies every deal in a stage and does not cancel as deals are added. Calibration costs a Professional subscription and a report per stage, and helps least where stages close few deals.

After calibrating, rewrite Deal probability on every open deal and confirm that one stage's total times its setting equals the sum of its Weighted amount to the dollar, which shows that the forecast multiplies what was measured.