How to Build a Bottom-Up Revenue Forecast Model

A bottom-up forecast model is the most defensible way to produce a revenue projection because every number traces back to a real deal with a named account. When leadership asks why the forecast is what it is, you can answer at the opportunity level. Building one correctly means getting the stage weights right, segmenting the model by the right dimensions, and being honest about what the model cannot predict.

Step 1: Define Your Stage Win Rates

The backbone of any weighted pipeline model is the historical win rate at each opportunity stage. This is not the same as your overall win rate. You need a conversion rate from each stage to closed won.

Pull every opportunity from the past four to eight quarters that reached at least one defined stage, and calculate what percentage closed.

Stage	Definition	Historical Close Rate
Qualified	Confirmed budget, authority, need, timeline	Your actual data
Discovery complete	Discovery call done, next step confirmed	Your actual data
Technical evaluation	Active proof of concept or pilot underway	Your actual data
Proposal sent	Written proposal submitted	Your actual data
Verbal commitment	Buyer has indicated intent to proceed	Your actual data

Do not use a single blended win rate applied to all pipeline. Stage-level rates are what make the model useful. A deal in late negotiation should carry a much higher probability than a deal that just entered discovery, and the model should reflect that.

If your volume of historical closed deals is too low to produce statistically stable rates, use a larger lookback window or segment less granularly. A less precise rate grounded in real data is better than a precise rate based on intuition.

Step 2: Segment the Model by the Right Dimensions

A single rate applied across your entire pipeline is more accurate than no model, but it misses important variation. Segment your stage win rates and average deal sizes by at least two dimensions:

Segment or market. Enterprise deals convert differently than mid-market deals. If you blend them, you will systematically overstate or understate depending on your current pipeline mix. Opportunity type. New logo deals typically have lower win rates and longer cycles than expansion or renewal opportunities. Model them separately.

If your volume supports it, segment further by deal source (inbound versus outbound), product line, or sales rep tenure. Add dimensions only when you have enough closed deals in each bucket to produce a stable rate.

Step 3: Build the Rep-Level Rollup

For each open opportunity, calculate the weighted value:

Weighted value = Deal amount x Stage close rate

If a deal is worth a certain amount and carries a close rate reflecting its current stage, the weighted value is that amount multiplied by the rate. Sum all weighted values at the rep level, then roll up to the team level.

A rep-level view lets you identify two problems at once: reps whose pipeline volume is insufficient relative to their quota, and reps whose pipeline is concentrated in a single large deal that will dominate the forecast whether it closes or not.

Step 4: Separate Committed Deals from Weighted Pipeline

Not all deals belong in the weighted calculation. Deals in late stages, particularly those with a verbal commitment or signed LOI, should be tracked separately as commit-category deals.

The commit category represents what the rep is willing to state publicly they expect to close. The weighted model represents what the math says should close. Both numbers matter:

- Weighted pipeline: Statistically expected revenue based on stage mix. - Committed forecast: Rep and manager judgment about which deals are actually going to land.

The gap between the two is a signal. If the weighted model is significantly higher than the committed forecast, either the model is generous or the reps are sandbagging. If the committed forecast is significantly above the weighted model, the reps are being optimistic. Investigate the gap rather than choosing one number over the other.

Step 5: Apply Cycle Length to Determine Period Eligibility

A deal's stage close rate answers the question of whether it will close at all. Cycle length answers whether it will close this period.

For each opportunity, calculate how long it has been open. Compare that to the median total cycle length for deals at that stage. Deals that are very early in a long cycle, regardless of their stage, are poor candidates for this quarter's forecast even if their stage probability looks good.

Cycle position	Forecast eligibility
Early (under half of median cycle)	Remove from near-term forecast unless late stage
Mid-cycle	Include in weighted model with stage probability
Late (at or past median cycle)	Flag for close-date review; may have slipped

This adjustment prevents the model from overstating near-term revenue by including deals that are realistically not going to resolve in the current period.

Step 6: Present the Model With Its Assumptions Visible

The output of the model is only useful if the people reading it understand the inputs. For every forecast presentation, show:

- The stage win rates applied. - The segment split used. - The pipeline total by stage before and after weighting. - The committed number alongside the weighted number. - Any large single deals that are disproportionately affecting the total (and whether they are included or excluded from the base case).

A model presented as a black box erodes trust. Showing the inputs produces real conversations about the assumptions, which is where forecast accuracy actually improves.

For related frameworks, see revenue forecasting models and weighted sales forecast. For how manager and rep commits roll up into a team number, see roll-up forecast.

Common Mistakes

Using a single win rate across all pipeline. Stage-level rates exist for a reason. Blending them produces a number that is accurate for no specific deal. Not segmenting by opportunity type. New logo and expansion deals behave differently. A model that treats them identically will mislead you about where to focus. Ignoring cycle length. A deal at a high-probability stage that opened two weeks ago is not a realistic close this quarter. Stage probability and timing are separate inputs; the model needs both. Updating rates with insufficient data. Recalibrating win rates on small sample sizes creates noise. Set a minimum closed-deal volume before changing the rates. Treating the model as the answer. The model organizes evidence. Rep-level deal inspection is still required.

Frequently Asked Questions

What is a bottom-up revenue forecast model?

A bottom-up forecast model builds the revenue projection from individual opportunity data rather than from a top-line growth assumption. Each deal is assessed by its stage, expected value, and probability of closing in the period. The aggregate of these deal-level projections becomes the forecast. This approach is more grounded than top-down models because it connects to named accounts and actual pipeline activity.

What inputs does a stage-weighted forecast model require?

The model needs four inputs per opportunity: current stage, estimated deal value, expected close date, and the historical win rate for that stage in your business. The stage win rate is the key calibration input. Without it, you are weighting deals by assumption rather than by evidence.

How do you handle the difference between the rep's forecast and the model's output?

Treat the model output and the rep's call as two separate inputs. Where they diverge significantly, investigate the individual deal. The rep may have context that the model cannot capture (a verbal commitment, an executive sponsor change). The model may be flagging deals the rep is overvaluing due to familiarity. Neither is automatically right. The manager's job is to resolve the gap with evidence.