How to Forecast Revenue by Segment

A company-level forecast that lands within a point of plan can still be wrong in every way that matters. Enterprise came in 20 percent short, SMB overdelivered by the same amount, the two errors canceled, and the number looked correct. Nobody learned anything, and the same offsetting misses will repeat next quarter. Segmenting the forecast is how you stop that pattern.

Why does a blended forecast hide risk?

Because averaging two different motions produces a rate that describes neither one.

An enterprise deal might run nine months at high value through a procurement process. An SMB deal might run three weeks at a fraction of the value with a single decision maker. A blended win rate sits somewhere between them and applies accurately to nothing in your pipeline.

The damage compounds when the mix shifts. If enterprise grows from 30 percent of pipeline to 45 percent, your blended rate is now wrong even though no individual segment changed behavior. The forecast degrades and the cause is invisible, because the model has no dimension that would reveal it.

How do you define segments that actually behave differently?

Test candidate splits against your closed history and keep only the ones that separate win rate, cycle length, or deal size.

Start with the obvious candidates. Company size, new business against expansion, and inbound against outbound source. Calculate the three rates for each split and look at the spread. If enterprise wins at 18 percent and mid-market at 21 percent, that split is not doing work and you should collapse it.

Those figures illustrate the structure rather than serving as benchmarks. The coverage column is the inverse of the win rate, which is the honest way to set a coverage target. Definitions for both measures sit in the win rate glossary entry and the pipeline coverage glossary entry.

Resist segmenting by sales team unless the teams genuinely run different motions. Territory boundaries change, and a forecast dimension that reorganizes every year cannot accumulate the history needed to produce stable rates.

What changes in the model for each segment?

Every rate, plus the definition of a stale deal.

Each segment needs its own conversion rates by stage, its own cycle length, and its own realization ratio between pipeline amount and closed-won amount. That last one is frequently ignored and consistently expensive. Pipeline amounts run above what deals close for, and the gap differs by segment. A pattern that shows up repeatedly is an $80,000 average deal size in open pipeline against a $40,000 average in closed-won.

Aging thresholds also differ. A ninety day old SMB opportunity has almost certainly died. A ninety day old enterprise opportunity is midway through a normal cycle. Applying one staleness rule across segments either purges live enterprise deals or leaves dead SMB deals sitting in the number.

The useful general rule is a twelve month ceiling on any open opportunity with no meaningful activity, where meaningful means a change to stage, close date, or amount. Across ORM's customer base, more than 10 percent of open pipeline has not been touched in twelve months. Within segments, tighten that threshold to something closer to two full cycle lengths.

How do you roll segments into one company number?

Sum the segment forecasts, then reconcile the total against a top-down check rather than forcing agreement.

The sum is the forecast. What matters is what you do with the difference between that sum and the plan. That gap is the planning output, and dividing it by segment tells you which motion has to change and how much.

Present the rollup with segment contribution visible. A quarter where 70 percent of the number comes from enterprise carries a different risk profile than one where enterprise carries 30 percent, even at identical totals, because enterprise revenue concentrates into fewer deals. Concentration risk is invisible in a single number and obvious in a segmented one.

How do you handle a segment with too few deals?

Pool it with its nearest behavioral neighbor, or forecast it deal by deal.

Rates calculated on small samples are unstable. Thirty closed opportunities is a reasonable floor for a conversion rate, and below that the quarter-to-quarter swing in the rate will exceed the effect you were trying to capture.

Two options work. Pool the thin segment with whichever segment has the closest cycle length and win rate, and accept a slightly blended rate for that portion. Or forecast the segment deal by deal, with a named owner giving a judgment call on each opportunity, which is workable when the deal count is small enough to review individually.

How does segmentation change the forecast review?

It turns the review from a debate about the total into a specific question about one motion.

Run the variance analysis by segment every period. When enterprise misses and SMB covers it, the review surfaces that immediately rather than a quarter later when the pattern finally becomes too large to hide.

Coverage discussion changes as well. The standard range runs from 3x to 5x. Across ORM's customer base most customers sit near 3.5x, with individual customers as low as 1.4x. Quoting one ratio across a segmented business means at least one segment is being measured against a target that has nothing to do with its conversion behavior. The full argument against a universal ratio is in why the 3x pipeline coverage rule is wrong.