the bottoms-up forecast is the most confidently wrong artifact in the founder's drive. it has tabs, it has formulas, it has a row labeled "conservative scenario" the founder genuinely believes is conservative. it gets opened once a quarter, at which point everyone is surprised it's drifted 40% from reality.
the tops-down version is worse. "tam is $60B, we'll capture 0.2%, that's $120M arr in five years." nobody believes it. it's in the deck because the partner asked, and the partner doesn't believe it either. both sides are pretending, and the pretending is somehow load-bearing.
the honest forecast comes from neither. it comes from triangulating three views and taking the minimum. the minimum, almost without exception, is the truth.
the three forecasts that actually triangulate
a bottoms-up forecast multiplies leads × conversion × acv × growth and produces a hockey stick. each variable gets set to a number that's defensible on its own — 80th percentile, "we've seen this in good months". four 80th-percentile assumptions multiplied together is a 99th-percentile outcome. the founder thinks they built a conservative case. they built the best month of the year, every month, forever.
the fix is not a more pessimistic spreadsheet. the fix is to compute three forecasts using three independent methods and lay them next to each other. when they disagree, the lowest one is the plan.
the gap between $8.0m and $4.8m isn't a rounding error. it's the difference between "accelerate hiring" and "extend runway six months before the next round closes." a team running against the bottoms-up hires three aes in q3 to chase the curve. by q4 they've spent $600k on people who weren't going to close enough pipeline to pay for themselves — the pipeline was never going to be there. the triangulation would have caught this in week one.
what each method actually measures
the cohort-extrapolated forecast takes the last six months of new-logo acquisition and asks what happens if you keep doing exactly what you've been doing. no acceleration, no new channel. if you closed $80k of new mrr in march, $95k in april, $110k in may, the trailing 90-day pace is roughly $95k/month — call it $1.14m of new arr over 12 months, plus expansion at current nrr. this is the no-heroics number — what the company prints if nothing changes.
the capacity-constrained forecast asks what the team can actually close given the people you have right now. eight aes carrying $500k of annual quota each is $4.0m of theoretical capacity. attainment runs 70% in a healthy team and 50% in a stretched one. realistic close from the existing roster is $2.0m to $2.8m of new arr. if your bottoms-up says $5m and your capacity says $2.8m, you have a capacity problem before you have a pipeline problem.
the historical compound forecast is the simplest. take your trailing six-month mom growth rate, compound it forward 12 months. if you grew 8% mom in h1, the run-rate forecast is current arr times 1.08 to the twelfth, plus gross adds from new sales motion. it's mechanical. it doesn't care about your story.
the three numbers will not agree. they're not supposed to. they're supposed to bracket the truth from three different directions, and the lowest of them is what you plan against.
why the bottoms-up always wins inside the founder's head
the bottoms-up is the only forecast that lets the founder be a hero. every other method just records what happened. the bottoms-up lets the founder pick the assumptions, defend each one in isolation, and arrive at a number that justifies the next hire. it's the financial equivalent of a deck slide — designed to be read, not stress-tested.
the same set of assumptions blows up the model across teams that get the forecast wrong by 2x or more.
every conversion rate is set to the best quarter, not the average. the founder remembers the month sdr-to-meeting hit 18%. it's averaged 9% for three years. the model uses 18% because "we've proven we can hit it" — it doesn't know 18% required a launch, a press hit, and the head of marketing personally writing the sequence.
ramp is assumed instant. new aes hired in q1 are modeled at full attainment in q2. real-world ramp is 6-9 months for series a saas. the model overstates revenue by roughly $400k per ae for the year.
no churn is modeled in the first 18 months. the founder assumes the q1 cohort is still paying in q4 at the same arpu. some are, some downgraded, some logo-churned. without a contraction line, gross revenue overstates net revenue by 8-15%.
the forecast that lets you be a hero is the one you should distrust most — heroes do not compound at the 95th percentile for four consecutive quarters.
what to do with the triangle
run all three monthly. plot the divergence. when the bottoms-up runs more than 30% above the historical compound, either the bottoms-up is wrong or the company is about to inflect. if it's the second, you'll have a leading indicator — a new channel firing, a new product hitting pmf, ramp closing faster than modeled. if you can't name the indicator, the bottoms-up is wrong. plan against the lower number.
the operating rhythm a few well-run teams use — the cfo (or the founder, if there isn't one) opens the forecast triangle at the monthly business review. "three numbers — $4.8m, $5.0m, $7.8m. planning against $4.8m. last month we said $4.6m. what changed." the discussion is short. the discipline is the point.
how zift handles this
zift computes the three-forecast triangle from your stripe revenue, ae headcount, and trailing-six-month pace, every fifteen minutes. monday morning the briefing shows the bottoms-up, capacity, and historical-compound numbers side by side, the delta from last month, and which assumption moved when the divergence widens.
if you're a finance lead at a series a team running this across multiple product lines or geographies, zift handles that too.
the forecast that gets you funded is rarely the one that ships. the one that ships is the lowest of the three you should have been running all along.