SaaS Forecast Model: Predict MRR With 3 Inputs

Most SaaS forecast models fail because they try to solve too much. Founders open a spreadsheet and begin adding acquisition assumptions, conversion rates, channel splits, churn by cohort, expansion by plan tier, hiring impact on capacity, and pricing sensitivity. Each input feels useful. Together they produce a model that is technically comprehensive and practically useless — too many variables to trust, too complex to update, and too optimistic because every individual assumption was slightly too generous.

The 3-input forecast model starts from a different premise: a founder who wants to know what next month’s MRR will probably look like does not need a model that predicts everything. They need a model that captures the three movements that actually drive recurring revenue change.

What is a SaaS forecast model?

A SaaS forecast model is a repeatable way to estimate how recurring revenue will change over the next period based on a small number of honest assumptions. It is not a prediction — no forecast is. It is a structured estimate that gives the founder a concrete number to hold the business against, plus a range of scenarios that shows how sensitive that number is to the assumptions that are most likely to be wrong.

The narrower definition distinguishes it from a full SaaS financial model: a forecast model focuses on MRR movement, while a financial model also covers costs, cash, and runway. For most founders, the forecast model is the first and most frequently used layer — updated monthly to produce the expected ending MRR, then compared to actuals to calibrate assumptions. Y Combinator’s startup financial guidance makes a similar point: start with the simplest model that answers the operating question, and earn complexity only when the business demands it.

For the broader model that includes costs and runway, see SaaS Financial Model: The Minimal Sheet That Predicts Runway.

Every forecast needs a clean MRR baseline. Get yours from Stripe in 90 seconds →

The 3 inputs that are enough

The three movements that explain almost all MRR change in an early SaaS product:

Input 1: New MRR. The recurring revenue expected from customers who are not currently paying. This is the acquisition contribution — how much fresh recurring revenue is likely to land from new paying customers this period. Use recent monthly averages as the starting point; do not model heroic conversion rates.

Input 2: Expansion MRR. The additional recurring revenue expected from existing customers upgrading or increasing usage. This can be zero if there is no upgrade pathway yet — but it should not be omitted from the model, because assuming it away makes the forecast less accurate as soon as upgrade behaviour exists.

Input 3: Churned MRR. The recurring revenue expected to be lost. This is the most commonly underestimated input — founders tend to apply last month’s churn rate optimistically or assume “this month will be better.” Use recent averages; in the downside scenario, use recent averages multiplied by 1.2–1.3 to test sensitivity.

The core formula

Next Month MRR = Current MRR + New MRR + Expansion MRR − Churned MRR

That is the complete model. Adding contraction MRR as a fourth input is worth doing once the business has measurable downgrade activity — but for most early SaaS, churn encompasses most of the leakage and the three-input version is accurate enough.

For the clean definitions of what belongs in MRR, see What Is MRR? The Clean Version.

Why the 3-input model works

More detail in a forecast model does not produce more accuracy — it produces more places for optimism to hide. A 30-input model where each input is 5% too optimistic produces a forecast that is systematically wrong in the same direction every time, with no obvious way to identify where the error entered.

The 3-input model has three places where assumptions can be checked and calibrated against actuals. When the forecast misses, it is immediately clear whether new MRR was overestimated, churned MRR was underestimated, or expansion was absent. That diagnosis is what makes the model improve over time.

This model also captures the structure of the problem correctly: the founder’s ability to influence MRR is concentrated in exactly these three movements — finding new customers, creating upgrade pathways, and retaining existing ones. A model with 30 inputs disperses the founder’s attention across dimensions that they have limited ability to change in the near term. The 3-input model focuses it. Bessemer’s State of the Cloud report consistently shows that the highest-performing SaaS companies optimise around a small number of operating levers — acquisition rate, expansion, and churn — rather than trying to manage dozens of variables simultaneously.

A worked example: forecasting MRR with 3 inputs

A SaaS analytics product in month four. Current state:

Current MRR: €10,000

Forecast inputs for next month:

New MRR: €1,500
Expansion MRR: €600
Churned MRR: €500

Base case forecast:

Next Month MRR = 10,000 + 1,500 + 600 − 500 = 11,600

Forecasted ending MRR: €11,600

The forecast is complete. Now the useful work begins: checking the assumptions.

Are these inputs honest?

New MRR of €1,500 — is this based on recent conversion rates applied to current pipeline? Or is it a round number that felt right? If the last three months of new MRR were €1,200, €1,100, and €1,300, then €1,500 is optimistic. A more conservative estimate uses the recent average: €1,200.

Expansion MRR of €600 — has expansion been consistent at this level, or is it variable? If expansion was €800 last month and €200 the month before, using €600 as the estimate may be reasonable but deserves scrutiny.

Churned MRR of €500 — is this split into voluntary and failed payment? Failed payment churn is partially recoverable with a dunning sequence; voluntary churn requires product work. If recent months show €200–250 of failed payment churn, a well-implemented dunning sequence could reduce realised churned MRR. The forecast should reflect achievable churn, not aspirational churn.

Three scenarios with the same formula

The value of the 3-input model multiplies when run across three scenarios rather than one. The scenario structure is simple: change the three inputs to reflect different operating realities, compute ending MRR for each, and compare the gap.

Base case — current operating assumptions:

New MRR: €1,500 · Expansion: €600 · Churned: €500
Forecast: €10,000 + 1,500 + 600 − 500 = €11,600

Upside case — slightly better acquisition, better retention:

New MRR: €1,800 · Expansion: €700 · Churned: €450
Forecast: €10,000 + 1,800 + 700 − 450 = €12,050

Downside case — slower growth, higher churn:

New MRR: €1,200 · Expansion: €500 · Churned: €700
Forecast: €10,000 + 1,200 + 500 − 700 = €11,000

Side-by-side table:

Scenario	New MRR	Expansion	Churned	Ending MRR	Gap vs Base
Base	1,500	600	500	11,600	—
Upside	1,800	700	450	12,050	+450
Downside	1,200	500	700	11,000	−600

The gap between base and downside is €600 in ending MRR — which is €7,200 annualised. That is the operating risk the founder should be thinking about, not the polished middle number.

For the full scenario stress-testing methodology, see Scenario Modeling for Bootstrappers: Stress-Test in 15 Minutes.

The forecast habit that makes the model useful

A forecast model that is created once and never compared to actuals is just a formatted opinion. The model becomes useful through repetition:

Before the month: set the three inputs for the period based on recent trends and current expectations. Produce a base, upside, and downside forecast.

During the month: track new MRR, expansion, and churn as they materialise. NoNoiseMetrics surfaces these in real time from Stripe — no manual tracking required.

After the month: compare forecast to actual. If the forecast missed, identify which input was wrong and by how much. Adjust the assumption methodology (not just the number) for next month.

This loop — forecast, track, compare, calibrate — is what makes a 3-input model produce progressively better estimates. The first forecast will be imprecise. The sixth or seventh will be calibrated to how this specific business actually behaves.

For the comparison layer, see Budget vs Actual: The Weekly Loop That Keeps You Alive. KeyBanc Capital Markets’ SaaS Survey data shows that founders who track forecast vs actuals monthly make materially better capital allocation decisions than those who review annually.

Common SaaS forecast model mistakes

Too many inputs. Each additional input is another place for optimism to hide. Founders who build 15-input forecast models rarely have the data to populate 15 inputs honestly — most inputs are guesses that compound into a structurally optimistic output. Three honest inputs beat fifteen guesses.

Anchoring new MRR on aspirations rather than actuals. The most reliable estimate of next month’s new MRR is a slightly adjusted average of recent months. Founders who forecast new MRR based on potential — “we have five good leads” — systematically overestimate this input. Use actuals as the anchor; adjust up only when a specific, concrete change in acquisition activity justifies it.

Ignoring expansion MRR. A common simplification that reduces model accuracy as the business matures. If the product has an upgrade path and customers are using it, expansion MRR contributes materially to month-over-month MRR growth. Omitting it makes the model underestimate MRR in good months and makes new MRR look more important than it is.

Underestimating churned MRR. The most pervasive error. Founders apply last month’s actual churn — which was often a good month — rather than a considered average. The downside scenario exists specifically to pressure-test this assumption. What happens if churn is 40% higher than expected? That answer should inform how much runway is genuinely comfortable, not the polished base case.

No comparison to actuals. A forecast that is never checked against what actually happened cannot improve. This is the critical step most founders skip, often because the comparison reveals that assumptions were wrong — which is uncomfortable but necessary.

When to add a 4th input: contraction MRR

Contraction MRR (recurring revenue lost to downgrades, without full cancellation) is worth adding as a fourth input once it becomes a measurable and material part of the business. In early-stage SaaS where most customers are on fixed tiers, contraction is often near zero and can be safely folded into churned MRR conceptually.

The signal that contraction deserves its own line: if downgrade events appear consistently in Stripe data — customers moving from Growth to Starter, or reducing seat counts — and the aggregate impact is more than 5–10% of churned MRR, track it separately. The intervention for contraction (pricing structure, packaging clarity, downgrade friction) is different from the intervention for full cancellation, and the model should reflect that difference.

SaaS revenue forecast model: the automation version

A useful revenue forecast tool does not require manual spreadsheet work. The minimal automated version:

Pull current MRR from one trusted billing source (Stripe subscription events)
Surface the three historical inputs — recent new MRR average, recent expansion average, recent churned MRR average — as defaults
Let the founder adjust each input and toggle between scenarios
Display forecast ending MRR for each scenario side by side
After the month closes, compare forecast to actual and surface the delta per input

NoNoiseMetrics is building this as part of the runway forecaster — a lightweight tool that produces a 3-input forecast from live Stripe data, requiring no spreadsheet work and no manual input population.

JSON model for a 3-input SaaS forecast

{
  "saas_forecast_model": {
    "period": "2026-05",
    "currency": "EUR",
    "current_mrr": 10000,
    "inputs": {
      "new_mrr": 1500,
      "expansion_mrr": 600,
      "churned_mrr": 500
    },
    "formula": "current_mrr + new_mrr + expansion_mrr - churned_mrr",
    "forecast_ending_mrr": 11600
  },
  "scenarios": {
    "base": {
      "new_mrr": 1500,
      "expansion_mrr": 600,
      "churned_mrr": 500,
      "ending_mrr": 11600
    },
    "upside": {
      "new_mrr": 1800,
      "expansion_mrr": 700,
      "churned_mrr": 450,
      "ending_mrr": 12050
    },
    "downside": {
      "new_mrr": 1200,
      "expansion_mrr": 500,
      "churned_mrr": 700,
      "ending_mrr": 11000
    }
  },
  "forecast_habit": {
    "before_month": "Set 3 inputs from recent averages",
    "during_month": "Track actuals from billing",
    "after_month": "Compare forecast vs actual; calibrate assumptions"
  }
}

FAQ

What is a SaaS forecast model?

A SaaS forecast model is a repeatable way to estimate how recurring revenue will change over the next period based on a small number of explicit assumptions. The minimal version uses three inputs — new MRR, expansion MRR, and churned MRR — to forecast next month’s ending MRR from the current base.

How do you forecast MRR for SaaS?

The formula is: Next Month MRR = Current MRR + New MRR + Expansion MRR − Churned MRR. Use recent monthly averages as the basis for each input rather than aspirational targets. Run the forecast across three scenarios (base, upside, downside) to understand the range of realistic outcomes.

What inputs matter most in a SaaS forecast model?

New MRR (acquisition contribution), expansion MRR (upgrade contribution), and churned MRR (leakage). Churned MRR is typically the most consequential input because it is the most commonly underestimated. For most early SaaS products, these three inputs explain the vast majority of month-over-month MRR change.

How detailed should a SaaS forecast model be?

Only as detailed as the quality of available data can support. Three inputs, honestly populated from recent actuals, produce more reliable forecasts than fifteen inputs populated with estimates. Complexity should be earned by specific business needs, not applied to create an impression of sophistication.

What is the difference between a SaaS forecast model and a SaaS financial model?

A SaaS forecast model focuses on projecting recurring revenue movement — new MRR, expansion, churn, and ending MRR. A SaaS financial model is broader and includes costs, burn, cash, and runway. Most founders should start with the forecast model (lightweight, updated monthly) and embed it in a financial model when cost and runway tracking becomes material to decisions.

How accurate is a 3-input SaaS forecast model?

Accuracy depends on the quality and honesty of the three inputs, not the number of inputs. A 3-input model calibrated against six months of actuals typically outperforms a 30-input model populated with first-guess estimates. The model becomes more accurate through the monthly forecast-vs-actual comparison loop, not through adding more inputs.

What is a SaaS revenue forecast model vs a SaaS revenue model?

A SaaS revenue forecast model projects future MRR movement based on explicit assumptions. A SaaS revenue model may refer to the broader pricing and monetisation structure — the model that determines how the product charges customers. Both are useful; they answer different questions.

Forecasting from dirty MRR is forecasting wrong. Start with numbers you can trust →