Most SaaS forecast models are really bookings models with a recurring revenue label attached. They project new business carefully and treat the existing base as a fixed block that carries forward. That works until contraction shows up, at which point the model has no line item to put it in. An ARR forecast model fixes this by making every movement in the revenue base explicit and forcing the whole thing to reconcile every month.
What is an ARR forecast model?
A monthly waterfall that starts with beginning ARR, accounts for every increase and decrease, and lands on ending ARR.The reconciling property is what makes it valuable. Ending ARR for one month becomes beginning ARR for the next, so there is no place for revenue to appear or disappear without a line explaining it. When the waterfall does not tie, something in the underlying data is wrong, and you find out in the month it happened rather than at the end of the year.
That structure also produces gross and net revenue retention as outputs rather than as separately calculated metrics that never quite match the revenue numbers.
What lines does the ARR waterfall need?
Eight, in this order.| Line | Direction | Definition |
|---|---|---|
| Beginning ARR | Base | Prior month ending ARR |
| Churned customer ARR | Contraction | Full customer departures |
| Churned product ARR | Contraction | Customer retained, product dropped |
| Product decrease ARR | Contraction | Same product, lower value |
| New customer ARR | Expansion | Logos that did not exist last month |
| New product ARR | Expansion | Existing customer buys an additional product |
| Increased product ARR | Expansion | Same product, higher value |
| Ending ARR | Base | The sum of everything above |
The same applies to expansion. New product ARR indicates cross-sell working. Increased product ARR indicates seat or usage growth, which is a different motion with different predictability. Seat growth is more forecastable than cross-sell because it follows customer headcount rather than a sales cycle.
How do you forecast the new customer line?
Feed it from the sales pipeline forecast, converted to annualized recurring value at the expected close month.The connection point between the sales forecast and the ARR model is here, and it needs two adjustments most teams skip.
First, convert bookings to ARR correctly. A three-year contract booked at $300,000 total is $100,000 of ARR, and multi-year deals with escalators need the first-year value, not the average. Getting this wrong inflates the ARR base and makes retention rates look worse than they are in the following year.
Second, time it by close month rather than by quarter. An ARR model that drops a whole quarter's new business into month one overstates the base for two months and distorts every retention rate calculated against it. Within a quarter, month three consistently produces more closed business than months one and two, so a flat monthly allocation is wrong in a predictable direction.
For the pipeline side of this input, see how to forecast revenue.
How do you forecast churn and contraction?
Cohort rates from your own history, overlaid with the actual renewal calendar for the period.A blended company churn rate applied evenly across months will be wrong every month. Renewals cluster, and a month carrying three large renewals has a different risk profile than a month carrying twenty small ones. Start by laying out contracted renewal dates and the ARR attached to each, then apply cohort-level rates by segment and tenure.
Then bring in leading signals rather than waiting for the renewal date. Support case volume is one of the more useful ones and it runs in both directions. Customers with no support cases at all carry elevated churn risk, because silence usually means nobody is using the product. Customers with seven or more cases in a year also carry elevated risk. The safer zone is three to five cases, typically lower severity, which indicates an engaged customer getting help and staying.
Contraction deserves its own forecast rather than being folded into churn. Product decrease ARR often precedes full churn by a renewal cycle, so a rising contraction line is an early warning about next year's churn line.
What does the model produce beyond ending ARR?
Gross revenue retention and net revenue retention, calculated directly from the waterfall lines.Gross revenue retention takes beginning ARR minus all three contraction lines, divided by beginning ARR. It never exceeds 100 percent and it measures how much of the base you keep before any expansion is counted.
Net revenue retention adds the three expansion lines from existing customers, excluding new customer ARR, then divides by beginning ARR. Excluding new logos is the rule people break most often, and including them produces a number that looks excellent and means nothing. See net revenue retention for the full calculation and its edge cases.
Both rates come out of the same waterfall that produces ending ARR, which means they always tie to the revenue number. Metrics calculated in a separate model drift from the revenue they describe.
How do you validate the model each month?
Close the prior month, tie ending ARR to the general ledger view of recurring revenue, and record the variance on every line.The reconciliation check is simple: beginning ARR plus expansion lines minus contraction lines must equal ending ARR, and ending ARR must equal what finance shows for recurring revenue at that date. A gap means an opportunity was miscategorized, a multi-year deal was annualized incorrectly, or a contraction was recorded in the wrong month.
Track line-level variance, not only total variance. A model that hits total ARR while overstating new customer ARR and understating churn is not accurate, it is lucky, and the offsetting errors will separate at some point. Measuring forecast accuracy line by line is what turns the waterfall from a reporting artifact into a model you can forecast the next four quarters with.
Frequently Asked Questions
What is an ARR forecast model?
An ARR forecast model projects annual recurring revenue forward by tracking every movement in and out of the revenue base. It starts with beginning ARR, adds new customer and expansion ARR, subtracts churn and contraction, and produces ending ARR. Ending ARR becomes the next period's beginning ARR, so the model reconciles month over month.
What components does an ARR waterfall need?
Eight lines: beginning ARR, churned customer ARR, churned product ARR, product decrease ARR, new customer ARR, new product ARR, increased product ARR, and ending ARR. Separating customer churn from product churn matters because a customer dropping one module is a different problem from a customer leaving entirely.
Should ARR be forecast monthly or quarterly?
Monthly. Quarterly buckets hide the timing of churn events and make mid-quarter contraction invisible until it has already compounded. A monthly waterfall also produces cleaner retention rates, since the beginning balance is closer to the movements being measured.
How do you forecast churn in an ARR model?
Use cohort-based rates from your own history rather than a blended company average, then adjust for known renewal dates in the period. Support ticket volume is a useful leading signal. Accounts with no support cases at all and accounts with seven or more cases in a year both carry elevated churn risk, while three to five non-severe cases usually indicates an engaged customer.
How does an ARR model connect to the sales forecast?
New customer ARR and new product ARR come from the sales pipeline forecast. The retention and expansion lines come from the customer base. Keeping them in one waterfall prevents the common failure where a team celebrates a strong bookings quarter that net retention quietly cancels out.
See how ORM turns these insights into action
ORM builds custom revenue forecast models for B2B SaaS companies. Not dashboards. Prescriptive analytics that tell you what to do next.
Schedule a Demo