How to Build an ARR Waterfall That Reconciles Every Month

An ARR waterfall is a monthly table that starts with beginning ARR, adds new, expansion and win-back ARR, takes out contraction and churn, and lands on ending ARR. Beginning ARR is always last month's ending ARR, so the table has to tie to the cent. Gross and net retention come straight off it. Updated October 2026: sign for each move, a month that ties to the cent, error checks, and NRR and GRR math.

At ORM we call this chart Gross and Net Retention, and we run it by month. It turns a pile of contract changes into a clear record of why revenue moved. Teams that run it every month can defend each retention number they report. Teams that skip it are guessing. Some call it an ARR bridge, and the ARR waterfall vs ARR bridge entry covers the naming. This guide covers the lines, the rules for each move, a worked month, the checks, and the ratios.

What lines belong in an ARR waterfall?

Two balances and five kinds of move, each with a fixed sign.

ORM splits the moves into eight lines, run by month. The table maps the common names to those lines:

The critical rule sits in the first line. Beginning ARR equals the prior month's ending ARR with no exceptions and no restatements. The moment you allow beginning ARR to be recalculated from the contract table, the series stops being a reconciliation and becomes twelve unconnected snapshots.

The split between churned customer, churned product, and product decrease is what makes the schedule diagnostic. Three very different failures produce three different lines, and each one points at a different owner.

For who owns each line, see the ARR waterfall that reconciles.

How do you classify each movement?

Classify by what changed for the account, and apply the rule the same way every month.

Work through the decision in order:

1. Did the account cancel entirely? Churned customer ARR. The full contracted ARR moves out. 2. Did the account drop a product but keep others? Churned product ARR, valued at the dropped product's ARR. 3. Did the account reduce quantity or tier on a product it kept? Product decrease ARR, valued at the reduction only. 4. Was this the account's first contract ever? New customer ARR, at full contracted value. 5. Did an existing account add a product it did not have? New product ARR. 6. Did an existing account increase quantity or tier on an existing product? Increased product ARR. 7. Did an account that churned in an earlier month come back? Reactivation ARR, at the full value of the new contract.

Here is an illustrative month. It starts at $8,412,386.40.

The month ties to the cent. Gains total $441,618.00 and losses total $165,365.28. Add the gains to beginning ARR, take away the losses, and you land on $8,688,639.12. Gross revenue retention for the month is $8,247,021.12 over $8,412,386.40, or 98.03 percent. Net revenue retention is $8,404,618.32 over $8,412,386.40, or 99.91 percent. Ending ARR grew 3.28 percent on new logos and one win-back, while the base you started with shrank. The waterfall is the only view that shows all three facts at once. To check one period of your own, enter the movement lines in the ARR calculator.

What checks catch errors in an ARR waterfall?

Run six checks every month before anyone reads the retention rates. Each one catches a different kind of error. The count tie is quick to run. In the example month, 412 customers to start, plus 9 new and 1 win-back, minus 6 churned, gives 416 at month end (illustrative).

The test a board applies to a forecast applies here too. A number on one slide has to match the next. It has to roll up from region to total. And it has to drill down to the accounts that produced it. A waterfall that passes all six checks meets that test.

Which ratios does the ARR waterfall produce?

Four ratios come straight off the lines: GRR, NRR, net new ARR and the quick ratio. New customer ARR and reactivation stay out of both retention rates. Neither account was in beginning ARR, and counting them lets strong sales hide a retention problem. GRR can never pass 100 percent. NRR can. Some teams add reactivation to the top of the SaaS quick ratio; if you do, say so next to the number.

These are monthly rates, and they compound. If every month looked like this one, a 98.03 percent monthly GRR would come to about 78.8 percent for the year, since 0.9803 to the twelfth power is 0.788. For context, Benchmarkit puts median yearly GRR at 88 percent and median NRR at 101 percent in its 2025 SaaS benchmarks. For the full method, see how to calculate monthly NRR.

How do you turn the table into an ARR waterfall chart?

Draw the first and last bars from zero, and float each move from where the last one ended.

Gains step up and losses step down. Keep the bar order the same every month, so a reader can compare one month to the next at a glance. Put GRR and NRR on the same chart, which is how ORM shows it. Excel has a built-in waterfall chart type that handles the floating bars, and how to make a waterfall chart in Excel walks through it with a free template. For a board deck, fold the lines into a four-bar bridge and keep the full monthly table in the appendix.

Which edge cases break the tie-out?

Backdated amendments, one-time fees, and mid-month changes, in that order of frequency. Late renewals are normal. In ORM's data, renewals tend to close within about 30 days either side of the renewal date. Booking churn and then a win-back for a renewal that lands two weeks late inflates both lines. A short grace window avoids that, as long as it is the same every month. The entry on when to count a customer as churned covers the date choices.

Never use a plug line. A balancing entry hides one misclassification in month one and eleven by December, and every retention rate derived from that series inherits the error.

If the month does not tie, see ARR waterfall reconciliation for the usual causes.

Why does the waterfall matter more than the retention rates it produces?

Because the rates are outputs and the classification is the actual work.

Gross and net retention are two divisions performed on the waterfall. If the categories are right, both rates are right and comparable across every period. If the categories are inconsistent, the rates are precise numbers built on shifting definitions, which is more dangerous than no number at all because it invites confident decisions.

Consistency also beats perfection in the underlying data. Most revenue teams believe their CRM data is uniquely bad and that this prevents accurate measurement. It rarely does. As long as the classification is applied consistently, imperfect records still support accurate prediction. What breaks prediction is changing the rules between periods.

How does the waterfall feed forecasting?

It gives you three separate empirical rates instead of one growth assumption.

Once you have twelve or more months of clean waterfall, you can model churn, contraction, and expansion as independent series with their own seasonality rather than as a single net retention percentage. That matters because they behave differently. Contraction lands mid-term without a renewal trigger. Cancellation clusters at renewal dates. Expansion follows product adoption curves.

Seasonality shows up in all three. Q2 and Q4 typically run stronger than Q1 and Q3, and the third month of a quarter runs stronger than the first two. A model that assumes flat monthly movement will consistently miss the shape of the year even when the annual total is close.

For how the retention side joins new business in a single number, see how to forecast revenue and net revenue retention. For the discipline that keeps the forecast itself honest, see forecast accuracy.