ARR means three things in the same company
Sales, finance and the board each use a version, none of them written down, and the differences only surface when an outsider reconciles them.
Software
Almost every software business we see has more than one definition of recurring revenue, and discovers this during a fundraise or a sale. Billing holds contracts, the product holds usage, and the ledger holds recognised revenue, on three different bases. We agree the definitions, write them down and build the reporting on top, so ARR, net revenue retention and cost to serve survive somebody else checking them.
Discuss your operationWhat we usually find
Sales, finance and the board each use a version, none of them written down, and the differences only surface when an outsider reconciles them.
A single churn percentage hides which cohorts, plans and segments are actually leaving, which is the only version that suggests what to do.
Infrastructure, support and success costs are pooled at company level, so gross margin by plan or customer size is guesswork.
Where we start
Scope depends on the state of your systems. These are the pieces that recur in this sector.
ARR, bookings and recognised revenue defined once, written down, and reconciled to the ledger so the three can be moved between.
Gross and net revenue retention by cohort, plan and segment, including the expansion that offsets churn.
Infrastructure, support and success cost attributed to plan and customer size, so gross margin is measured rather than assumed.
The reporting an acquirer or investor will ask for, built now rather than reconstructed under time pressure later.
What changes
You own all of it: the code, the written definitions and documentation aimed at whoever maintains this after us.
The builds behind it
Your finance, customer and operational records joined into one dataset, with each figure defined once and traceable back to the system it came from.
Reporting built on top of the joined data, aimed at the few revenue and cost drivers that actually change the result.
One defined task, automated inside a process that already exists, measured against whatever it replaced.
Driver-based models built on the same definitions as your reporting, so the forecast and the actuals stop disagreeing.
Worth asking