The CFO of a PE-backed SaaS company should agree written definitions of ARR, net revenue retention, gross margin and customer with the finance team, and put them where every reporting and AI tool reads them, before any AI assistant answers a board member's question about those numbers.
AFP, the Association for Financial Professionals, published its 2025 FP&A Benchmarking Survey on January 14, 2025, twenty months before this post, from 362 FP&A and finance practitioners surveyed in fall 2024. Spreadsheets were the dominant tool, with 96% of survey respondents using them for planning, 93% using them for reporting purposes on a daily or weekly basis. AFP's release gives no industry breakdown, only organizations of varying sizes worldwide, though the spreadsheet figures hold across company size, geography and ownership type.
BCG, a consultancy that sells AI work, published The Widening AI Value Gap in September 2025, from a survey of 1,250 CxOs and senior executives who are AI decision makers across nine industries and 68 countries, sorted into four AI maturity stages. Technology, media and telecommunications is one of the nine industries surveyed. Self-reported, covering only the business area respondents know best, and subject to perception bias, it finds that more than 50% of future-built firms operate on a single enterprise-wide data model, compared with just about 4% of their stagnating peers, future-built being its top stage and stagnating its bottom one.
The CFO writes the four definitions first, one entry each, naming the rule, the system and field it is computed from, the owner, and the effective date. The ARR entry decides whether monthly contracts are annualized, whether a signed but unbilled contract counts, and on which date a cancelled customer leaves; the net revenue retention entry names the cohort and the period; the gross margin entry says whether hosting, third-party software inside the product and customer success sit in cost of revenue; and the customer entry says whether a customer is a billing account or a parent legal entity, and which ID joins it across the billing system and the CRM. At a company with two products, the customer entry might read: a parent legal entity with at least one active subscription on either product, joined across billing and CRM by one customer ID; owner, the controller; effective, the first day of the next quarter.
With ninety days, the CFO, the chief revenue officer and the head of customer success sign the four entries in the first month, and the CFO decides any split. In the second, the data team loads them into the semantic model, one of the metadata management tools that enable organizations to define key metrics and standardize the terms that describe their data, and the board pack, the BI tool and the assistant all read that model, and the assistant reads it alone. This assumes a data team and a BI tool the smallest portfolio companies may lack. No source prices either, or says which budget carries it. In the third, the head of FP&A asks the assistant every question from the last two board meetings and compares each answer with the board pack; a disagreement goes back to the entry, and the CFO signs the changed one. A company that already runs an assistant runs the same three months, with its board answers logged from day one and board questions answered from the board pack until the comparison is done.
Prior periods keep the old definition since the change applies only from the date recorded, so the series is not comparable across the effective date. No source prices what a restatement would cost.
