51 min read

Why Revenue Has Multiple Definitions (And What to Do Instead)

Why sales and finance report different revenue, the inclusion rules behind it, and how a metric contract makes one revenue definition reusable.

Why Revenue Has Multiple Definitions (And What to Do Instead)

Two revenue numbers sat in the same board deck last quarter, and both were defended. Sales showed the annual contract value it had booked. Finance showed the revenue the general ledger had recognized. Nobody in the room could say which one was wrong, because neither was. A revenue definition is the set of inclusion rules that decides what a number counts, when it counts it, and what it leaves out. Different teams apply different rules, so they produce different totals from the same business. This article sets out the rules that make two numbers diverge, what breaks when they do, and the acceptance checklist that turns one definition into something dashboards, applications, analysts, and agents can all reuse. The governance principle behind it is one metric, one definition.

Why can two teams report different revenue and both be right?

four panels showing the same annual contract as seen by the CRM, the billing system, the payments ledger, and the general

The two numbers answer different questions, and each is correct for its own. FASB's revenue standard (ASC 606) requires an entity to recognize revenue to depict the transfer of promised goods or services in an amount reflecting the consideration it expects to be entitled to, applied through five steps: identify the contract, identify the performance obligations, determine the transaction price, allocate it, and recognize on satisfaction (FASB, ASU 2014-09, 2014). That is recognized revenue, and it is a timing answer.

IFRS 15, the international twin, states the same core principle and five-step model and adds that a performance obligation may be satisfied at a point in time or over time (IASB, IFRS 15, 2014). The standard became effective for public entities for annual periods beginning after 15 December 2016, with early application not permitted for public entities (FASB, ASU 2014-09, 2014), and internationally for periods beginning on or after 1 January 2018 (IASB, IFRS 15, 2014).

Bookings, billings, and cash collected are different answers again. Take an illustrative annual contract of $120,000 signed in March, billed quarterly at $30,000, collected net 30, and recognized at $10,000 a month. The CRM records $120,000 booked, the billing system $30,000 billed, the payments ledger $30,000 collected, and the general ledger $10,000 recognized. The founder consequence: the sales dashboard and the finance report are not in conflict, they are measuring different events.

What inclusion rules actually differ between revenue numbers?

Four inclusion rules account for most divergence, and the arithmetic underneath them is usually sound. The first is gross versus net, which turns on the control test behind gross versus net: the entity asks whether it controls the specified good or service before transfer, and indicators of an agent position include another party being primarily responsible for fulfilment, little or no inventory or return risk, no pricing discretion, commission-based compensation, and no credit risk on the underlying good (IASB staff, principal versus agent considerations, 2024). A reseller that never holds control reports the commission; one that does reports the gross amount.

The second is refunds and credits. Variable consideration is estimated and constrained, included only when highly probable under IFRS 15 or probable under ASC 606 that no significant revenue reversal will occur, and reassessed each reporting date (IASB staff, post-implementation review of IFRS 15, 2024). A team that nets expected refunds and one that does not will disagree by design.

The third is intercompany elimination. Oracle's consolidation documentation explains that a sale from one unit to another is revenue for the seller and cost for the buyer, but not revenue for the group, so consolidated revenue equals the sum of unit revenue minus intercompany eliminations (Oracle, Intercompany Eliminations, 2026). The fourth is the time boundary: a point-in-time versus over-time judgment moves the same contract across periods. The two standards remain substantially converged on variable consideration, principal versus agent, and contract cost, so the gap is rarely the framework (IASB staff, post-implementation review of IFRS 15, 2024). It is the inclusion rules each team chose, which is why a valid query can still answer the wrong question.

What breaks when revenue definitions diverge?

Divergence stops being a reporting annoyance and becomes an operating failure, and the evidence is near-universal. In a 2026 survey of senior data leaders, 99% said defining metrics separately in each BI or analytics tool is a challenge, with 61% calling it moderately to extremely challenging; 42% cited inconsistent definitions across tools and 37% inconsistent definitions across teams as top obstacles, and 79% said consistency of metrics across platforms matters (Studio by Informa TechTarget for Strategy Inc., 2026, 2026; vendor-commissioned, 100 respondents).

What fractures metric consistency Inconsistent definitions across tools (42%) and across teams (37%) are cited as obstacles, behind too many data sources (49%). What fractures metric consistency. horizontal bar data: Too many data sources 49; Inconsistent defs (tools) 42; Lack of governance 38; Inconsistent defs (teams) 37.Source: Data, AI and Analytics Trends Across Organizations in 2026 2026. What fractures metric consistency Too many datasources 49 Inconsistentdefs (tools) 42 Lack ofgovernance 38 Inconsistentdefs (teams) 37 Source: Data, AI and Analytics Trends Across Organizations in 2026 (2026)

Trust erodes alongside it. Precisely's 2025 data integrity survey, run with Drexel LeBow, found 67% of data professionals do not completely trust the data they rely on for decisions, up from 55% in 2023, and 45% named inconsistent data definitions and formats as a top barrier to high-quality data (Precisely with Drexel LeBow, 2024; vendor-commissioned, 565+ respondents). The same 2026 survey found 39% cite data quality issues as a top AI-implementation challenge (Studio by Informa TechTarget for Strategy Inc., 2026, 2026).

The founder consequence lands in three places. Forecasts built on one number break when finance reports another. Fundraising conversations stall on definitional questions before they reach valuation. And an agent answering "what was revenue last quarter?" returns whichever definition its query happened to encode, which is why consistent definitions make AI answers testable. Finance teams still tie out CRM, billing, and general ledger by hand each close; no named survey in this evidence puts a percentage on that reconciliation effort.

What is a metric contract, and how do you make one revenue definition reusable?

a single definition box feeding four consumers — a dashboard, an embedded application, an analyst notebook, and an AI agent

A metric contract is a written definition with clauses: what the metric counts, what it excludes, its time grain, its owner, and its version. It works because the same clauses can be read by a person and executed by software. dbt's semantic model specification requires an aggregation time dimension, entities, and dimensions, while measures and simple metrics are optional (dbt Developer Hub, semantic models, 2026). How dbt defines a metric adds required name and type, type-specific parameters such as agg for simple metrics or numerator and denominator for ratio metrics, and a config block supporting meta, group, and tags, where owner and governance metadata live (dbt Developer Hub, creating metrics, 2026).

Cube's semantic layer centralizes metric definitions, joins, access rules, and caching upstream of every BI tool, application, and AI agent, and serves all of them from one model (Cube documentation, 2026). In Cube's data modeling overview, cubes define measures and dimensions while views expose a curated, governed subset that humans and agents query (Cube documentation, 2026). Both vendors describe the same promise, and the 2026 survey shows 99% of enterprises still struggle to realize it, so the contract is the discipline the tooling assumes.

The acceptance checklist is five steps:

  1. Name the question the metric answers, in one sentence.

  2. Write the inclusion rules: gross or net, refunds and credits treatment, entity scope, and time boundary.

  3. Assign one business owner accountable for the definition.

  4. Version it, with release notes and backward-compatibility rules, so a change is traceable and reversible.

  5. Test it against a second system and reconcile the difference before anyone reports the number.

A semantic layer is the natural home for that contract, which is what an enterprise semantic layer is built to provide.

How we apply this at Singdata, the publisher of this guide: our analytics agent writes and runs the query, retrieves governed metrics, and returns a trusted answer, so a revenue definition written as a contract can be executed rather than re-typed into each tool. It does not decide the clauses for you: which inclusion rules apply, who owns the metric, and when a version changes remain your team's calls, and the same contract discipline works with any semantic layer or BI tool, ours included.

Which revenue number should I show investors, and is ARR revenue?

Show recognized revenue as the primary figure and ARR beside it, labeled as an operating metric. ARR is not an officially recognized accounting metric and was not developed by FASB or any accounting board, and companies use interchangeable variants that make cross-company comparison unreliable without the definition (Ordway Labs, review of 125 public SaaS filings, 2022). Ordway's review found 53% of 125 listed SaaS, cloud, and fintech companies referenced ARR or a variant in their 10-K or 20-F, and 35 of those 66 reported it as a key business metric in MD&A (Ordway Labs, 2022).

Distrust in decision data, 2023 to 2024 The share of data professionals not completely trusting their decision data rose from 55% to 67%. Distrust in decision data, 2023 to 2024. line data: 2023 55; 2024 67.Source: Precisely / Drexel LeBow 2025 Data Integrity Trends 2024-09-18. Distrust in decision data, 2023 to 2024 0 25 50 75 100 55 2023 67 2024 Source: Precisely / Drexel LeBow 2025 Data Integrity Trends (2024-09-18)

ARR also carries two common meanings: annualized run rate, which multiplies MRR by 12 and includes monthly contracts, and annual recurring revenue, which counts only contracts of 12 months or longer and divides total contract value by years (ChartMogul, ARR reference, 2026). The same company can report two ARR figures a year apart in meaning and both be arithmetically correct.

The disclosure rules reinforce the ordering. The SEC's non-GAAP guidance requires a non-GAAP measure to be presented together with the most directly comparable GAAP measure with equal or greater prominence, and lists as improper presenting the non-GAAP measure first or omitting the GAAP measure, including in a release headline (SEC Division of Corporation Finance, Non-GAAP C&DIs, updated through 2021). PwC's count of S&P 500 filings found 97% use non-GAAP measures and KPIs when communicating results, so the practice is standard; the prominence rule governs how it is shown (PwC, Deconstructing non-GAAP, 2024). The founder consequence: lead with GAAP, state the ARR policy in a footnote, and never let the operating metric appear alone.

Conclusion

The two numbers in the board deck were never a sign that someone was careless. They were a sign that no one had written down the inclusion rules, assigned an owner, or versioned the definition, so each team built its own contract and defended it. The evidence says this is the default state rather than an exception: 99% of senior data leaders report difficulty defining metrics consistently across tools, and 67% of data professionals do not completely trust the data behind their decisions (Studio by Informa TechTarget for Strategy Inc., 2026, 2026; Precisely with Drexel LeBow, 2024).

The takeaway is that a revenue definition is a governance asset, not a formula. Once it has an owner, a version tag, and written inclusion rules, it can be reused by a dashboard, an embedded application, an analyst's notebook, and an AI agent without any of them reinterpreting it. The next step is small: run the five-step acceptance checklist on your own revenue number this week, and see which of the four inclusion rules your current definition leaves unstated. The reasoning behind shared meaning across systems is set out in why AI needs shared meaning.

FAQs

What is a revenue definition?

A revenue definition is the set of inclusion rules that determines what a revenue figure counts, when it counts it, and what it excludes. It covers gross versus net treatment, refunds and credits, entity scope such as intercompany elimination, and the time boundary between billed and recognized. Two teams can apply different rules to the same contracts and both produce arithmetically correct totals.

Why do sales and finance report different revenue numbers?

Sales typically reports bookings or annual contract value, which is the total value signed, while finance reports recognized revenue, which spreads that value across periods as performance obligations are satisfied. ASC 606 and IFRS 15 require recognition to depict the transfer of promised goods or services, so a $120,000 annual contract signed in March appears as $120,000 to sales and $10,000 a month to finance (FASB, ASU 2014-09, 2014).

Is ARR the same as revenue?

No. ARR is an operating metric, not an accounting figure, and it is not recognized by FASB or any accounting board (Ordway Labs, 2022). It also has two common definitions: annualized run rate, which includes monthly contracts, and annual recurring revenue, which counts only contracts of 12 months or longer (ChartMogul, 2026). Recognized revenue under GAAP or IFRS is the figure that appears in financial statements.

What is gross versus net revenue?

Gross versus net revenue depends on whether the company controls the good or service before it is transferred to the customer. A principal that controls the item reports the full amount; an agent that arranges for another party to provide it reports only its commission, judged against indicators such as fulfilment responsibility, inventory risk, pricing discretion, and credit risk (IASB staff, 2024).

How do I make one revenue definition reusable across tools and AI agents?

Write the definition as a contract with named inclusion rules, one accountable owner, and a version tag, then store it where every consumer reads from the same place. Semantic layers such as dbt and Cube define metrics in version-controlled files and serve them to BI tools, applications, and agents from one model (dbt Developer Hub, 2026; Cube documentation, 2026). Test the definition against a second system before publishing it.

Does a metric contract settle which revenue number is correct?

No. A contract makes the definition explicit, owned, and traceable; it does not resolve judgment calls such as whether a company is principal or agent in a given arrangement. Those remain accounting judgments made against the control indicators, and the contract records the conclusion and its version so a later change is visible rather than silent (IASB staff, 2024).