Why sales and finance report different revenue, and how to find out which number is right

Sales says $14.2M, finance says $12.8M, and both are right. The gap comes from five rules nobody wrote down: timing, grain, scope, adjustments and currency. Here is how to find which one is producing your number, and why a fourth dashboard makes it worse.

Sales reports revenue of $14.2M. Finance reports $12.8M. Both numbers are correct, and that is the part that usually gets missed in the meeting. Neither team made an arithmetic error, and neither is using a broken report. They applied different rules to different source systems, and each rule is defensible on its own terms.

Which is why the next dashboard does not settle it. A third team builds a third report, applies a third set of rules, and now the board has three numbers instead of two.

The four places a revenue number comes from

Four systems, four defensible revenue numbers

None of these is wrong. Each answers a different question, and each books the same deal on a different day.

A deal signed on 29 September sits in a September CRM number, an October billing number, a four-quarter GL number and a November bank number. Same deal.

Source: Thinklytics engagement pattern across the 11 semantic layer and metric governance engagements in the case library.

Most companies above a few hundred staff have four systems that can each answer the question "what was revenue last quarter", and they are not meant to agree. The CRM records what the revenue team committed. The billing platform records what was invoiced at contract terms. The general ledger records what is recognisable under the accounting policy. The bank records what was collected.

A deal signed on 29 September sits inside a September CRM figure, an October billing figure, a four-quarter ledger figure and a November bank figure. Same deal, four correct answers, four different periods.

That structure has always existed. What changed recently is how many rules now sit on top of it.

Why the gap widened in 2026

Why the gap widened in 2026

Finance leaders reporting on their own revenue operations. Each added model and channel adds another booking rule.

• Plan to add more consumption-based products
• Sell through more channels than a year ago
• Now manage multiple revenue models
• Complete at least one in five workflows manually

Source: Salesforce double-blind survey, fielded 4 to 15 May 2026, n=865 CFOs, chief accounting officers, EVPs of finance and accounting and VPs of finance across France, Germany, Japan, the UK and the US.

Salesforce ran a double-blind survey between 4 and 15 May 2026 across 865 CFOs, chief accounting officers, EVPs of finance and accounting and VPs of finance in France, Germany, Japan, the UK and the US. It found 69% now managing multiple revenue models, 73% selling through more channels than a year earlier, and an average of three channels in use. Ninety per cent said their company plans to add more consumption-based products.

Every one of those additions carries its own booking rule. A usage overage charge, a reseller margin, a multi-year contract with a mid-term upgrade and a consumption commitment that was only partly drawn all have to be assigned to a period, a customer and a line. Each assignment is a decision, and in most companies those decisions were made separately inside each system by whoever needed an answer that week.

The same survey found 60% of finance teams completing at least one in five workflows manually, which is where the rules get re-decided by hand each cycle rather than looked up.

It is not a reporting problem, and that is why reporting work keeps failing

Intuit's Future of Finance 2026 report, fielded in May 2026 by CatalystMR across 2,000 US CFOs, controllers and VPs of Finance at businesses above $2.5M revenue, found 70% saying their critical business data is scattered across systems, spreadsheets, dashboards and static reports with no single source of truth, and 51% of weekly finance time going to manual work including reconciliation, exports, error-fixing and report stitching.

It is a vendor-commissioned, US-only, self-reported survey and should be read that way. The direction matches the independent work: EY's 2026 Global DNA of the CFO survey, fielded between 16 February and 30 March 2026 across 1,610 finance leaders at organisations above $1B revenue across 28 countries, found 47% of CFO capacity going to operational tasks and only 27% saying finance is perceived as a strategic value partner.

The reason more reporting does not help is that the disagreement is upstream of the report. Both teams are already producing a technically correct figure. More reporting capacity produces more technically correct figures.

The five rules that are actually in dispute

The five rules that produce the gap

Work down this list with both teams in the room. The gap is almost always one or two of these, not all five.

Write the answer to each question down once, name an owner for it, and the two numbers converge without anyone rebuilding a pipeline.

Source: Thinklytics engagement pattern across the 11 semantic layer and metric governance engagements in the case library.

When we sit two teams down with a disputed number, the gap is almost never all five of these. It is one or two, and finding which one takes an afternoon rather than a project.

Timing. On what date does a deal count? Close date, invoice date, or spread across the recognition period. This is the single most common cause and the easiest to test, because the gap moves when you change the period boundary.

Grain. One row per what? Per contract, per subscription line, per customer, per legal entity. A customer with three subscriptions is one row or three, and the choice changes every per-customer average downstream.

Scope. What is in and what is out. Professional services, one-time fees, usage overage, reseller margin and intercompany lines are each either revenue or not, and the two teams have usually made opposite calls on at least one of them.

Adjustments. Credits, refunds and discounts get subtracted somewhere. If one system applies a discount at the line and another applies it at the invoice, the totals differ without either being wrong.

Currency. Deal-date rate, month-end rate or budget rate. For a company with meaningful non-domestic revenue this alone can account for the whole gap.

What the gap costs while it is open

The direct cost is reconciliation labour, and it is measurable. APQC's general ledger reconciliation benchmark, reported by CFO.com in September 2024, puts the median at six hours with the 25th percentile at five and the 75th at 10. CFO.com notes the slower end tracks the number of systems and connectors involved, and that liability and revenue accounts take longer because entries have to be traced back to source documents.

At Kaiser Permanente the reconciliation labour attached to 14 regional definitions of one metric ran $2.1M a year, and two internal attempts to resolve it had already failed. See the Kaiser Permanente metric governance engagement.

The larger cost is the decision that did not get made. The Intuit survey found 57% of finance leaders had missed a time-sensitive strategic action in the previous six months because financial visibility arrived too late, and only 14% had same-day data available for their most recent major business decision. Reconciliation hours are a finance-team cost and easy to dismiss. A decision that arrived late is a company-level cost.

How to find out which number is right

The question is the wrong shape, and reframing it is most of the work. Neither figure is the company's number, because the company has not chosen a rule. The CRM figure is right for quota and capacity planning. The ledger figure is right for the audited accounts. Both are wrong as an answer to "how are we doing", because that question has never been defined.

So the test is not which number wins. It is this: take one disputed metric and one period, have each team write its rule down before looking at any data, then run both rules against the same source table. If the two figures now agree, the plumbing was never the problem and the five rules above are where the budget belongs. If they still disagree, you have a second problem underneath the first, which is the subject of fix your KPI definitions or rebuild your data pipelines.

What we would do first

Take the last board pack. For each revenue figure in it, ask one question: can a named person state the definition and the source without opening a spreadsheet. Count how many pass.

That ratio is more useful than any maturity assessment, because it tells you exactly how much of the pack is currently defensible and how much is a number somebody assembled and nobody owns. Then pick the single figure with the largest gap between versions and run the half-day test on it. One resolved definition is worth more than a governance framework nobody reads, and it is the evidence you need to fund the rest.

The sequence after that is covered in the metric definition problem, the enforcement layer in semantic layer engineering, and the duplicate-customer version of the same problem in master data management. If the dispute is really about which records refer to the same entity rather than which rule applies, data cleaning and preparation is the starting point instead. The full set of work in this area sits under we cannot trust the numbers.

Frequently asked questions

Why do sales and finance report different revenue?

Because each team computes a defensible figure from a different source system under a different set of rules. Sales reads the CRM, which counts what the revenue team committed and books it on the day the opportunity is marked closed won. Finance reads the general ledger, which counts what is recognisable under the accounting policy and spreads it across the service period net of credits. Neither has made an arithmetic error. The disagreement is in five rules that were never written down: on what date a deal counts, what one row represents, what is in and out of scope, what gets subtracted and when, and at what exchange rate.

Which number is right, the CRM number or the GL number?

Neither, and that is the useful answer. Asking which is right assumes one of the two rule sets is the company's rule, when in most cases no rule has been chosen. The CRM number is right for forecasting capacity and quota. The GL number is right for the audited accounts. The question that moves things forward is not which figure wins but which rule the company wants for the question being asked, written down once, with a named owner.

Is this a data quality problem?

Usually not, which is why data quality work often fails to fix it. A data quality problem means the same rule applied to the same source gives an unreliable answer. Here, two different rules applied to two different sources give two reliable answers. The test takes half a day: have each team write down its rule before looking at the data, then run both rules against the same source table. If the numbers now match, nothing is broken in the plumbing.

Will a new dashboard fix it?

No, and it reliably makes things worse. A third team building a third report applies a third set of rules and produces a third number, so the board now has three versions instead of two. In our work this pattern is consistent: tooling does not create metric disagreement, it publishes it. One client had 220 store reports resting on only 40 distinct metrics, and cutting to 14 certified dashboards is what resolved it, not report number 221.

How big are these gaps in practice?

Large enough to reach boards. At a growth-stage SaaS platform, five competing ARR definitions produced a $2.1M discrepancy that the board had noticed, and the CFO could not answer which figure was correct. At an enterprise SaaS company, six revenue metrics produced a $1.4M gap on a $22M base. At Kaiser Permanente, 14 regional definitions of a patient encounter cost $2.1M a year in reconciliation labour on their own. Two internal attempts to fix it had already failed.

How long does it take to resolve?

In the case library, the engagements that resolved a disputed metric by writing the definitions down and certifying them on the platform already in place ran 8 to 20 weeks, and the duration tracked the number of contested metrics rather than the size of the company. Five ARR definitions took 8 weeks. Fourteen regional definitions took 11. Forty-two metrics across eight administrative units took 20.

Who should own the revenue definition?

One named person, not a committee and not a function. Finance normally owns the recognised revenue definition because it already carries the audit obligation for it, but the definitions for pipeline, bookings and committed revenue are usually better owned in revenue operations. What matters more than the choice is that the name is written next to the definition, and that a standing forum exists with the authority to settle the next dispute. Without that forum the answer reverts to whoever argues hardest in the meeting.

Does this get worse with AI?

Yes, and faster. A natural-language assistant pointed at three definitions of revenue will answer the same question three ways, with more fluency and more apparent confidence than the dashboards had. The disagreement is upstream of the interface, so a better interface distributes it rather than resolving it. EY's 2026 Global DNA of the CFO survey, fielded between 16 February and 30 March 2026 across 1,610 finance leaders at organisations above $1B revenue, found 61% citing data quality and bias as the top barrier to AI investment.

The work behind this

Eleven engagements in the case library carry semantic layer and metric governance, and three more carry master data management. They range from five ARR definitions reduced to one certified figure in eight weeks to 14 regional definitions of a patient encounter consolidated across a health system.

Every one names the client where we are permitted to and states the measured outcome: semantic layer and metric governance, 11 engagements.