A pitch deck, a financial model and a data room can all carry the same revenue figure and still disagree about what it means. The reconciliation question has a direct answer: the three documents agree when a reader can move from the headline claim to the calculation behind it to a dated record that reproduces it, using the same as-of date, the same unit and period, and the same definition of revenue. This article shows how to trace one revenue assumption through all three documents, which dates, units, definitions and supporting files have to line up, how to tell an ordinary discrepancy from one that warrants escalation, and why a mismatch is a question rather than proof of wrongdoing. The method is source-backed cross-checking. Its output is not a verdict. It is a shorter, sharper list of questions.
What Does Consistency Actually Mean?
Consistency does not mean every document shows an identical number. A deck headline, a model driver and a ledger export are different kinds of objects serving different purposes. Consistency means a reader can move from the claim to the calculation to the evidence without hitting an unexplained jump.
Three things have to align. The as-of dates must match, or be labeled clearly enough to reconcile. The unit and period conventions must match. The definition of what counts as revenue must be the same. When those three conditions hold, the supporting file in the data room becomes checkable rather than decorative.
Treat the comparison as a check of the reasoning, not the formatting. A forecast may use a narrower metric than a deck headline, but the reader needs an explicit bridge between them before the forecast can be interpreted.
Which Revenue Assumption Should You Trace?
Do not audit everything. Choose one revenue assumption that appears in all three places and carries weight in the story: a current run rate, an annual recurring revenue figure, a top-customer contribution, or a renewal rate that drives the forecast.
Write it as a single sentence with four slots: value, unit, as-of date, definition. Filled from the hypothetical example later in this article, the reference line reads: $200,000, in US dollars per month, as of September 30, 2026, normalized subscription MRR excluding non-recurring implementation services. Every slot carries a real value, so any document that omits or changes one of them stands out immediately. If your own figure is not yet pinned down, use the same four slots as a fill-in template and complete each one before you open the documents.
Find the metric in the deck, the model and the supporting files, then note any difference in date, calculation or scope. Record the questions the available evidence does not answer. The time this takes depends on the records and the business.
Choosing a load-bearing figure is also a test of the story itself. Series A pitch deck guidance warns against cherry-picking the strongest numbers and recommends figures that give a complete picture of the economics. Tracing one figure end to end, with its full context, is a practical way to check that you are doing that.
Do the Dates Refer to the Same Thing?
A figure without a date cannot be compared at all, so start here. A deck slide stating current ARR makes a point-in-time claim: it describes one moment. A model line labeled FY2026 revenue makes a period claim: it covers a range of time. Neither is wrong, and they are not the same claim. Each needs a label that says which one it is.
Ask four questions. What is the as-of date behind the headline figure? What period does the model line cover? When were the underlying records cut, as distinct from when the file was exported or printed? Do any of the figures cover a period that has not closed yet?
An export date and an as-of date are different things. A file exported three weeks after a headline was written can still describe the same period, and a file exported the same day can describe a later one. The export date tells you when the file was produced. The as-of date tells you what period the records actually describe. Ask for the second, and record it next to the figure.
A forecast period sitting beside historical figures is normal. A forecast period presented without a label next to historical figures slows the review down, because the reader cannot tell which claim is being made.
Do the Units and Periods Match?
Units fail in predictable ways: dollars versus thousands, a monthly figure annualized by multiplying by twelve versus a trailing twelve-month total, currency converted in one document and not another, and booked or contracted amounts set beside revenue recorded in the accounts.
Multiplying one strong month by twelve produces an annualized figure for that month. It is a unit and period decision, not a growth result. If the deck annualizes and the model does not, the two numbers will diverge even when each is internally correct.
Write the unit next to every figure you compare: currency, scale, period, and whether the figure is actual, annualized or forecast. Some apparent mismatches disappear once the units are written out. The ones that remain are worth asking about.
Does Revenue Mean the Same Thing in All Three Documents?
Revenue labels carry rules, and the rules are the company's to state. Ask which measure each figure represents: an amount recorded in the accounting records, a recurring-revenue metric such as ARR or MRR, a booked or contracted amount, an invoice, or cash received. The review is not applying an accounting standard from the outside. It is establishing that the three documents are using the same answer.
Ask the definition question directly: what is inside this number, and what is outside it? Then check whether the deck, the model and the data room use the same answer. Ask for the company's written rules on subscriptions, services, discounts, taxes and unsigned renewals, because a recurring-revenue figure without stated rules cannot be reproduced by anyone else.
If a headline labeled recurring revenue includes non-recurring services while the model excludes them, reconcile the calculation and correct the metric or the label. Then check whether the forecast still starts from the intended base. The difference alone does not tell you whether the problem is limited to presentation.
Evidence should be specific enough to be checked. A definition that cannot be tied to a policy document, a contract term or an invoice line is an assertion, not yet a definition.
A Clearly Labeled Hypothetical Example
The following is a hypothetical example created for this article. It does not describe any real company.
As of September 30, 2026, a fictional company has normalized subscription MRR of $200,000, so its current subscription ARR is $2.4 million. For this simplified example, September subscription revenue is also $200,000, and the company earns a separate $50,000 in non-recurring implementation services that month. The deck incorrectly labels the $250,000 total multiplied by 12 as ARR of $3.0 million. The model starts from $200,000 MRR and forecasts subscription ARR of $4.0 million as of December 31, 2026. The data room contains a dated active-subscription schedule and contracts supporting the $200,000 MRR, plus separate billing and accounting records for subscriptions and services.
The bridge is explicit: $200,000 × 12 = $2.4 million of current subscription ARR. Annualizing the additional $50,000 from services contributes $600,000 to the arithmetic total, but it does not make that amount recurring or predict future service work. The resulting $3.0 million is an annualization of September's total revenue, not ARR. The $4.0 million figure is a later forecast and must be labeled separately. Comparing that forecast with current ARR implies about 66.7% growth in the recurring run rate over those three months; it is not a measure of year-over-year recognized revenue growth.
Ask the company to reconcile the $200,000 subscription base to the dated contract schedule, separate the services amount, and correct the deck's ARR label. The mismatch calls for an explanation. It does not, by itself, establish why the error arose or whether anyone intended to mislead.
What Should the Data Room Prove?
For the chosen metric, gather a dated schedule of the underlying records, the written calculation rules, relevant contracts or order forms, and the billing and accounting records needed to reconcile the inputs. The necessary files depend on the metric and the business; this focused cross-check is not a complete revenue-recognition audit.
The test is recomputation. A reviewer should be able to take the supporting files, apply the stated definition, and land on the figure in the deck or the model. If they land somewhere else, the gap needs a written explanation, not a verbal one.
Collecting the files early is a practical advantage. Carta's Series A pitch deck guidance notes that interested investors ask for supporting materials, most often a cap table, a financial model and access to a data room, and that organizing these before pitching saves time and lowers the risk of a deal falling apart during diligence. Start with the documents needed for the selected assumption and expand the review when the evidence raises further questions.
A Discrepancy Is Not Proof of Wrongdoing
A discrepancy is an observation: two documents do not line up. Reading intent or misconduct into that observation requires evidence the mismatch itself does not contain. The narrower point is the useful one. A mismatch alone does not establish misconduct, and treating it as though it does ends a productive conversation before the records have been read.
Use four working labels: timing difference, metric-definition difference, transcription or update error, and unresolved. Write down the evidence for each explanation, and keep an item unresolved when the records do not settle it.
These labels organize the investigation; they do not settle its outcome. A single important error can merit follow-up, while repeated mismatches still need evidence before conclusions are drawn about their cause.
Make the explanation traceable: identify the observed inputs, separate them from estimates, and show how the forecast uses each one. That gives a reviewer specific points to verify.
When Should an Unresolved Discrepancy Be Escalated?
Escalation is a decision about how much weight a figure can carry, not a judgment about anyone's intent. Three questions decide it, and each is judged on its own.
Significance: does the gap change a number a reader would rely on? A gap that changes the starting point of a forecast matters, and so does an unsupported historical revenue or traction claim, even when the forecast does not depend on it. A headline figure a reader would repeat is load-bearing whether or not any model line references it. A gap confined to a slide caption or a formatting detail carries less weight. Judge the figure itself first, then note separately whether the forecast also rests on it.
Supporting records: is there a dated record that settles it, such as a subscription schedule, a contract, an invoice register or a written definition, or is the proposed explanation unverifiable with what the data room currently holds? Here the distinction between two outcomes matters. When the records show the numbers agree once the definitions are stated, you have a documented reconciliation: nothing was wrong; the bridge was missing. When a record shows a figure was wrong and the figure is changed, you have a correction. An explanation backed by a record is one of those two. An explanation with no record behind it stays open.
Effect on the forecast: does the model use the disputed figure as its base? If the base is corrected, does the forecast move, and by how much? This question changes the action, not the significance. A discrepancy can be material to the review because of what the headline claims while leaving the forecast untouched.
The next action follows from the three answers. When a record reconciles the gap and the forecast base is unaffected, note the reconciliation and continue. When a record shows a correction and the base changes, ask for the revised model line and re-run the trace on the new figure. When a record shows a correction but the forecast does not depend on it, correct the figure in the deck and keep the corrected claim in view, because the historical or traction statement still has to be right. When no record settles it, keep the item labeled unresolved, state whether the forecast depends on it, and put it in front of a human reviewer before it is used in any decision. That last step is a decision to withhold reliance. It is not a finding about intent, and it should be written up that way.
What Questions Does Source-Backed Cross-Checking Produce?
Source-backed cross-checking does not produce a score. It produces better questions, and better questions are what move a review forward.
Instead of asking how fast the company is growing, ask which revenue definition is used to calculate the growth rate and whether the deck headline uses the same one.
Instead of asking whether the forecast is realistic, ask which line in the model drives it, what date that line was last updated, and which data room file supports the base period.
Instead of asking whether the traction is real, ask for the invoice register behind the stated figure and whether services, taxes and unsigned renewals are included.
Instead of asking only why two documents differ, identify the proposed explanation and the record that would confirm it. Keep the question open if the evidence is incomplete.
Explain how the revenue assumption feeds burn and runway rather than relying on a confident headline. A reviewer should be able to follow the calculation and see the remaining uncertainties. Seed-stage fundraising guidance makes the same point from the founder's side: what matters is whether the founder can separate facts from assumptions, and documents that contradict each other are one of the mistakes that stall outreach.
A Practical Checklist Before You Send Anything
1. Write one sentence containing the chosen assumption's value, unit, as-of date and definition. Evidence that closes it: your own reference line, agreed before you open the documents.
2. Locate that assumption in the deck, the model and the data room, and mark every place the four slots differ. Evidence that closes it: the three documents side by side, each showing its own date.
3. Label every figure as actual, annualized or forecast, and mark each date as a point in time or a date range. Evidence that closes it: the label printed in the source document, not your inference from context.
4. Gather the written definition and the dated schedule, plus the contract, billing and accounting records needed to reproduce the metric. Evidence that closes it: recomputation, where applying the stated definition to the files lands on the figure in the deck or model.
5. Classify each mismatch as timing, definition, error or unresolved, and write one line of explanation for anything not resolved. Evidence that closes it: the record you cite, or an explicit note that no record settles it yet.
6. Run the three escalation questions on every unresolved item and record the resulting action. Evidence that closes it: a written note naming the significance of the figure itself, the missing record, and separately whether the forecast depends on it.
7. Keep dated master versions of the deck and the model. After editing either, re-check whether the relevant assumptions and supporting files still reconcile. Evidence that closes it: the version dates on the files themselves.
8. Send the explanation with the documents. A gap that arrives with its own note reads as diligence. The same gap discovered later reads as doubt.
The action to take now is small and specific: pick one revenue figure that appears in all three documents, run the four-slot trace on it, and send the deck, the model and the supporting files together with a one-page note listing anything still unresolved.
Where Structured Analysis Fits
This kind of tracing is manual, and it is worth doing manually at least once, because following one number through three documents teaches a founder where the company's own story is thin.
Structured analysis tools, including DDScore, are useful for organizing what a review finds: which gaps are presentation problems, which are substantive, and which need evidence before anyone can judge them. They support the reviewer's judgment rather than replacing it.
DDScore provides analytical tooling and scoring to support due diligence. It does not provide investment advice, guarantee returns or replace a full review, and no output from any tool should be read as a finding of misconduct.
Explore DDScore for Founders · Explore DDScore for Investors