Why Did Gross Profit Change? Build a Margin Bridge Finance Can Review

Finance AI can help explain a change in gross profit when it reads a reconciled bridge, the method behind that bridge, and evidence for the proposed explanation. It should not choose an attribution method while writing the management commentary. A bridge that adds up establishes a numerical relationship; it does not, by itself, establish why the business changed.

The CFO's question has two parts

Imagine a management review in which revenue has grown but gross profit has fallen. The CFO asks whether the problem is pricing, product mix, input costs, or execution. A useful answer must first account for the movement between two accepted reporting balances. It must then distinguish what the calculations show from what the available evidence can explain.

Those tasks are related, but they require different checks. The controller may confirm that gross profit moved from one balance to another while FP&A is still investigating the commercial reasons. An AI summary that merges those states into one confident explanation can make unfinished analysis look settled.

The practical objective is a review package: the movement, its defined contributions, the records behind each explanation, and the questions that remain open. That gives the leadership team something it can challenge and act on through its normal decision process.

Start with the reporting contract

Before decomposing a movement, agree exactly what is being compared. Gross profit amount and gross margin percentage are different measures. A movement in currency cannot be described interchangeably as a movement in percentage points. The bridge in this article concerns gross profit amount.

The reporting contract should identify the two periods, the reporting currency, the entities and products included, the accepted source-report versions, and the treatment of restatements. It should also specify whether the comparison uses reported or constant exchange rates and how acquisitions, disposals, and reclassifications enter the analysis.

These choices affect the answer before any model is involved. If one period includes a newly acquired business and the other does not, a clean arithmetic bridge can still be an unhelpful comparison. If the opening balance has been restated, the reader needs to know which version is authoritative.

A practical contract also names the business owner and the person who accepts reconciliation. That ownership should be visible in the review process, not buried in an analyst's working notes.

Agree how contributions are attributed

Volume, price, mix, and cost are useful labels only when their calculation rules are explicit. Where price and quantity both change, there can be an interaction term. Different attribution methods can allocate that interaction differently while reconciling to the same total.

The implementation therefore needs an agreed method: calculation grain, baseline prices and costs, sequencing or interaction treatment, and treatment of new or discontinued products. It should state what belongs in identified adjustments and what remains an unexplained residual. A residual must not be renamed "mix" simply to make the bridge appear complete.

This is where finance expertise belongs. The platform can execute the chosen method consistently; it cannot decide which management interpretation the organization intends. Once agreed, the method should be versioned and tested against cases selected by Finance, including combinations that challenge its assumptions.

Reconciliation

Product mix remains a broader attribution question when more than one product is in scope. It is not a contribution in this single-product example. Finance can say which supplied line is largest. It cannot yet say that a supplier increase caused the decline. That explanation would require evidence linking the relevant cost movement to the supplier change, with appropriate treatment of timing, inventory, and other possible contributors.

Nor does the favorable volume contribution establish that every part of the business sold more units. Aggregated contributions can conceal offsetting movements. A useful review lets the reader inspect the relevant business slice without changing the agreed method behind the total.

Separate calculation evidence from causal evidence

A calculation record should identify the inputs, method version, scope, and reconciliation result. Evidence for an explanation serves a different purpose. It might include an approved price schedule, a product-mix breakdown, a purchasing record, or a documented accounting adjustment.

An attached document is not automatically support for a claim. The reviewer must assess whether it concerns the relevant products and period, whether it supports or contradicts the statement, and whether it accounts for the amount being discussed. A general supplier email is not necessarily evidence for a specific gross-profit contribution.

For the synthetic example, compare two sentences. "The bridge attributes the cost-rate line" is a statement about the supplied calculation. "A supplier price increase caused that line" is a causal claim. Without appropriate linked evidence, the second sentence should remain a question for investigation rather than become the executive headline. The purchase record in the example is input evidence. The business explanation is unconfirmed context.

The distinction should survive export into a slide, email, or briefing. A caveat hidden in a detail panel does not help if the copied summary removes it.

Build the review path around reconciliation

A proposed implementation starts with accepted finance reports, applies the agreed bridge method, and checks the result. If the opening balance plus contributions does not match the accepted closing balance within the agreed tolerance, the review path should stop at investigation. It should not manufacture an explanation for an unreconciled result.

After reconciliation, the interpretation layer can summarize defined contributions, retrieve permitted evidence, and draft questions. It should retain the calculation's units, scope, method version, and unresolved conditions. It may call an approved calculation tool where the architecture supports that; the boundary is authority over the method, not a blanket ban on executing queries.

Finance then reviews the proposed explanation. The recorded outcome might be accepted wording, a correction, a request for further analysis, or an explicitly unresolved cause. None of those outcomes should silently rewrite the underlying calculation contract.

Test the exceptions before the executive summary

Useful acceptance cases include a rounding difference, a missing product mapping, a duplicate source record, a restated opening balance, and a new product with no comparable baseline. Each case should produce a defined result or a visible exception, rather than disappear into an adjustment column.

Evidence tests should include a document from the wrong period, an unresolved reference, and a record that contradicts the suggested cause. Narrative review should check sign, currency, amount, period, and the difference between correlation and causation. Numerical accuracy alone is too narrow an acceptance criterion.

The review should also test access. A user who can see a divisional total may not be entitled to every underlying commercial document. A correctly calculated answer can still be inappropriate if its supporting narrative exposes information outside the reader's permissions.

How Orisdale would scope this work

For a margin-review engagement, Orisdale's proposed starting point is one recurring management question, one agreed reporting scope, and a small set of representative periods. The work would establish the bridge contract with Finance, map available source evidence, implement the bounded calculation and review path on the customer's chosen platforms, and verify it against agreed examples.

The useful deliverables are concrete: a definition of the calculation, reconciled example outputs, an exception register, an evidence-to-claim structure, acceptance checks, and named owners for ongoing changes. The scope should establish which of those artifacts already exists and which needs to be built.

That approach follows the delivery journey without assuming a ready-made product will resolve every accounting or data issue. A proposed implementation is separate from the current maturity of Orisdale's in-development Workflow Agent.

Make the handover useful after the first review

The calculation needs an operating owner as well as an initial approver. Specify how a changed product mapping, a revised cost allocation, or a reporting restatement reaches that owner. Identify which checks must run again and whether previous explanations need to be revisited. A new method version should not silently replace the basis of an earlier management decision.

The handover should let a finance analyst reproduce a selected movement and understand its limitations without asking the original developer to reconstruct the work. That is a stronger acceptance standard than a successful demonstration of one convenient period.

What to settle before proceeding

Before asking for a demonstration, ask whether Finance agrees the opening and closing balances, the attribution method, and the treatment of residuals. Identify who can approve an explanation and which evidence that person would expect to inspect. If those answers are disputed, resolving them is part of the work, not a reason to hide the disagreement behind a fluent interface.

The next step is to bring one accepted report and one difficult movement to a Finance Intelligence discussion. The same review boundary sits in the shared architecture. The existing executed synthetic variance example illustrates a narrower calculation-and-evidence boundary. It does not prove this margin bridge has been implemented, but it shows the kind of inspectable distinction a serious finance experience should preserve.