Margin is not lost in one decision. It is lost in ten thousand
Module · Revenue, Pricing and Margin
The quote says one discount, the contract says another and the invoice matches neither.
Prophesee Commercial decomposes every basis point of movement into price, volume, mix, rate and currency, catches the leak at the transaction rather than in the quarterly review, and lets commercial finance test a pricing move before it is offered.
Margin is not lost in the pricing meeting. It is lost in the exceptions after it.
The list price is set carefully once a year. Then come the discounts, the rebates, the freight allowances, the deductions and the special terms, each defensible on its own and none of them visible together. By the time the erosion shows in a quarterly review, the contracts that caused it have another year to run.
Sources: World Commerce & Contracting with Deloitte, The ROI of Contracting Excellence, published 2023 (n=1,236 organisations, fieldwork Apr 2021 to Dec 2022, age disclosed) · Simon-Kucher, Global Pricing Study 2025, published 2025 (n=2,200+ business leaders across 28 countries, research by a pricing consultancy and identified as such) · Deloitte, CFO Signals Survey Q3 2025, published 15 Oct 2025 (n=200 North American CFOs at companies above $1bn revenue).
From explaining margin to defending it
Margin is reported by product and region because that is how the ledger is cut. Nobody can say which customer, contract or channel will be dilutive next quarter.
Gradient-boosted models project realised margin by customer, product, channel and contract, each with a confidence band and a ranked driver list. Dilution is visible while the contract can still be renegotiated rather than after it renews.
Each concession is approved on its own merits by email. Nothing looks at the accumulation, so a customer can drift outside policy one defensible step at a time.
Write the guardrail in plain English: this floor, this discount band, this rebate accrual, this freight allowance. Every transaction is tested as it happens, with peer group deviation catching the concession that is normal for the rep and abnormal for the customer.
A price move is argued on a spreadsheet with an assumed volume response. There is no no-action baseline, so the outcome can never be attributed to the decision.
Model the price move, the mix shift or the rebate redesign against a predicted no-action baseline, and watch the projected margin curve update. Then track actuals against that plan with counterfactual modelling separating the move from the market.
The commercial terms that govern the margin sit in signed documents nobody has read since signature, in language that does not match the ledger.
Contracts, price lists, rebate agreements and invoices resolved into one graph, with the clauses extracted by language models and tied to the transactions they govern. Ask what a customer is actually entitled to and get the clause, not an opinion.
Turning margin challenges into decisions
Margin fell one hundred and eighty basis points and four people spend three days arguing whether it was price, mix or currency.
Nobody agrees what it costs to serve this customer, so nobody can say whether the contract is worth renewing.
Each concession is defensible alone. The accumulation is invisible until someone runs a report a quarter later.
Revenue is forecast and margin is inferred from it, so the mix that determines the margin never enters the number.
A price move is argued with an assumed volume response and no baseline, so the result can never be attributed to it.
12 AI applications that could be relevant
A sample of what becomes possible on the decision layer, not a fixed list: each application draws on the same data foundation and audit trail, and new ones are configured on the engines, not built from scratch.
Every basis point of movement decomposed into price, volume, mix, rate and currency.
Margin projected by customer and product, with a band and the drivers behind it.
Shared cost carried to the customer, product and channel that actually consumed it.
The customers and contracts contributing most to erosion, ranked by size and trend.
Concessions outside band, policy or peer group flagged at the transaction.
Accruals and claims tested against the agreement that actually governs them.
Where the mix is shifting against plan, and what the shift is worth in basis points.
Model a price or mix move against a no-action baseline and watch the curve update.
Test a rebate structure on predicted behaviour before it is offered.
Actuals tracked against the modelled plan, separating the action from the market.
True profitability per contract, including freight, rebate, service and financing cost.
What a customer is actually entitled to, answered with the clause behind it.
A day when margin is defended, not explained
Today: A margin report cut by product and region, because that is how the ledger is cut.
One hundred and eighty basis points decomposed into price, mix, rate and currency, with the three customers driving most of it named.
Today: Ten concessions approved by email, each defensible, none visible together.
A discount normal for the rep and abnormal for the customer, flagged at the quote rather than at the quarter.
Today: Argue the price move on a spreadsheet with an assumed volume response.
A price increase tested against a rebate redesign on the projected curve. She takes the one that holds volume.
Today: Margin recovered. Nobody can say whether the pricing action did it.
Last quarter's price move tracked against the modelled plan, with the counterfactual separating it from the market.
Tomorrow's margin becomes today's decision.
Increase realised margin
We agree the metric and the baseline in week one, and measure the result on your data.