AI trade promotion analytics needs a shared record

The post-promotion review is booked for an hour, and the first twenty minutes go to deciding whose number is right. Sales brought a lift figure from retailer point-of-sale data for the feature period. Finance brought spend from deductions that arrived after the display ended, measured against an accrual booked on shipments. Both describe the same promotion at the same account, and the meeting has to decide whether to run it again in the fall.

AI trade promotion analytics starts with a shared promotion record. The plan, shipments, sell-through, and spend must resolve to the same account, products, promotion window, and mechanic. Once those inputs agree, AI can classify deductions, compare outcomes, and answer questions about the review. Before they agree, a model may select one conflicting input without identifying the conflict.

The source records that have to resolve

The plan lives where it was made: a workbook by account, sometimes a promotion module in the ERP, often with an email thread that amended both. It holds the mechanic, dates, products, and committed money. It may not hold a stable identifier, so two labels can describe the same feature only because one person remembers the connection.

Shipments sit in the ERP and carry your product codes, account structure, and invoiced money. They can run ahead of the feature, so a shipment total limited to the shelf dates can omit the pre-build. A total cut at the wrong account level can also mix a banner agreement with a division agreement.

Point-of-sale data provides sell-through, meaning units purchased by shoppers rather than units shipped into the channel. It names products by retailer item number and follows the retailer calendar. Distributor depletion reports, which record product leaving the distributor for its customers, can arrive later and without store detail, so they support a coarser measurement grain than direct retailer data.

Spend arrives through accruals, invoice reductions, and deductions. Off-invoice spend reduces the invoice when product ships, while a billback is claimed after the agreed activity occurs. Those records need the same promotion identifier as the plan and volume records or return on investment remains partly unmatched.

Five fields in a promotion identity

The shared record needs five fields, each tied to a distinct measurement failure.

1. The account at the level the money was agreed, which is the banner for some retailers and one division for others, because spend committed at the banner cannot be split across divisions afterwards without inventing the split.
2. The products at pack grain, so a club pack and a grocery pack of the same product are not averaged into one lift number by a case conversion applied at the wrong level.
3. Two windows rather than one: the weeks the promotion ran at the shelf, and the earlier weeks the promotional volume shipped in.
4. The mechanic and spend basis, because invoice reductions and later deductions affect timing differently even when both fund the same promotion.
5. The baseline, meaning the estimate of what would have sold without the promotion, labeled as a modeled number with its method recorded beside it.

The five fields are spread across the ERP, retailer portals, distributor files, and planning workbook. When a person rebuilds those joins for each review, the same question can produce a different answer after a code, calendar, or assumption changes. A durable record makes those changes visible instead of hiding them inside a new lookup table. Trade spend management software covers the separate question of which tool category should own the planning and deduction workflow.

AI can classify deductions and compare promotions

Given a joined history, AI can propose which promotion agreement a free-text deduction reference belongs to and route uncertain matches for review. It can scan account, pack, and mechanic combinations for promotions that ran below plan. It can also answer plain-language questions during the review and draft a calendar variant for the team to evaluate.

AI is strong at classification, comparison, and explanation once the inputs are governed. A person still approves uncertain matches, chooses the baseline method, and commits the next calendar after negotiating with the buyer.

How Permute supports promotion measurement

Permute is the context layer we sell between promotion data and the AI and reporting tools that use it. We connect the ERP, retailer and distributor files, and the spreadsheets holding the calendar and fund tracker, then resolve their accounts and products to the same promotion record.

Promotion window, incremental volume, and net spend are defined once as explicit rules with effective dates. Incremental volume is the units attributed to the promotion above the recorded baseline. A change to a mechanic in March therefore applies from March without rewriting the first quarter.

A trade manager can ask which promotions ran below plan and receive the result with its shipment, sell-through, and deduction rows. The same definitions feed the scheduled review and dashboard. We sit underneath Claude, ChatGPT, and Copilot so they reason over governed promotion data rather than replacing those assistants. What a context layer actually does explains that mechanism.

Where Permute and AI stop

We read from connected systems and do not write back, so an approved calendar is still committed in the planning suite and an accrual entry is still raised in the ERP. We are not a forecasting engine. We preserve the baseline definition and its inputs, while the counterfactual remains a modeled estimate with a named owner.

Unmatched spend must remain visible as unmatched, and missing account files must appear as incomplete coverage. Neither Permute nor an AI assistant should convert those gaps into a complete ROI figure. How the data is handled explains how evidence and account access remain scoped.

Borrowed volume also requires human judgment. A promotion can pull demand forward, causing lift during the feature and a decline afterward. The measurement window should include the shipment pre-build, shelf period, and an agreed payback period, with the exact rule recorded by retailer rather than rediscovered in each analysis.

Before you trust a promotion ROI number

Take one promotion that went badly and one remittance into the next review, and put the number through five tests. Anything that survives them can be quoted to a buyer.

• Show this deduction matched to the promotion that caused it, in the account reference format, with no manual mapping step in between.
• Show the lift for a four-week feature whose volume shipped in the three weeks before it, and show which weeks that volume was attributed to.
• Show the same promotion ROI to a trade manager and to finance, then point at where net spend and incremental volume are written down.
• Restate one retailer month after the review was presented, then show which reported figures moved and which held.
• Show the shipment, depletion and deduction rows behind one lift figure, by account and pack, without an export.

Run the tests against the same stored promotion record. If a restatement changes the result, the review should show which source changed, when it changed, and which published figures moved with it.

Promotion analysis starts with agreement

The review can now return to its original decision: whether the promotion should run again. A defensible answer identifies the same promotion across its plan, volume, and spend records, applies a recorded baseline and measurement window, and exposes unresolved gaps. AI can accelerate classification and analysis after that agreement exists. The account owner still decides what to commit next.

Questions trade and finance teams ask about promotion ROI