In the second quarter of 2025, the brand finance team at a top-twenty CPG company flagged that trade spend was running 18 points above the planned ratio. The CFO asked the trade marketing team to explain. The trade marketing team produced a sixty-slide presentation that pointed in several directions at once. The CFO asked for a number. There was no number. We were brought in three weeks later with one question: where is the money actually going, and is it doing anything.
Twenty-one business days later we had the answer. 73% of the unplanned variance came from a single retailer. Not the largest retailer. Not the highest-volume retailer. The third-largest customer, which had been quietly absorbing escalating trade investment across four programs without producing corresponding lift.
This is what the forensic looked like.
The setup
The client is a top-twenty CPG firm in food and beverage. Annual trade spend across the portfolio is roughly $180M against a planned ratio of 19% of gross revenue. In Q1 and Q2 of 2025, the actuals came in at 22.4% and 22.7% respectively. Over six months, that’s about $11M of unplanned trade investment with no corresponding plan or expected return.
The internal team had spent eight weeks trying to attribute the overage. The work product was a deck showing trade as a function of category dynamics, competitor pressure, retailer mix shifts, and promotional calendar density. The deck was directionally accurate. The deck did not produce a number, a recommendation, or a course of action. The CFO closed the deck and called us.
The brief
The brief was three sentences: tell us which retailers and which programs are responsible for the overage. Tell us which of those investments produced lift and which did not. Tell us what we should stop doing by next quarter.
What we did not get was clean data. Trade investment data at this client was spread across the trade marketing system (gross deal value at the program level), the syndicated data (volume and price impact at the SKU-week-retailer level), the deduction reconciliation system (what actually got paid against each program), and the field execution tool (compliance reporting at the store level). None of those four systems talked to each other in any reliable way. The first week was data assembly.
The decomposition
The trade variance was decomposed across three dimensions: retailer (8 major accounts), program type (display, TPR, EDLP funding, slotting), and product portfolio (5 segments). That produces 8 × 4 × 5 = 160 cells. We computed planned vs. actual spend in each cell.
What emerged when we ranked the cells by absolute variance was striking. The top 4 cells — out of 160 — accounted for 73% of the total unplanned trade investment. All four were the same retailer. All four were the same program type: conditional rebates against shelf compliance metrics, escalating across four product segments.
The mechanism
The retailer in question — call them Retailer 3 — had been operating under a master trade agreement that included escalating performance rebates tied to shelf-set compliance. The agreement said, in effect: if you maintain 90% facing compliance and 85% promotional execution rate across our store base, you earn back a portion of slotting and trade investment. The percentages tier upward.
Three things had happened over the previous four quarters, none of which the trade marketing team had connected.
Compliance reporting had improved. The client’s field team had rolled out an updated mobile compliance app in late 2024. The reporting rate — not the actual compliance — jumped about 11 percentage points in Q1 2025. Retailer 3’s contract treated reported compliance as if it were actual compliance.
The agreement had escalator clauses. Each compliance tier above 85% increased the rebate by 1.5 percentage points of program spend, up to 92%. Reported compliance hit 89.4% in Q1 and 91.1% in Q2. The escalators triggered.
No one was watching the integrated math. The compliance team owned the field app. The trade team owned the rebate agreements. The finance team owned the line item. None of the three saw the interaction. The agreement was paying out against a metric that had been artificially inflated by a reporting change, and the rebate line was rising mechanically each quarter.
The agreement was paying out against a metric that had been artificially inflated by a reporting change, and the rebate line was rising mechanically each quarter. No one had connected the three systems. The mechanism, in one sentence
Did the spend produce lift?
We constructed a synthetic control comparison using matched DMAs in non-Retailer-3 trade areas. The estimated lift attributable to the Retailer 3 incremental rebate spend — the part above what the agreement would have paid at pre-2025 reported compliance levels — was statistically indistinguishable from zero. 90% credible interval: −$0.4M to +$0.6M against $7.9M of incremental rebate investment.
This is the cleanest possible failure mode: trade investment was rising mechanically, with no underlying behavior change at the retailer or in-market that produced commercial value. The client was paying for a reporting artifact.
The fix
The recommendation had three components.
Short-term: pause the escalators. The trade team initiated a discussion with Retailer 3 to suspend the escalator tier until a third-party audit of actual compliance could be completed. This is a politically sensitive move — it implies the retailer was being paid for reporting that didn’t match reality — but the agreement had a provision for audit-triggered review, and the audit was already overdue.
Medium-term: renegotiate the metric. Replace self-reported facing compliance with a sample-based third-party audit, conducted quarterly, scaled to a tighter range. The new metric does not have the reporting-artifact problem and is less susceptible to the same mechanical escalator.
Long-term: integrate the systems. The trade rebate model, the compliance reporting, and the finance close all need to share a single source of truth on what compliance is and what it pays. The fact that three independent systems can each tell a coherent internal story while the integrated outcome makes no sense is a structural problem that recurs across the client’s top accounts.
The handoff
Three weeks of work. The deliverable was twelve pages, three SKUs of supporting analysis, and a recommendation that quantified the recoverable trade investment at $6.2M annualized, with a 90% credible interval of $5.1M to $7.4M. The actual recovery against pre-renegotiation run-rate over the next two quarters came in at $6.8M.
The pricing — or in this case, trade — committee did not need a transformation. They did not need a new tool. They did not need a quarterly process. They needed a number, the math behind the number, and a course of action that would survive the conversation with the customer. The work was the answer. The dashboard was a side effect.
What this generalizes to
Trade variance forensics is a specific instance of a general failure mode: an organization with three or more independent systems that each look internally consistent, but whose integrated behavior produces an outcome no one designed for. We see this pattern across pricing, trade, demand planning, and customer retention. The fix is rarely a new system. It is almost always a forensic that establishes the integrated math and a recommendation that pulls the right system into alignment with the others.
The 73% concentration in this case was extreme. The pattern — that the bulk of the unexplained variance lives in one or two accounts, and the rest of the variance is essentially noise — is the rule, not the exception. The first move on any commercial-variance question should be to decompose, rank by absolute variance, and look hard at the top of the list. The answer is usually there.