What Does a Promotion Cost? Trade Spend and Deal Mechanics

A promotion costs the brand whatever it pays the retailer to run the event, and how it pays (off-invoice, bill-back, scan-back, or an accrual fund) determines who carries the risk that the money never reaches the shopper. That cost does not appear anywhere in the syndicated file. It lives in the brand's own records, and without it the return layer of promo analysis cannot be computed.

POI's working definition is the plain one: trade spending is the money a manufacturer spends with a retailer to encourage promotion through a discount or feature-and-display support and to drive volume [4]. POI also separates working trade, which funds events, from non-working trade such as new-item and introductory allowances [4]. Slotting fees for shelf space belong in that second bucket and stay out of event-level return math.

The four ways a brand pays

The names vary by company and by broker, but the mechanics reduce to a short list. The first two rows carry the strongest external evidence; the rest is standard practice described in plain terms.

Deal type	What triggers payment	Who carries the risk	Watch for
Off-invoice	A discount on every case the retailer buys during the deal window	The brand. Payment is made whether or not the shelf price moved	Forward buying: the retailer loads up at deal cost and sells later at full price [2]
Scan-back	A reimbursement per unit scanned at the promoted price, verified after the fact [1]	The retailer. No shopper discount, no payment	Retailers prefer off-invoice when the terms look identical; brands prefer scan-back [2]
Bill-back	An invoice from the retailer after the event, on proof of performance (the ad ran, the display was built)	Shared. The brand pays for activity, not for sell-through	Deductions that arrive months later and never reconcile to a specific event
Accrual or MDF fund	A per-case or percentage accrual that builds a fund the retailer draws on for events it chooses	The brand, over time. The fund is spent whether or not any single event paid back	Funds spent on events nobody analyzed

The scan-back mechanism matters because it is the one deal structure where the trade dollar and the shopper's discount are tied to the same unit. The pass-through study describes it directly: transactions are audited and the retailer is reimbursed for every unit sold at a promotional price [1]. Drèze and Bell's paper on scan-backs argues that manufacturers often lose money on off-invoice deals because retailers forward-buy, that scan-backs pay on units sold rather than units bought, and that a scan-back can be designed so the retailer is no worse off and the manufacturer is strictly better off; in the beverage category they studied, scan-backs did not trigger excess ordering and produced higher retail sales through lower prices [2]. Those are the authors' findings in one category, not a law, but they explain why the deal type belongs in the analysis as well as in the contract.

Pass-through: how much of the dollar reaches the shelf

Pass-through is the share of a manufacturer's trade discount that reaches the shopper as a lower shelf price. It is the single most important number the syndicated file cannot tell you directly, and the largest published study of it is sobering.

Nijs and colleagues assembled two years of price and shipment data through the entire channel for more than 100 products in one CPG category, sold through over 1,000 grocery and drug stores in more than 30 states, with Nielsen providing the store data [1]. Their mean pass-through elasticities were 0.71 from manufacturer to wholesaler, 0.59 from retailer to shopper, and 0.41 for the channel as a whole. In plain terms, a 10% reduction in the manufacturer's price reached the shopper as a 4.1% reduction on average [1]. Expressed as a rate rather than an elasticity, when the wholesaler offered a $1 discount, retailers cut the shelf price by $0.69 on average and kept 31 cents [1].

The averages are less useful than the spread around them. The authors found large variance at every level of the channel and give the example of a retail chain whose pass-through elasticity might be 98% in California and 42% in Nevada, concluding that aggregate estimates are of limited tactical value to a manufacturer [1]. Your account-level number is the only one that matters, and it is estimable from your own data: compare the wholesale discount you offered to the promoted price the syndicated file recorded.

Two of their findings bear directly on how you structure deals. Trade deal frequency lowered pass-through: the more often a product was on deal, the less of each discount reached the shopper [1]. And featured or displayed goods received less pass-through, which the authors read as retailers recovering the cost of the support from the trade dollar [1]. A brand that runs frequent deals with heavy support should expect a lower share of its money to reach the shelf, not a higher one.

The paper also documents the dispute this chapter opened with. A Nielsen survey it cites found retailers reporting that roughly 15% of trade deals goes straight to their bottom line, while manufacturers believed the figure was upward of 30% [1]. The study's own estimate, 31 cents on the dollar at the retail step, lands on the manufacturers' side of that argument.

The 80 to 90% assumption

When Nielsen built the return framework for its 2015 benchmark, it had no cost data for 76 million event weeks, so it applied an industry-standard assumption: manufacturers pay 80% to 90% of the shelf discount, with standard costs for cost of goods, feature, and display [3]. That assumption was reasonable for a cross-industry benchmark. It is not reasonable for your recap, because you have the actual number, or someone at your company does. Chapter 5 returns to this point: a return figure built on an assumed funding rate is an estimate and should be labeled as one.

Where the cost actually lives

Trade cost is not in the SPINS, Circana, or Nielsen extract. The syndicated file records the retailer's shelf price and the provider's estimate of base and incremental volume. What you paid for the event sits in one or more of your own systems: the deduction ledger where the retailer's bill-backs land, the broker's commission and deal statements, the TPM export if you run one (POI reports 79% of its surveyed manufacturers do [4]), or a spreadsheet the sales lead keeps.

That is the practical reason promo analysis stalls at lift for so many brands. The lift is in one file; the cost is in another, in a different shape, keyed to a different calendar. Suppose the syndicated data shows a Sprouts TPR sold 3,000 incremental units. Only your deduction ledger can say it cost $4,200. Both numbers are illustrative; the shape of the problem is not.

Sous connects internal files alongside syndicated data for exactly this reason: a deduction export or broker statement loads next to the SPINS or Circana file, and the return math in Chapter 5 becomes a query rather than a reconciliation project. The rule that follows is worth stating for any tool and any analyst. Without the cost file you can compute lift. You cannot compute return, and a system that produces one anyway has assumed a number it did not have.

Trade spend at the scale POI reports, above 15% of revenue for nearly 68% of manufacturers in its 2026 survey [5], deserves better than an assumption.

Common questions

What is the difference between off-invoice, bill-back, and scan-back? Off-invoice discounts the cases the retailer buys, whether or not the shelf price moves. Bill-back pays after the event on proof of performance. Scan-back pays per unit scanned at the promoted price, so the trade dollar follows the shopper.

What is forward buying? A retailer buying more cases than it needs during an off-invoice deal, at the deal cost, and selling them later at full price. The brand funds a discount the shopper never sees.

What is pass-through in trade promotion? The share of a manufacturer's trade discount that reaches the shopper as a lower shelf price. In the largest published study, a 10% manufacturer cut became a 4.1% shelf cut on average, with wide variance by retailer.

Does promoting more often help pass-through? The evidence points the other way. In the same study, higher deal frequency was associated with lower pass-through.

Where do I find what a promotion cost? In your own records: deduction ledger, broker statements, TPM export. It is not in the syndicated file.

Can I estimate return without the cost file? Only as an estimate, using a funding-rate assumption like the 80% to 90% Nielsen used for its benchmark. Label it as such, and replace it with the actual cost the moment you have it.