Why the Last Promotion in the Plan Earns the Least: The Case of Conagra

What did Conagra change in its promotions?

On 30 September 2026, reporting the first quarter of its fiscal 2027, Conagra Brands, an American maker of packaged foods, said it had "opted not to repeat certain merchandising events" on its Sandwich Bros brand, "which resulted in top line headwinds but improved profitability".¹ Merchandising events are the promotions a manufacturer pays a retailer to run: temporary price cuts, features in the retailer's advertising and displays in store. Asked on the call whether more such decisions would follow, the chief executive, John Brase, said the company would not "repeat promos that have been dilutive to the company", meaning promotions that lowered its profit. He added that the effect was already "embedded into the fiscal year forecast".²

Conagra added promotion money in the same quarter. Its merchandising spending rose across the company while particular events came out, and it raised advertising and promotion (A&P) spending by 15 percent.¹ A&P sits in its selling costs rather than in payments to retailers. It did not say that money moved from one brand to another.

Conagra dropped promotions that cost it profit while its merchandising and advertising spending rose in the same quarter, so its promotion money as a whole was not cut.

Why does the average return hide the weakest events?

The Academy's lesson on promotion investment allocation observes that almost every trade review reports one return on investment (ROI) per retailer. The lesson calls this the average return and sets it against the marginal return: "The average return tells you how the money you already spent performed. The marginal return tells you what the next dollar will earn."

In an illustrative worked example, a salty snack manufacturer spends 1,000,000 dollars on seven events at Retailer A. Each event's return is its net incremental profit divided by the money spent on it, so break-even is zero. Net incremental profit is the manufacturer's own margin on genuinely new units, counted on its net price rather than on the shelf price, less the discount paid on every unit sold on deal and the sales lost in the weeks after. The manufacturer decides which events it funds; the promoted shelf price is the retailer's decision.

Sorted by return, the first five events earn 60, 45, 33, 20 and 10 percent. Their budgets are 100,000, 120,000, 150,000, 150,000 and 130,000 dollars, 650,000 dollars in all. The last two lose 5 percent on 150,000 dollars and 12 percent on 200,000 dollars. Total profit is 175,000 dollars (60,000 plus 54,000 plus 49,500 plus 30,000 plus 13,000, less 7,500 and 24,000). The average return is 17.5 percent (175,000 divided by 1,000,000), which looks acceptable until the list is read from the bottom.

The lesson's return on the last slice of a budget is the change in profit divided by the change in spend. The top five events alone earn 206,500 dollars, so the last 350,000 dollars lost 31,500 dollars (175,000 less 206,500). Its return is minus 9.0 percent (31,500 divided by 350,000). A third of the budget lost money inside a plan that averaged 17.5 percent.

Kraft Heinz's published promotion returns are plan-wide readings. On 5 August 2026 it told investors that its "Return on Promotions" rose 3.4 points and its "% of Promotions with net positive ROI" 6.6 points between the first halves of 2025 and 2026.³ The closest measures in the Academy's lesson on the promotion performance grid are the portfolio return and the portfolio efficiency ratio, although Kraft Heinz's presentation does not say how it defines them. Neither reading says which event to drop next.

A plan's average return describes money already spent, so a manufacturer should sort its events by return and judge the plan by the bottom of the list.

Why do the last events in a plan earn the least?

The investment allocation lesson draws a plan as a response curve, profit against money spent, which "bends because you fund the best events first and the list gets worse". Each extra event at the same retailer also tends to win fewer genuinely new sales: more of its volume goes to shoppers who would have bought anyway, some is bought early and stored at home, and frequent deals can lower the price shoppers expect to pay.

A public Nielsen benchmark from 2014 shows the same pattern for the depth of discount. Trade efficiency is Nielsen's measure of trade return per dollar invested. Categories that used deep discounts (more than 25 percent off) on a small share of their events had the highest, 74.3 percent. Where every event was a deep discount, it was 37.0 percent, half as much (37.0 divided by 74.3 is 49.8 percent).⁴ The data compare categories rather than events within one plan, but the shape matches the curve.

Nielsen's improving categories cut their share of deep-discount events by 39.2 percent and gained 3.6 points of trade efficiency. The categories losing ground raised that share by 21.4 percent and lost 8.5 points. Both groups ran slightly more events (2.8 and 2.5 points more, on Nielsen's measure), so what separated them was the depth of discount: the winners ran shallower discounts more often without shrinking the calendar.⁴

In 2026 several manufacturers are adding promotion money, and each added event sits further down its plan's curve. Kraft Heinz told investors it was "Increasing investments in price and driving stronger returns on promotions",³ and General Mills said in July 2026 that shoppers were "buying more on promotion and less at everyday prices".⁵

The last events in a plan tend to earn the least, so a growing promotion budget makes the return on the next dollar more important.

Where should the money go, and when should it stop moving?

Dropping the two losing events would raise profit by 31,500 dollars and cost their volume at Retailer A. The better question is where the money would earn more.

A second retailer, Retailer B, has events nobody funded: a display that would cost 150,000 dollars and return 41 percent, a multibuy at 200,000 dollars and 26 percent, and a price reduction at 120,000 dollars and 8 percent. Retailer A has no better version of its plan on offer at the same spend.

Canceling the 200,000-dollar event at minus 12 percent and funding the multibuy gains 76,000 dollars (200,000 times 38 percent, the gap between 26 and minus 12). Canceling the 150,000-dollar event at minus 5 percent and funding the display gains 69,000 dollars (150,000 times 46 percent, the gap between 41 and minus 5). The same 1,000,000 dollars now earns 320,000 dollars (175,000 plus the two gains), an average return of 32.0 percent. The gain comes from moving the money, since no event was run better.

Before the moves the funding gap was 53 points: Retailer B's best unfunded event at 41 percent against Retailer A's weakest funded one at minus 12. After them, the best unfunded event is Retailer B's price reduction at 8 percent and the weakest funded one is Retailer A's 10 percent event. The gap is minus 2 points (8 less 10), so the plan is finished.

The cut-off also stops a manufacturer from dropping events just to raise the average. Dropping the 10 percent event as well would lift the average return to 35.3 percent (320,000 less 13,000, divided by the remaining 870,000) while profit would fall by 13,000 dollars. The event stays unless the money has another use with a measured return above 10 percent.

Money should move from the weakest funded event to the best unfunded one until no unfunded event earns more than any funded one. The cut-off stays at break-even until another use of the money has a measured return.

What if dropping an event costs share or the retailer's goodwill?

The strongest objection is that an event's return understates what it does: drop it and a rival may take the slot, the retailer may push back, and lower volume may raise unit costs in the factory. The method moves money rather than cutting it, and Nielsen's improving categories ran slightly more events while cutting deep discounts. Costs that are real belong in the return: if lost volume raises unit costs, that cost goes into the event's net profit and the event is ranked again.

The lesson allows an event below the cut-off on four conditions: "Name what it is buying, specifically", "Time-box it" (give it an end date), "Say what it must prove" and "Price it". "Building the relationship does not qualify."

Estimates are uncertain too. A 3 percent error in the baseline, the sales the product would have made without the event, can move a 50 percent return by about 18 points, the lesson shows. Retailer B's events are forecasts for events that have not run. If both were 18 points worse than estimated, the moves would still gain 82,000 dollars (200,000 times 20 percent plus 150,000 times 28 percent, the 38- and 46-point gaps each less 18).

For the retailer losing the events, the lesson's advice is to "Bring the events, not the total" and to "Offer the substitution before the reduction", meaning a better plan at the same spend, if one exists. In the example none does, so Retailer A's budget falls by 35 percent (350,000 of its 1,000,000 dollars), a cut the lesson would phase over more than one year where the relationship needs it.

Conagra has put the volume it gave up into its forecast for fiscal 2027. Such a decision rests on five steps: put a return on every event, sort the list, read the bottom, compare it with the best event nobody funded, and move the money while the gap is positive.

An event below the cut-off should be dropped or its money moved unless it buys something named, for a fixed time and at a stated price, and the volume it carried should be agreed in the forecast before it goes.

References