Use case Promotions & cannibalizationConvenience chain, 86 stores

How a convenience chain found the promotions that were quietly losing money.

Sixty-one promotions were live, and every one looked like a win on units. Cortex measured what each one really added after the discount and the sales it took from the shelf next to it.

The challenge

Every promotion looked like a win.

The weekly report showed units up on almost every deal. Supplier funding covered part of each discount, so the team assumed the rest was paying for itself.

Nobody had time to check what each deal took from the products beside it. And because signage stays up for three to six months, a losing promotion kept losing until the posters came down.

Before Cortex

What the team saw first

  • Promo unitsUp on 58 of 61 deals
  • Supplier fundingCovers part of every discount
  • CannibalizationNever measured
  • InstinctMore promotions, more traffic

Same question. Different results.

The question“Which promotions aren’t generating value, and which ones cannibalize the most?”

Claude / GPT / Public LLMs

Uploaded file

1 sourceThe file you uploadedpromo_sales.csv

Answer

58 of your 61 promotions increased unit sales. Chips 2-for-$4 had the biggest lift. Keep it running.

Just an answer.

Cortex AI

Connected intelligence

Cortex AI
  • Basket data
  • Cannibalization
  • Competitor shelf prices
  • Supplier funding
  • Price elasticity
  • Footfall
  • Weather
  • Category margins
  • Signage dates
  • Rebate programs
  • 23 deals losing money
  • Chips 2-for-$4: 58% cannibalized
  • 3 rebates never claimed

Answer

23 of 61 promotions lose money or need rework. Chips 2-for-$4 is the worst: 58% of its extra bags came from other chips. End 9, rework 11, add rebates to 3.

A decision.

The mechanic mattered more than the discount.

Cortex compared actual units with expected units for every promotion, store by store. Then it priced the discount, the extra margin, and what each deal took from neighboring products and pack sizes.

Multi-buys on snacks and energy drinks mostly moved shoppers between brands. Meal bundles and single-item price points brought new spend. Three losing deals had supplier rebates available that nobody had claimed.

What Cortex found

The promotions that need attention

PromotionNet value / monthCannibalizedCortex says
Chips 2-for-$4Salty snacks−$8,90058%End
Energy 2-for-$5Energy drinks−$6,20047%Rework: $2.99 single
Candy 3-for-$3Confectionery−$3,10044%Rework
Water 24-pack $4.99Water−$1,80012%Add supplier rebate
Cold brew + pastry $5.49Bundles+$4,7006%Keep
Coffee + breakfast sandwich $4.99Bundles+$6,3004%Keep

Six of 61 live promotions, sorted the way Cortex flags them: losing money first.

End the losers, fix the mechanics, claim the money on the table.

  1. Ended 9 promotions at the next signage change.
  2. Reworked 11 from multi-buys to single-item prices or meal bundles.
  3. Added supplier rebates to 3 deals, which turned them positive.

By the numbers

Live promotions23 of 61losing money or needing rework
Redirected$290Ka year, to deals that pay
Net promotion value+$38K / moup from −$24K a month
MeasureBeforeAfter
Promotions losing money234
Net promotion value, per month−$24,000+$38,000
Average cannibalization on multi-buys52%21%
Spend redirected to deals that work—$290K a year
Review of all live promotions2 days a monthEvery Monday, automatically
Leadership takeaway

Units aren’t value. Net promotion value is.

Every deal looked like a win because we only counted the units. Cortex counted what they cost.

A promotion is a bet that the extra margin beats the discount and the sales it takes. Now every bet gets checked.

How many of your live promotions are really paying?Book a demoSee all use cases
Next use caseHow a wholesaler found the dealers its own prices were pushing away
See Cortex on your data

Bring your hardest question.
Leave with a decision.

Target implementation2–4 weeks
EnvironmentPrivate enterprise
Timeline depends on data readiness and integrations.

See Cortex AI in Action

What would you like to explore?
We’ll only use your information to make your experience more valuable.