Retail AI loves discounts. Your CFO doesn’t

AI recommendation engines are one of retail’s biggest success stories. They personalize at scale, boost engagement, lift conversion, and make revenue charts trend in all the right directions. Leadership sees growth, teams celebrate adoption, and the algorithm gets more traffic.

Then the quarter closes.

Revenue is up, but profit is thinner.

Margins feel squeezed, but no one can quite point to why.

The AI “worked”… so where did the money go?

This is the quiet margin killer hiding inside many recommendation systems.

Most retail AI is trained to optimize what’s easy to measure: clicks, add-to-cart events, and conversion rates. The model learns fast. Discounted products convert better than full-price ones. Familiar SKUs outperform new launches. Low-margin bestsellers generate reliable engagement.

So the engine does what it’s told. It aggressively pushes discounted items, repeatedly surfaces the same high-velocity SKUs, and sidelines higher-margin alternatives that don’t convert as instantly. It doesn’t know, or care, that a slightly lower-converting product might be far more profitable. It ignores inventory holding costs, supplier funding rules, and the long-term impact of training customers to wait for discounts.

The result is a dangerous illusion of success.

Top-line revenue grows. Engagement metrics look healthy. But margin erodes underneath. High-margin products get cannibalized. Inventory ages in warehouses. Finance teams start asking why growth no longer translates into profit, and the AI team struggles to explain, because technically, the model is performing “better than ever.”

The failure here isn’t bad AI. It’s narrow optimization.

Recommendation engines are usually blind to business economics. They don’t understand contribution margin. They don’t see inventory decay. They don’t factor in supplier rebates, co-op funding, or fulfillment costs. Click-through rate becomes the north star, even though CTR has no concept of profit.

Fixing this doesn’t mean ripping out personalization. It means redefining success.

Retailers who course-correct move toward multi-objective optimization. Instead of ranking products purely by likelihood to convert, they balance relevance with margin, inventory age, and supplier economics. Constraints are baked directly into ranking models, minimum margin thresholds, discount caps, exposure limits for low-profit SKUs. The AI still personalizes, but within guardrails that protect the business.

Equally important, mature teams introduce regular profit sanity checks. Not just A/B tests on clicks, but deliberate reviews that ask: Are our AI recommendations making us money, or just activity? When recommendations drift toward margin-destructive behavior, they’re corrected before the P&L takes the hit.

This is why this topic resonates so strongly with executives. They already feel it. They see the contradiction between “AI success” and shrinking profitability. What’s been missing is a clear articulation of the problem, and a practical path to fix it without killing growth.

AI doesn’t ruin retail margins.

Optimizing the wrong objective does.

A large global apparel retailer deployed an AI recommendation engine across its website and email campaigns. Within months, conversion rates climbed sharply and online revenue hit record highs. Leadership initially viewed the rollout as a clear win.

However, gross margin dropped by nearly four percentage points year over year.

A detailed analysis revealed the issue: the AI had learned that discounted items converted far better than full-price products. As a result, it aggressively promoted sale items, even when full-margin alternatives were available and relevant. New seasonal collections were cannibalized, while older inventory with growing holding costs sat untouched.

The retailer recalibrated the system. Margin contribution was added as a core ranking signal, inventory age was factored into product prioritization, and supplier-funded promotions were weighted more heavily than retailer-funded discounts. They also instituted monthly reviews where merchandising and finance teams audited AI recommendations for profit impact.

Conversion dipped slightly, but gross margin recovered, inventory write-downs fell, and overall profitability improved. The AI didn’t become less effective. It became more aligned with the business.

#Retail #Ecommerce #AI #RetailTech #Profitability #DigitalTransformation #Merchandising