Go Narrow, Stock Deep: AI makes smarter holiday assortment bets possible

Conventional wisdom holds that more shopping choices equals more transactions equals more profits for retailers — in other words, breadth is king.

But ahead of this year’s holiday season, retail leaders would do well to consider a strategy better designed to withstand today’s uncertain spending environment: Going narrow and deep by buying more of the SKUs proven to turn fast and resonate with their customer segments.

The challenge, of course, is determining which bets to make, because in retail the details matter. That is especially true during the holidays, when demand is concentrated and heavily influenced by promotions. Historical data may be less reliable as trends and economic conditions shift from one season to the next; must-have products can emerge quickly, leaving past winners behind; and short selling windows allow little time to recover from a misstep.

What makes a good SKU hasn’t changed much in the past 25 years, but AI tools now enable organizations to fundamentally redefine it by empowering them to predict what the customer will want as opposed to what they’ve wanted in the past.

With the right KPIs, analytics, and benchmarks in place, AI can also help retailers get granular about what each individual SKU’s Adjusted Contribution Margin, or what it earns after every hidden cost — on trade, freight, promotions, warehousing — is stripped out.

Here’s what retail executives need to know to get started.

Why narrow and deep is the right strategy

The evidence is there. Consider Lowe’s strategy last year: on its Q3 2025 earnings call, the company announced that SKU rationalization drove a 50-basis-point increase in its Q3 gross margin and lower inventory levels.

Dollar General and Advance Auto Parts also claimed to be taking aim at unproductive SKUs of late, with the former’s CEO calling SKU reductions “a big win.” AI solutions can support these efforts, though more accurate — and efficient — demand planning, allocation, and replenishment are also a natural outgrowth of a narrow-and-deep assortment strategy.

More signals are pointing in this direction. The explosion of specialty and third-party marketplaces allow retailers to offer breadth without tying up their own capital in slow movers. For the SKUs you actually buy, warehouse, and mark down this holiday season, narrow and deep becomes the natural approach once you've offloaded risk to outside sellers.

Freeing up working capital by reducing spend on long-tail SKUs also enables retailers to be more flexible in chasing winners. Industry-standard carrying costs run to about 18–30% of average inventory value on-hand. This means that a brand sitting on $5 million of overstock for 12 months pays $900K–$1.5 million just to hold it. That’s $5 million that can't fund faster-turning categories, marketing, or growth. Case in point: back in 2023, Target’s operating profit jumped 237% after reducing bloated inventory levels.

AI breathes new confidence into SKU decisions

A narrow-and-deep buying strategy can be supported by the bevy of new AI tools now on the market. One analysis indicates that modern retail demand forecasting cuts inventory costs by 20–35% and prevents 65% of stockouts through AI-powered predictions. AI’s impact on replenishment and allocation will only further support these improvements.

Some retailers are getting even more creative. At Walmart, for instance, employees use predictive AI and machine learning models to factor in historical weather patterns and real-time data to help planners reposition inventory or reroute shipments before severe weather hits.

Again, these efforts are more successful with a narrow-and-deep approach in place: repositioning becomes much harder when you don’t have enough inventory depth (e.g., you have just enough to meet presentation minimums across your stores); by contrast, substantial depth on key items provides the flexibility to shift inventory according to AI-powered insights.

Crucially, AI only works when the underlying operating model supports it. Those who win with AI will therefore be the ones who are willing to change everything about the way they work — from organization structure and roles and responsibilities to processes, calendars, and more. If retailers use AI to support a legacy operating model, it will never realize its full potential.

The downstream costs of getting it wrong

Those who fail to make the most of this holiday season risk incurring major downstream costs in the form of heightened markdown exposure and thinner margins heading into Q1 2027.

For example, a unit costing $20 and meant to sell at $80 (60% margin) — sold instead at $40 — yields $20 of margin instead of $48; across 30% of production sold at half off, that's $1.5M–$2M in annual markdown cost on a $20M brand.

This happened to Levi Strauss a few years ago. When the company ended 2022 with inventories up by 58%, they were forced into heavy discounting. By the end of Q1 2023, they had cut inventory growth by about half, though this came at the expense of 360 basis points of gross margin erosion.

Moving beyond a ‘one-size-fits-all’ mindset

Retailers have to get specific not only with regards to the products they sell but inventory depth at the store level. For instance:

• Consider adjusting historical performance analyses to account for lost sales—and thereby allow higher-volume doors reach their full potential. AI-powered demand sensing tools can help.
• Adjust fulfillment strategies for high-volume doors that, as a result of added depth, may be pushing up against capacity (e.g., smaller quantities delivered more frequently).
• Reduce inventory in lower-volume stores to fund the additional inventory going to high-volume stores and minimize markdowns.

This mindset aligns with the winning strategy for retailers this holiday season: using AI to get specific and understand the right SKUs to invest in and where to hold them — instead of a broader hedging of bets.