In 2026, American retail is undergoing one of its most significant transformations in decades: according to public data, more than 2,000 stores and restaurants will close across the country. And this isn’t limited to the mass-market — the luxury segment is affected too. Dozens of Saks Fifth Avenue locations — one of America’s key luxury department stores — have already been caught up in closure and optimization programs. Even the owners of global fashion houses are acknowledging the need to rethink their strategy: Kering, which owns Gucci, has openly discussed reducing its store network.
What drives these decisions — a crisis, or something less visible to the average shopper — we asked Anna Kistner, one of the leading experts in luxury retail analytics. In her strategic decision-making, Anna draws on her proprietary methodology, the Predictive Buying Intelligence Platform (PBIP) — a data-driven system built on demand forecasting and data analysis in luxury retail. The methodology is already in use by a number of major industry players, including Galeries Lafayette, Majid Al Futtaim, and Versace, and is currently being evaluated by Silicon Valley technology companies. The framework has been covered by trade publications Elle China, Retail News Asia, and Retail Tech Innovation Hub, and in February 2026 it was recognized with the international Eurasian Entrepreneurship Award for its contribution to the digital transformation of the industry and improved accuracy in buying decisions.
In this interview, Anna Kistner discusses what actually drives brand success today — and why the modern buyer is no longer simply someone who tracks trends, but an analyst with deep technical and data expertise.
— Anna, you’ve spent more than 15 years at the intersection of buying, analytics, and strategy in luxury retail. In your view, what’s driving the wave of brick-and-mortar closures in 2026?
I see a fundamental structural mismatch between the traditional retail model and modern consumer behavior. This isn’t simply a shift from physical to digital commerce — it’s a crisis in the underlying economic model on which retail was built.
The core disconnect, as I see it, is time. Retail long operated on a planning horizon of several months: buying decisions were made well in advance, based on past sales and the buyer’s experience. But consumer behavior has changed. Today, demand is born in real time — through TikTok, influencers, the news cycle — and can disappear just as quickly. As a result, products often arrive in stores after the peak of interest has already passed, which inevitably leads to markdowns and margin erosion.
This is what makes the classic brick-and-mortar model economically fragile. When I see that roughly 30% of assortment ends up marked down due to forecasting errors — a figure I’ve validated through industry research — it becomes clear that fixed costs in real estate, staffing, and inventory can no longer be justified. The math of the traditional model has simply stopped working.
— Does that mean physical retail will disappear entirely?
What I observe is that the brands closing stores — even large, long-established ones — are those that failed to build an omnichannel system. In plain terms, that means a seamless connection between online and offline, where the website, social media, and physical stores operate as one: demand and inventory data stays synchronized, and customers can move fluidly from browsing online to buying in-store.
I predict the future trajectory is not the elimination of physical retail, but its transformation into intelligent, algorithmically curated spaces — smaller footprints, precisely stocked based on predictive demand rather than historical pattern recognition. This is what I am building at Saks now.
— Currently you analyze data across more than 1,000 brands and 70 retail locations — effectively managing one of the most comprehensive analytics maps of luxury retail in the US. Which brands are performing well today, and which aren’t — and what makes the difference?
I want to emphasize that brand performance isn’t a static quality — it’s a dynamic system shaped by many factors. The industry often oversimplifies this by talking about “hot brands,” but that framing doesn’t explain the actual reasons behind their success.
In practice, the highest-performing brands consistently show a few defining characteristics: strong conversion from social platforms to sales — particularly from TikTok and Instagram — controlled distribution with limited availability and strict pricing discipline, and targeted seasonal marketing investments that amplify demand at precisely the right moments.
I observe underperforming brands typically exhibit:
– Over-distribution across channels — availability diluting perceived scarcity and pricing power;
– Absence of social signal monitoring — reliance on creative direction without algorithmic trend detection;
– Slow supply chain responsiveness — inability to react to demand acceleration within commercially viable timeframes.
— Not only American retailers, but leading players from around the world — Europe, the Middle East, and beyond — have repeatedly turned to you to transform their buying processes. Galeries Lafayette, the French department store with nearly 150 years on the market, is one such example. What systemic mistakes in buying approaches do you encounter when working with international players of this scale?
– The most common issue I see — even among the largest players in the market — is that buying decisions are still largely built on intuition, trend interpretation, and retrospective sales data. What this means in practice is that the buyer is essentially trying to predict the future by looking at the past through a subjective lens. My work is precisely about shifting that balance toward data and predictive analytics.
— Your algorithmic approach delivers forecast accuracy of up to 87% — significantly above the industry average of 50–60% — with a meaningful impact on margin growth. What is the logic behind the system, and what accounts for results like these?
Through my Predictive Buying Intelligence Platform (PBIP) methodology, I move the buying process away from an intuition-driven model toward an algorithmic one, where decisions are grounded in a combination of digital signals, behavioral patterns, and historical dynamics rather than subjective trend assessment. I still leave room for expert judgment, but the driving force is 48 weighted variables — spanning digital signals through to historical sales performance. This minimizes the influence of human bias and allows for more accurate, consistent decision-making.
The second important element is the shift from a static model to an adaptive one, and from an aggregated approach to a granular one. Rather than a single decision per season, I work with continuous optimization: as demand signals shift, the system allows for rapid inventory reallocation. Decisions are made at the level of individual stores — accounting for local demand rather than averaged figures.
— Beyond improving buying accuracy, PBIP is also associated with tackling the problem of excess inventory, which in the US alone is estimated at around $25 billion annually and carries a significant environmental cost. How does the methodology address that?
This is indeed one of retail’s core challenges — a large portion of inventory ends up distributed inefficiently. What customers actually want sells out quickly or is understocked, while low-demand items sit on shelves.
From there, a familiar chain of compensatory measures kicks in: discounts, promotions, markdowns. In the luxury segment, this is particularly damaging — a customer buying something that genuinely aligns with their needs and the brand’s value proposition is willing to pay full price. Frequent sales, by contrast, erode perceived value rather than stimulate demand. The retailer ends up losing margin and, at the same time, quietly undermining the brand’s positioning.
My methodology addresses this through more precise demand forecasting at the level of individual stores and time windows. This makes it possible to redistribute inventory proactively — sending it where it will actually sell — and to reduce the volume of misjudged purchases that eventually become markdowns or deadstock. The result is not only a lower share of unsold inventory, but a reduced need for aggressive clearance sales as a mechanism for offloading stock.
— Indeed, following the implementation of PBIP, several major retailers reported a 20–30% reduction in markdown levels. How universal is your PBIP methodology? Is it applicable only to the US market, or do you also deploy it in other regions — luxury retail in the Middle East, for example?
PBIP applies equally to both markets, but its strength lies in accommodating regional differences through parameter configuration, not by changing the underlying logic of the model.
On Middle Eastern markets, for instance, I found that demand is event-driven and relationship-dependent. I observed demand concentrated around Ramadan, Eid, and the wedding season, with personal relationships between clients and sales associates driving a significant share of purchases. By my estimates, around 60% of transactions are tied to established client-advisor relationships. Within PBIP, I applied a depth-oriented buying strategy in that environment — hero products, proven performers, and size ranges calibrated to regional demographics. The physical store served as the primary venue for both the customer experience and the transaction itself.
In US markets, by contrast, I see demand shaped by trends and the digital environment. Viral moments generate distributed, rapidly emerging demand patterns that require agile, real-time inventory reallocation. Product discovery happens through social platforms, and fulfillment must be omnichannel. Here I build a strategy around breadth of assortment with surgical precision — accounting for micro-segments defined by climate, demographics, and regional preferences. I apply my store-level allocation formula within PBIP to determine differentiated volumes: for example, Store A receives 150 units, Store B receives 40, and Store C receives zero until online demand is confirmed.
— In many ways, you’ve moved beyond classic fashion into deep technical expertise. How do you see the role of the buyer in five years — still a person, or something more like a human-system hybrid?
In five years, the buyer function will, in my view, be significantly less operational and far more architectural. Algorithms will handle the bulk of routine and predictable decisions — demand analysis, assortment allocation, inventory management within standard scenarios. The human role will shift toward configuring the logic of the system: defining parameters, weighting factors, interpreting model outputs, and making adjustments where needed.
That said, there will remain areas where human judgment is critical — above all, responding to genuinely novel and unconventional trends, navigating complex negotiations with partners and suppliers, and situations where trust, intuition, and emotional intelligence matter — things that can’t yet be formalized.
Essentially, the buyer will evolve from an operational participant in a process to a specialist who manages a commercial intelligence system. It’s no longer about selecting products — it’s about governing how a system makes decisions about products. And perhaps in time, the word “buyer” itself will feel too narrow. A more precise descriptor will take its place — something like architect of demand and commercial systems.
Presented by DN NEWS DESK
