From Gut Instinct to AI: The New Era of Data-Driven Merchandising

• Predictive AI is replacing gut-driven merchandising by turning real-time consumer signals into confident, forward-looking decisions on demand, pricing, and product strategy.

• First Insight’s two-layer AI system—Predictive AI plus a Generative AI interface—promises to translate massive data inputs into instant, actionable recommendations grounded in validated consumer behavior.

• Early insights help retailers engineer winning products, avoid over/underbuying, optimize pricing and shift merchandising teams from instinct-based decisions to evidence-based execution.

Traditional merchandise planning was governed by instinct, spreadsheets and the risk of the “gut decision.” But as retail faces unprecedented volatility, relying on historical sales data and personal hunches is no longer a viable strategy.

Here, Viki Zabala, CMO of First Insight, discusses how Artificial Intelligence is fundamentally transforming the retail landscape. Zabala explains how First Insight’s two-layer AI architecture, which combines Predictive AI for modeling demand with Generative AI for instant, natural-language recommendations, is closing the gap between data collection and actionable, confident decisions. She details why “Value Score” intelligence is more critical than a simple sales forecast, and how early consumer signals are the key to engineering winning products, optimizing pricing and finally mastering the delicate balance between overbuying and underbuying.

HGI: Historically, merchandise planners have relied heavily on intuition and spreadsheets. How is AI transforming those gut decisions into data-driven, real-time intelligence?

Viki Zabala: For decades, merchandising ran on instinct and Excel. AI didn’t just modernize the process; it fundamentally changed the idea that “going with your gut” was a strategy.

AI finally closes the gap on what companies collect and what they actually decide. Mountains of data with scattered signals from sales to sentiment to competitor pricing to product feedback can all now be turned into a clear forward-looking intelligent signal that shows planners where demand is heading — not where it’s been.

The result? Buy calls stop being bets. Allocations stop being guesses. Inventory stops being reactive.

Merchants don’t need more data. They needed a way to turn data into decisions. AI gives them that — instantly.

“Predictive insight shapes the buy. That’s how retailers avoid costly misses and build assortments with confidence instead of hope — or large markdown budgets.” —Viki Zabala, First Insight

HGI: AI is a broad term. What type of AI specifically is most useful in demand planning and pricing and how does it differ from, say, a typical ChatGPT query?

VZ: Retail doesn’t need a chatbot that summarizes last season. It needs AI that can predict what shoppers will do next. That’s why First Insight’s AI architecture has two layers. The first is Predictive AI that models demand, elasticity, sentiment, and SKU behavior using real consumer data. And then there’s Generative AI, which serves as the interface, where [First Insight’s AI growth co-pilot] Ellis delivers instant, actionable recommendations in natural language.

ChatGPT can give you a response, but it can’t validate accuracy. It relies on historical text and will always have to guess about future demand. Ellis doesn’t guess. It’s grounded in real-time consumer signals, so retailers get recommendations they can trust.

HGI: With so much data, from sales to social media and beyond, retailers are inundated. How does First Insight turn that overwhelming data into actionable insights that drive product and pricing decisions?

VZ: Retailers don’t have a data problem — they have a decision problem. Teams have dashboards everywhere and answers nowhere. First Insight fixes that by grounding decisions in real-time consumer signals, then translating those signals into clear, predictive outcomes.

Our AI growth co-pilot Ellis accelerates that process by collapsing millions of inputs into a single next step. Teams can ask:

• Which new products have the highest value score for next season?
• Where are we over- or underpricing based on consumer thresholds?
• Which items are most sensitive to markdowns right now?
• What attributes are driving the lift in this category?
• Which concepts should we fast-track — and which should we kill?

…and the AI gives you a predictive recommendation in seconds — not a spreadsheet, not a report, but an action. The power of AI isn’t the summary. It’s the speed between insight and execution.

HGI: How do you further parse this information to pinpoint the product attributes—like color, fit, or quality—that most influence buying decisions?

VZ: Everyone thinks they know which products will sell — until the data says otherwise.

We stress-test decisions from concept to shelf, and we don’t just reveal the winners — we surface the Value Score intelligence behind them. That includes silhouette, fit, color, material, quality cues, pricing thresholds, sentiment patterns and perceived value.

Our Value Score algorithms combine those signals into one predictive output, helping retailers see not only which concepts are strongest, but why. And while teams can compare their products to market alternatives, the Value Score insights themselves come directly from how consumers respond to their product.

Most products aren’t failures; they’re a 5 or 6 with the potential to become a 9 (out of 10). Value Score data reveals the combination of changes — price shifts, color adjustments, detail updates, material improvements — that collectively move demand.

It’s rarely one tweak. It’s the mix of signals that changes consumer behavior. With Value Score intelligence and attribute-level clarity, retailers don’t just identify the winners — they engineer them.

HGI: Merchandise planning has always been a balancing act between overbuying and underbuying. And these days, disruption and uncertainty are the new normal. How can retailers get it right?

VZ: Planning has always been a gamble because most retailers depend on historical sales data — which works for basics, but fails every time you introduce something new. You can’t predict next season with last season’s receipts. What actually reduces overbuy and underbuy risk is early demand signals, captured before inventory dollars are committed.

At First Insight, we measure consumer reaction at the concept stage and convert it into a Value Score that reliably forecasts which products will win, at what price, and at what buy depth —seasons before launch.

And here’s the part most people miss: that early signal is more accurate than history, and far more stable than people assume. The consumer usually tells you the truth — the hard part is accepting it when it contradicts instinct.

If consumers value a product nine months out, they’ll still value it when it hits shelves — especially once you refine the design and pricing around what they told you.

So, “getting it right” in today’s volatility isn’t about real-time trend chasing. It’s about:

• validating demand upstream
• planning newness with evidence
• and using predictive consumer data instead of post-mortems

Historical data explains the past. Predictive insight shapes the buy. That’s how retailers avoid costly misses and build assortments with confidence instead of hope — or large markdown budgets.

HGI: One of the main challenges with markdowns is they can have a negative impact on brand perception. How can retailers adjust pricing without this negative outcome?

VZ: Retailers don’t lose margin because they mark down — they lose margin because they mark down the wrong products, at the wrong time, for the wrong reasons.

First Insight predicts a product’s model price, its realistic AUR and how demand will hold or fall as price changes. That means retailers can establish the right starting price and the right markdown cadence before the product even launches — avoiding the two killers of brand perception: discounting too fast and discounting too deep.

The data is consistent: when retailers over-discount, shoppers learn to wait. It erodes price integrity, pulls demand forward and cheapens the brand. We’ve seen many leading retailers reverse course and return to full-price strength because they realized the discount addiction was conditioning customers.

Predictive consumer insight fixes this by showing exactly which items will convert with a small price shift, which ones need a deeper cut, which shouldn’t be discounted at all, and the precise timing for markdowns that sustains demand without signaling distress.

When a discount is grounded in consumer demand rather than panic, it feels intentional — not desperate. That protects margin and brand value. Smart pricing isn’t a race to the bottom. It’s knowing the price consumers will pay — and staying aligned with it from day one.

HGI: With fashion leaning more into AI, how is this changing the types of talent retailers need to bring on board to keep up?

VZ: AI is absolutely changing the talent retailers need, but the real shift isn’t just about hiring new people. It’s that decades-old roles, org charts and approval chains can’t survive in a world where the consumer signal shows up instantly. Some tasks will be automated, some roles will evolve and some will disappear. But the work that remains becomes more strategic and more connected to the customer.

The talent retailers need now are people who can work with clarity: talent that can read what the customer is actually telling them, challenge long-held assumptions and adjust assortments, pricing or creative direction without waiting for a hierarchy to bless every step. Not technical experts — decision-makers who know how to use better information.

But this only works with top-down leadership. AI adoption in retail is a change-management problem, not a hiring problem. CEOs have to break the silos, reset old processes and give teams permission to move faster. Without that, the best talent in the world will still get trapped in systems built for a different era.

HGI: With so many options, what’s the best way for retailers and brands to vet AI solutions today?

VZ: Retailers are testing every kind of AI right now…creative tools, workflow shortcuts, forecasting engines, and plenty of shiny objects with big claims. But the most important question isn’t “What can this AI do?” It’s “Will this AI help us make better decisions — not just faster ones?”

In retail, true ROI isn’t instant. Some wins show up in-season, but many reveal themselves over a full merchandising cycle, because planners want to validate predictive accuracy against final sales. And that’s the right instinct — the stakes are too high for guesswork. That’s why early demand signal matters: it gives retailers evidence before dollars are committed, and accuracy they can confirm later.

The AI worth investing in must do three things: improve decisions in the moments that matter — buys, pricing, product, planning; produce predictions retailers can measure against real sales months later; and fit into how teams actually work, while helping them evolve outdated processes — not reinforce them.

Retail doesn’t need more tools. It needs AI that changes decisions and proves it.

Read article on HGI.