Pricing with AI: giving merchandisers more impact

In this article, we explain how merchandising departments responsible for pricing can improve their effectiveness using AI technology. We will use a markdown pricing example combined with AI, as this approach is particularly clear and relevant for retailers.

Retail merchandising: what to improve  

Markdown is crucial for clearing inventory and maintaining cash flow. Yet most merchandising teams struggle with the same fundamental issues: 

Issue №1: Treating sales history as a rule, not a guide 

Teams do the right thing by starting with historical data. But it is possible to go further by identifying which parts of that history still matter and which don’t. So, markdown decisions stay consistent as conditions change. 

Issue №2: Measuring markdown impact by one metric instead of outcomes 

Sell-through helps, but it’s just one gauge. To see if a markdown really worked, check a small scorecard: margin, revenue, inventory left, and recovery. This keeps teams focused on results, not just a rate. 

Issue №3: Applying discounts to the entire product groups to simplify calculations 

Applying the same discount to an entire group is efficient. However, checking item-level signals helps surface better price points and prevents unnecessary depth on healthy SKUs. 

Issue №4: There’s a widespread fear in the industry that implementing AI tools means merchandisers won’t be needed anymore 

This misconception holds many companies back from adopting helpful technology. 

Let us show you how each of these problems can be solved, and why the human merchandiser remains essential in this process. 

Solving issue №1: cleaning historical data to reveal true demand 

Here’s what’s really happening in modern merchandising departments: we’re making pricing decisions on “dirty” data. Think about it. Historical data is contaminated by price promotions, marketing campaigns, and seasonal spikes (e.g., “2+1” offers or Black Friday). And it’s possible to work with some of this because marketing can provide data on seasonal spikes.  

But we also have forces outside the marketing calendar—think inflation or COVID-level shocks. These signals move demand, but they aren’t captured in promo calendars. Or when a product is out of stock, the system records zero sales and assumes there’s no demand. But we know that’s not true customers wanted to buy; they just couldn’t. 

This is where modern AI pricing tools make a difference. Instead of working with this “contaminated” history, it is possibly taking massive data sets and clean them: strip out the effects of promotions, adjust for seasonality and macro shifts, and account for stock-outs.  

The result? Clean history that shows true demand patterns.  

Solving issue №2: aligning markdowns with real business goals 

Now that we have clean data, we need to connect it to what really matters: business objectives. 

In an ideal scenario, management comes to the merchandising team with clear goals: “We want to earn one million euros from this markdown campaign. We can use four discount levels, but we cannot accept negative margins.” These become our north star. 

But it doesn’t stop there. Real-world constraints add layers of complexity. Perhaps you need at least 10% of products at each discount level to maintain price architecture. Maybe certain categories can’t go below specific margins to protect brand perception. Each business has its unique requirements based on whether the goal is clearing dead stock, maximising profit, or repositioning the brand. 

This is where the modern merchandiser’s role evolves. Instead of manually calculating possibilities, they capture these business requirements and constraints, then input them into the AI tool. The merchandiser becomes the translator between business strategy and technological capability. 

Solving issue №3: personalised pricing through intelligent optimisation 

With clean data and clear goals, we can now move to the optimisation phase — and this is where the magic happens. 

The AI tool doesn’t just apply blanket discounts to product groups anymore. Instead, it calculates scenarios for each level you need: the level of country, the level of store, and the level of product.  

What happens if we discount this item by 0%? What about 10%, 20%, or 30%? The algorithm runs through thousands of combinations, always checking against the business rules the merchandiser defined. 

The system considers: Will this combination achieve our revenue target? Does it respect our margin constraints? Is the discount distribution balanced across categories? 

Within seconds, we have optimised recommendations that would have taken weeks to calculate manually — and with far greater precision. 

Solving issue №4: AI is a partner to merchandisers, not a replacement 

This brings us to the industry’s biggest misconception. Many believe that implementing AI means merchandisers become obsolete. The reality is exactly the opposite. 

Yes, we receive lists of AI-generated discount recommendations with expected results. Yes, the technology does the heavy computational lifting — quickly, consistently, and transparently. But here’s what technology cannot do: understand context, make strategic trade-offs, and apply business judgment. 

The AI engine guides us toward the goal, but humans complete the journey. Merchandisers decide where to deepen discounts to accelerate stock clearance for an upcoming collection, or where to hold back because a category is still profitable. They understand that sometimes accepting lower margins on one product protects the pricing integrity of an entire range. 

In our real-world experience with retailers, when 60-70% of AI recommendations are accepted, it shows the algorithm understands the business well. We’ve even seen acceptance rates reach 86%. But that remaining 14-40%? That’s where human expertise creates value that no algorithm can replicate. 

AI + merchandising: how it works in real life, by the numbers 

Above we showed where AI truly helps merchandising. Together with an international sports brand, ADC Сonsulting implemented the technologies described above during markdowns — and here are the results: 

• 450% ROI in the first year 

• +2 percentage points to margin 

• +8% increase in sales 

• −10% reduction in average discount depth 

You can read the whole case study on their website, but back to the story.  

The collaboration started with a proof of concept — one month and one product category. This let ADC Consulting measure elasticity, demonstrate impact, and build confidence. Once ADC Consulting proved value, the work scaled to more categories, automated workflows, and advanced dashboards. 

ADC Consulting validates every price impact through controlled experiments — think advanced A/B tests. ADC Consulting is comfortable with outcome-based contracts because the technology delivers. 

Below is what the system is made of and why it’s stable:  

• The foundation: glass-box architecture
  ADC Consulting deploys entirely within your IT perimeter. Your data, code, and tools stay with you — full IP ownership, complete development control, zero vendor lock-in. It’s a transparent process you can understand and control. 

• The engine: causal modelling and optimisation
  At the core, ADC Consulting builds causal models for elasticity and demand forecasting. ADC Consulting creates price response curves down to the individual level. These curves then drive price selection and markdown optimisation, always respecting your constraints. 

• The interface: clear, actionable insights
  Results appear in an intuitive dashboard with slices by item, country, and aggregated curves. Your team immediately sees which discount produces which uplift, transforming insights into action. 

You’re invited to play the pricing game with ADC

A hands-on simulator where you’ll try your hand at the real trade-offs pricing teams juggle every day