Industry
Retail
Omnichannel CX and Intelligent Innovation – The Future of Retail
HIGHLIGHTS
- Retail decisions built only on sales data often miss customer intent, trade-offs and choices that never resulted in a purchase.
- Non-purchase behaviour, when analysed properly, can reveal hidden demand, substitution patterns and missed opportunities that traditional metrics overlook.
- Incorporating these signals into everyday merchandising decisions helps teams move from reactive analysis to more anticipatory and confident decision-making.
On this page
On this pageinpage
The blind spot
Merchants are deluged with data more than ever before. Every transaction, price change, and promotion is captured, with attributes measured at granular levels. Yet, despite all the available data, there’s still a gap in merchandising decisions. That’s because traditional merchandising strategies across pricing, assortment, and demand planning largely rely on observed outcomes such as conversions, sales velocity, basket composition, promotion uplift and inventory sell-through.
While this approach explains what sold, it offers limited insights into unmet demand, abandoned choices, and latent customer intent. These missed signals don’t show up clearly in standard reports; they stay buried, often treated as noise rather than information. Since there are no sophisticated means to capture such dormant signals, , these blind spots add up over time, affecting merchandising decisions. As a result, decisions often become reactive, driven by visible outcomes such as sales data, while the underlying reasons for conversions that never happened often get overlooked.
This paper proposes how to leverage signals embedded in unsold products and ‘no-buy’ decisions to enrich merchandising decisions.
Beyond sales data
Typically, sales data shows retailers what customers purchased, not what they may have considered but left on the shelf. Therefore, with limited insights into no-buys, retailers often end up relying on assumptions. For example, a product showing low sales might be attributed to low demand. But the reality is often more nuanced. The conversion may not have happened because the price was slightly off. Or the size available on the shelf was not the right fit for the customer. Or customer loyalty was more towards a competing brand. Because there are no means to capture these scenarios, reports tend to be broad-brush and low on detail, even though the underlying reasons can be quite different. Retailers often try to bridge this gap through surveys and market studies. While these approaches are useful, they are conducted only occasionally and are limited in scale. Moreover, they don’t capture real-time decision behaviour at a granular level.
At the same time, customers are now exposed to more information than ever before, and their preferences change quickly, often across different age groups. This makes buying behaviour much harder to understand. The new imperative is to systematically understand no-purchase signals: those moments where a customer considered something but eventually did not buy.
Decoding no-buys
Advances in analytics, AI/GenAI models, and agentic AI now make it possible to detect no-purchase signals at a much more granular level. Instead of relying on broad averages, these models powered by agents analyse micro-level interactions—how different customer segments respond to small changes in price, assortment, or availability. Segmentation itself can be tuned to reflect meaningful business contexts, making insights more relevant and actionable.
Beyond these interactions, agents can surface a wide range of non-purchase behaviours, including browse-to-buy drop-offs, repeated product comparisons without conversion, promotion blindness within specific segments, switching between private-label and national brands, and interaction gaps such as searches that don’t lead to engagement. They can also identify sequence patterns, for instance, when customers consistently delay a purchase until a markdown, or when intent weakens after exposure to substitutes. Over time, these patterns help distinguish true lack of demand from unmet demand driven by pricing, assortment, or positioning gaps.
This layer of insight reveals what traditional metrics miss. A men’s shirt priced at INR 1,499 may underperform while a similar option at INR 1,299 sells well—not due to low demand, but because customers are comparing and choosing the lower-priced alternative. The demand exists; it’s simply being redirected. Similarly, a private-label product may receive high visibility but low conversion compared to a national brand, indicating perception or trust gaps rather than absence of interest. In this sense, no-buy signals act as a potent second layer of data, uncovering demand that is present but not realised.
However, invisible signals alone are not enough. To make these signals actionable, retailers need a unified decision framework (Figure 1) that connects fragmented inputs across channels and functions. This is enabled through a centralised decision intelligence layer powered by specialised AI agents across pricing, promotions, customer behaviour, assortment, and inventory. Together, they convert hidden patterns of non-purchase behaviour into coordinated, optimised merchandising decisions—shifting merchandising teams from reacting to outcomes to proactively addressing missed intent.
Figure 1
Decision intelligence architecture for smarter merchandising
Figure 1
From signals to actions
What matters isn’t just spotting why customers don’t buy; it’s using those insights to drive outcomes. The value lies in turning signals of hesitation, substitution and unmet demand into specific merchandising actions that lift conversion, protect margin, improve forecast accuracy and reduce lost demand. When these signals are built into everyday merchandising decisions, they don’t just give insights, they actively improve business performance.
In pricing, these signals show where shoppers are interested but not buying, helping merchants fine-tune price points, making even small adjustments that can significantly improve conversion. In assortment planning, they highlight gaps where demand exists, but the offer doesn’t fit, guiding better choices on products, sizes, or alternatives to reduce lost sales. In demand planning, they provide early visibility into changing demand, improve forecasts, and help teams stock the right products before stockouts occur. The key is to move from insights to action in a structured way, connecting insights with decision systems rather than treating them as standalone analysis. This ensures the learning loop becomes tighter, signals are continuously captured, interpreted and fed back into decisions. Over time, this shifts how decisions are made; they will no longer be based only on past outcomes but also on missed opportunities.
However, the real challenge lies in how quickly retailers can move from insight to action, improving KPIs. This is where modern merchandising optimisation engines come in. With the ability to operate at scale, speed, and at a granular level, and to simulate business scenarios in near real time, they help merchants act on weak signals before they become visible in topline sales. The commercial payoff is clearer and faster: higher conversions, stronger margins, better forecast accuracy, and reduced lost demand.
Making it work
Making the shift from relying only on sales outcomes to incorporating deeper signals of customer intent is not just about introducing new AI or GenAI models. It involves rethinking how data is used across various teams across the organisation and following a structured approach:
Pricing, assortment, inventory, and customer interaction data often sit in different systems. The first step is to unify data in a usable form across functions. Otherwise, signals will remain fragmented, and decision-making will be siloed.
Analytics must support micro-level decision-making. Broad averages are not enough; insights should be actionable at a product, store, or segment level.
If insights remain as reports, their impact is limited. If they are part of autonomous decision systems, they have the potential to influence outcomes directly while freeing merchants to focus more on strategising instead of firefighting. This is where modern, cloud-based autonomous decisioning tools play a key role. They help scale these capabilities without adding complexity.
With the right data strategy in place, the impact becomes visible at the category and SKU levels, as seen in the following example.
Consider a grocery retailer noticing consistently low conversion in a mid-priced ready-to-eat category. Traditional reports suggested weak demand. However, by analysing non-purchase signals, the retailer discovered that customers frequently compared these products but switched to lower-priced alternatives during checkout. By slightly adjusting price points, introducing smaller trial packs, and repositioning select SKUs, the retailer was able to improve conversion within the category, increase overall category penetration, and reduce substitution-driven revenue leakage. This illustrates how capturing and acting on non-purchase signals can convert missed intent into measurable business outcomes.
What changes next
As retail becomes more dynamic, relying only on past sales will become increasingly limiting. Many more factors than ever before influence customer choices today. Price comparisons happen instantly, substitutes are easier to discover, and preferences shift faster than expected. In this context, understanding why customers didn’t buy a product becomes just as important as what influenced their purchase decision.
The choices customers don’t make often carry signals that are easy to miss but difficult to ignore once identified. Non-purchase signals don’t replace traditional analytics; they make them commercially sharper. They help explain where conversion is leaking, where margins are getting eroded, where demand is being misread, and where customers are walking away despite clear intent. Over time, that enables merchandising teams to reduce lost demand, improve forecast accuracy, and make pricing and assortment choices better aligned with how customers shop.
Retailers that systematically capture and act on non-purchase signals typically see measurable improvements across key metrics. These include conversion uplift in high-friction categories, margin improvement through more precise pricing decisions, reduction in lost demand due to better assortment alignment, and improved forecast accuracy driven by early detection of demand shifts. The impact is not from a single intervention but from continuously capturing and acting on previously ignored signals. This also marks a gradual shift from purely operational execution to more intelligent, guided decision-making, in which systems support choices more proactively. This shift from visible outcomes to invisible signals is the next frontier in decision-making.