Understanding Shoppers: the Real Driver of Product Performance

Design is often reduced to its visible layer: mockups, interfaces, components. Behind that surface lies less spectacular but more decisive work: understanding what really happens when a shopper is alone in front of a screen. That is what separates a high-performing purchase journey from one that merely looks good.

Le design n'est pas une succession d'écrans

Many personalisation tools promise to improve conversion. Few focus on what is actually holding it back. Behind every design decision at Matcha, there is a straightforward question: is this choice grounded in a real understanding of purchase behaviour, or is it based on assumption?

Building an effective product recommendation engine is not about assembling algorithms. It starts with understanding how shoppers phrase their search, which criteria they actually use, and at what point they drop off. That foundational work is what makes our solutions useful rather than incidental.

Our Approach: Quantitative and Qualitative, Hand in Hand

Effective product research relies on two complementary types of signal. Neither is sufficient on its own.

Quantitative data: it shows where

Analytics are our compass. They show us precisely where shoppers drop off in the purchase journey: at which stage, on which product type, with which filter configuration. Without a rigorous reading of the data, there is a real risk of optimising journeys that nobody follows.

AI and data analysis tools can significantly accelerate this work: detecting patterns at scale, identifying anomalies, cross-referencing variables.

User interviews: they explain why

If the dashboard shows the symptom, the interview reveals the cause. We use structured interview protocols to surface what the numbers cannot tell us: the mental load involved in choosing between products, implicit barriers, budget trade-offs, and the cognitive shortcuts shoppers rely on.

A concrete example: data may flag a high abandonment rate on a given category. An interview can reveal that the available search criteria do not match the shopper's natural vocabulary. That is not a code bug. It is a comprehension problem that only qualitative research can detect. This type of insight also feeds directly into the quality of our product recommendation engine and improves the relevance of results for the shopper.

Test early, to avoid paying later

As soon as a prototype is ready, we put it in front of real users. The goal of guerrilla testing is straightforward: validate or invalidate our assumptions before they cost development time.

In practice, we ask users to complete a specific task on the interface with minimal guidance, while narrating their thought process aloud. Every hesitation, every unexpected click, every remark is documented.

It is often these unplanned moments that expose the real usability issues, and sometimes the best product opportunities.

A methodology built for performance

Every qualitative insight is cross-referenced with quantitative data to turn an assumption into a measurable growth lever. The cycle is continuous: test, analyse, iterate, repeat. A product is never finished; it evolves alongside shopper behaviour and usage patterns.

This rigour is what guarantees to our retail and brand partners that Matcha's solutions do not stay at surface level. Behind our recommendation engine, our retail media layer, and our shopper data tools lies a detailed understanding of what actually happens in the purchase journey, category by category.

AI changes the tools and can accelerate certain stages of this work. But deciding which problems to solve, for whom, and in which specific purchase context, remains work that requires industry expertise that data alone cannot replace. That is work we do not delegate.

---

Questions fréquentes

Why does user research matter for a product recommendation engine?

A recommendation engine only performs well if the criteria it surfaces match the real vocabulary and needs of shoppers. Without user research, you end up optimising an interface for what you imagine purchase behaviour to be, rather than what it actually is.

How do you improve an online purchase journey?

Improving the purchase journey operates on two levels: identifying where shoppers drop off (through quantitative analysis) and understanding why (through qualitative interviews). Combining both is what allows you to act on the right levers, whether that involves e-commerce personalisation, e-merchandising, or simplifying filter criteria.

What is the difference between quantitative and qualitative data in shopper analysis?

Quantitative data (analytics, conversion rates, abandonment rates) shows where friction occurs in the purchase journey. Qualitative data (interviews, user tests) explains why that friction exists. The most useful shopper data is the kind that brings both together.

Can AI replace user research?

AI is a useful accelerator for analysing large volumes of behavioural data and detecting patterns. It does not, however, replace the detailed understanding of shopper motivations, barriers, and mental load that structured interviews and observation require.

How does Matcha integrate product research into its solutions?

At Matcha, product research is ongoing: regular user testing, engagement and conversion data analysis, and interviews to understand purchase behaviour category by category. These insights feed directly into the development of our Advisor, Boost, and Insights solutions.