We knew illuminating buyer behavior would be tricky. Traditional research methods would easily tip off users to our test, or simulate a shopping experience so removed from the real thing that the results wouldn’t be reliable. Instead, we built a hyper-realistic virtual shopping environment that captured what shoppers actually do, not just what they say they’ll do. This turned ambiguous consumer behavior into predictable science.
The blind spot in traditional approaches
How do you measure the subtle influence of a logo’s position or a single line of messaging in the final moments of an online purchase? The client needed authentic data based on observed behavior, which most standard research methodologies would be unlikely to provide.
A routine discrete choice model (DCM) would explicitly ask shoppers to trade off different options, making them hyper-aware and diverging from the natural, often subconscious, decision-making process. Similarly, the questions used in standard message testing might be leading, inflating the importance of messages that might go totally unnoticed in the real world.
The challenge: capture buyer behavior without revealing what we were truly measuring.
Research that doesn’t feel like research
Our custom-built solution was a simulated e-commerce store for an invented brand, offering coffee machines, coffee cups, and gift sets. To the user, the experience felt like a simple UX test. We gave them a budget and asked them to purchase a coffee product for a friend. However, beneath this familiar interface, we were rigorously testing something else.
We embedded an integrated DCM in the architecture of the online store itself—not as a series of abstract questions. Each shopper was randomly assigned to a different “version” of the checkout page using a monadic design. We varied:
- Whether brand messaging was present on various web surfaces
- Which of the five core brand messages was displayed
- Which payment brands were available
- The order in which they appeared
This design allowed us to isolate the precise behavioral impact of each element. By complementing it with a follow-up MaxDiff exercise on a wider set of messages and an open-ended Narrative Intelligence probe, we were able to capture a far more complete picture. We were able to reveal not just what shoppers did, but why they did it.
Two methods, one comprehensive picture
The integrated DCM allowed construction of a market-based simulator, capable of predicting how shoppers would respond under different scenarios. But that was only one part of the story. The Narrative Intelligence survey provided critical context by engaging users in real conversations at scale. For example, the DCM-derived simulator alone might have suggested that most of the credit card usage was habitual, but Narrative Intelligence revealed that about 1 in 5 shoppers were specifically motivated by loyalty points and rewards associated with their cards, not just convenience.
A purely quantitative model might have missed this lightbulb moment. Instead, authentic customer stories identified a crucial messaging opportunity for our client.
Clear findings meet competitive strategy
The discovery about loyalty points led directly to a powerful, data-backed recommendation for this financial services client: Launch messaging that explicitly reassured shoppers they could continue earning their existing card benefits when choosing our client’s option at checkout.
But our approach gave the client something else too—a powerful quantitative tool in the market simulator. This will allow them to war-game countless “what-if” scenarios, predicting how changes in messaging and brand placement by themselves or competitors will impact their market share at checkout, all without additional costly real-world testing.
The final result? Research that cut through the ambiguity, and a tool for tomorrow’s questions.
Do you have a research problem that traditional methods can’t solve? Get in touch with us—we’d love to help.
