Before building products, I study how people respond to them.
My research background helps me understand user behavior, trust, personalization, relevance, and decision-making. I use research not as theory alone, but as a way to improve product judgment.
How personalization and social presence shape user trust in chatbot experiences
A 2×2 controlled experiment on human–chatbot interaction in e-commerce — built in Landbot, tested on real users via Prolific. I treat this as evidence of how I design experiments, isolate product variables, and translate user behavior into product insight.
The product question
As AI chatbots become standard in e-commerce and banking, a critical design question emerges: what actually makes users trust them enough to act? Is it making the chatbot feel more human (social presence)? Is it making it feel more relevant (personalization)? Or does the combination create something greater than the sum of its parts? I designed a controlled experiment to find out.
Experiment design
Built four distinct chatbot prototypes in Landbot — each simulating an e-commerce product recommendation experience — using a 2×2 factorial design:
- Factor 1: Social Presence (high vs. low)
- Factor 2: Personalization (high vs. low)
This created 4 conditions: High+High · High+Low · Low+High · Low+Low. Participants recruited via Prolific completed a pre-interaction survey in Qualtrics, the chatbot interaction in Landbot, then a post-interaction survey measuring perceived social presence, personalization, trust, and purchase intention. Analysis included reliability checks, manipulation checks, mediation analysis, and regression-based hypothesis testing.
Key findings
- Both social presence and personalization independently and positively influenced user trust — through different mechanisms.
- Trust was strongly associated with purchase intention.
- Personalization had both a direct effect on purchase intention AND an indirect effect through trust — working on two levels simultaneously.
- The interaction effect between social presence and personalization was not significant — they work independently, not multiplicatively.
Why this matters for AI product design
- Gives product teams an empirical basis for chatbot design decisions, not just intuition.
- Directly applicable to AI advisory tools, authentication flows, and customer service bots in banking and fintech.
- The 2×2 design is a template for how to run controlled product experiments — isolating variables and drawing conclusions beyond "users preferred version B."
Tools
Landbot · Qualtrics · Prolific · Mediation analysis · Regression-based hypothesis testing
Targeted Advertising & Product Involvement
A quantitative study on when personalized online advertising actually becomes more effective — and how product type changes the impact of personalization on click intention.
The product question
When does personalization actually make users more likely to click on an online ad? Most teams treat personalization as a binary capability — either an ad is targeted or it isn't. I wanted to test the more useful product question underneath that: under what user and product conditions does personalization actually translate into behavior?
Research design
- Quantitative survey-based study of targeted online advertising.
- 307 complete responses from everyday web users.
- Measured perceived personalization, perceived personal relevance, perceived intrusiveness, and product involvement.
- Used Partial Least Squares (PLS) analysis to test mediation and moderation relationships.
Key findings
- Personalization doesn't work just because an ad is technically personalized. It works when users perceive the ad as personally relevant — perceived personal relevance mediates the relationship between perceived personalization and click intention.
- Product involvement changes how personalization works. For higher-involvement products, perceived personalization has a stronger effect on perceived personal relevance — meaning the same personalization strategy should not be reused across every product category.
Why this matters for product work
- Shows how customer behavior shifts across product categories.
- Pushes product and growth teams beyond generic, one-size-fits-all personalization.
- Connects segmentation, relevance, product context, and conversion behavior into one coherent model.
- Provides a research-backed way to think about ad personalization, recommendation systems, and user targeting.
- Reinforces my product perspective on experimentation, user behavior, and data-informed decision-making.
Tools & methods
Survey design · Quantitative research · PLS analysis · Mediation analysis · Moderation analysis · Advertising effectiveness · Customer behavior research