AVWA · Quantitative Research

The starting point

This case is the direct sequel to AVWA's service conception, the conceptual peer-to-peer carsharing service under the Volkswagen brand. In the previous stage I had stacked up hypotheses from secondary research: that cars sit idle most of the time, that there is pent-up demand in Brasília, and that plenty of people drive without owning a vehicle. They were good hypotheses, but still educated guesses. What was missing was the voice of the people who live the problem.

That gap is what the quantitative stage came to close. Since this was the first time I approached users directly about the P2P sharing model, the survey had a double job: to validate what the desk research suggested and to reveal what I did not yet know I needed to ask.

The challenge

Move from perception to data. I needed to understand the target audience across three layers: who these people are, how they get around today, and what holds them back when it comes to renting or lending a car. Without that map, any flow I designed later would be a bet on an imaginary user.

Why quantitative research

The question at this stage was not "why do people do it" but "how many people do it". I wanted to size behavior and measure how often certain motivations and frustrations showed up, not interpret them in depth. A structured survey, distributed as a form, was the right instrument to turn hypotheses into comparable percentages and to show where it would later be worth digging deeper with qualitative research.

The data journey

I built a research plan for the P2P sharing model and ran a questionnaire that gathered 61 responses. The profile of who answered already told a story: mostly women, nearly half between 30 and 39 years old, income concentrated between R$2,001 and R$6,000, and nine out of ten living in the Federal District. It was exactly the Brasília audience I wanted to hear.

The first contrast showed up right away. The vast majority hold a driver's license, yet fewer than half own a car. Almost 80% use ride-hailing apps, a large share depends on public transport, and around 30% rely on a car borrowed from a relative or friend. The desire to drive is there; it is access to the vehicle that fails.

The data point that anchored the entire service thesis came from usage: 76% said the car is used for up to six hours a week. Translated, that means vehicles sit idle at least 91% of the time. On one side, licensed people with no car; on the other, cars parked in the garage. AVWA exists to connect those two ends.

I also asked about perception and fear. Among the car's advantages, freedom and autonomy lead by a wide margin. Among its downsides, high cost and the difficulty of parking are the most pressing pains. As for the P2P model, the fears sort themselves by role: those who lend worry about accidents and damage, and whether the driver will take good care of the car; those who rent worry about cost and about the billing process. Those four fears became direct trust requirements for the service.

Process and role

I ran every front of this stage, organized into four phases:

1. Research plan. I set the objective, the hypotheses to test and the audience cut based on what desk research had surfaced, so the questionnaire would answer decision questions rather than loose curiosities.
2. Instrument design. I structured the form to cover demographic profile, mobility behavior, car usage, perceived advantages and disadvantages, fears about the P2P model, and prior experience with traditional rental agencies.
3. Collection. I distributed the survey and gathered the 61 responses from the Federal District audience, the territory the service intended to serve.
4. Analysis and synthesis. I cross-referenced the percentages to turn isolated answers into a reading of opportunity, always returning to the business question: is there room for peer-to-peer carsharing in Brasília, and for whom.

I used AI tools to help organize and read the data, but the judgment of what to ask, what to cross and what it meant for the service was mine. Data does not decide on its own; it informs the decision.

What I delivered

The output of this stage was not an interface, it was a quantified portrait of the audience and a list of opportunities prioritized by evidence. The core validation held up: there is a pool of licensed people without a car living alongside a fleet that stays parked most of the time, and that slack is AVWA's raw material.

Prior experience with traditional rental agencies gave me the map of pains to avoid. Three out of four respondents had already rented from a conventional agency, and the complaints clustered around long waits for service, slow reservation confirmation, mismatches between the car booked and the car delivered, and incidents of accidents or damage. Each recurring complaint became a point the service had to solve better than the current market.

The cross-tabs also lit up the exclusion angle I had been carrying from the previous stage. A good share of those who depend on public transport could use an acquaintance's car, but not owning the vehicle means no guarantee of availability when they actually need it. And among that same group, nearly 88% either drive or want to get their license. The demand is there, dammed up by lack of access, not by lack of will.

Learnings

1. Good data starts with a good question. The quality of the insights came less from the analysis and more from the research plan. Tying the questionnaire to the desk research hypotheses is what kept me from collecting pretty, useless numbers.
2. Quantitative sizes, it does not explain. The survey told me how many people fear lending their car, but not the deep why behind each fear. I left this stage with a clear map of what would need qualitative listening later, and seeing that limit is part of the method.
3. User fear is a product brief. The four P2P fears, accidents, care of the car, cost and billing, stopped being obstacles and became the list of guarantees the service is obligated to offer. Hearing the objection early is what separates a desirable product from an ignored one.
4. A small sample calls for humble reading. These were 61 responses, enough to steer direction, not to settle statistical truth. Today, as a Product Designer with a data lens, I treat numbers of this size as a strong signal to confirm, and I communicate that uncertainty instead of hiding behind the percentage.