How to help people choose the right car for them…
… when they don’t know a thing about cars!
Introduction
In 2018 I was an in-house UX Architect for a publisher company in the automotive industry. There are many big companies that invest a lot in research in this sector, so I had loads of available documentation on different types of car buyers and their needs.
My conclusion after getting my head around this market was that there was a big need that was yet to cover amongst the readers of our magazine. Let me give you some context… There are mainly three types of car buyer:
- The enthusiast, who knows everything about cars and knows what’s buying. These people don’t need any advice choosing a car, they are the ones giving it!
- The speedy-chooser, who buys by impulse and cares more about what the car says about them than about its performance. These people could use some advice, but they would probably not listen to it. They would listen to a good marketing campaign though!
- The benefit-maximiser, who just wants to buy a car that fits their needs, at the best possible price. And, in order to do this, researches, researches (often asks the enthusiast), and eventually ends up feeling overwhelmed and most likely buying the same car they had before, but a newer model. These people could actually use some advice.
Which one are you? I’m definitely the benefit-maximiser.
In a nutshell, I was working in a company full of enthusiasts who wrote car reviews for other enthusiasts, but somehow they were hoping these reviews were helpful for benefit-maximisers who didn’t understand most of what they were reading.
So the idea was to create a tool to help people who don’t know too much about cars pick the best car for them, selecting the exact model they need according to their driving style, their needs, and their budget.
Main tasks
- Research to learn about how other people are doing picking tools.
- More research to learn enough about cars to actually do this work 😅
- Copywriting, as it was key to translate the technicalities about cars into plain English that everyone could understand.
- Wireframing and creation of prototypes to quickly test with users.
- Usability testing, both moderated and unmoderated.
- Algorithm writing, testing and refinement to define the logic behind the wizard so developers’ work made sense.
Design process
Research
The first wizard that I had a look at was Thread, a wizard to help you find your clothing style by picking images you like. I liked that it was very visual, but cars have many features that you cannot easily represent in a photo.
Then I looked at car picking tools, but they all had the same problem: the questions weren’t actually helping the users choose what they need. For example, you could choose between petrol and diesel, but do you really know which fuel is best according to your driving habits and your concerns (price, pollution, etc)?
Another question that I found in all car picking tool was your yearly mileage. It was fun to discover that not even the enthusiasts had an accurate idea of how many miles they drove in a year.
I also needed to talk to car experts to help define the main characteristics that I was looking to define in a car. Basically budget, type of body, engine and fuel, engine power, miles per gallon, and of course carbon footprint.
Prototyping and testing
I worked in Axure and did rapid prototyping and guerrilla testing at the beginning to quickly define what direction I should take.
Sometimes, images were useful to help convert qualitative (understood by people) into quantitative data (understood by the algorithm).
I run loads of moderated and unmoderated testing until I realised that the interface wasn’t as important as the algorithm behind it. However, we were also going through a complete redesign of our website, so there were no developers available to help me write and refine a working prototype connected to a database.
Iterative algorithm creation
I decided to create my own database, creating a spreadsheet with all the details for all the available models of Volkswagen Golf. I chose this model because there are so many lines, variants and engines, so it would be complex enough to start defining an algorithm, and afterwards I could expand it to other models and brands.
I wrote a formula to quantify and weight the qualitative responses of the user, so every time a user answered a question, all the cars on the list would get a score and would be re-ordered from best to worst fit.
I created a quick form in the spreadsheet that would return the perfect Golf for the user and shared it around to get enough data to analyse the performance of the algorithm.
It took many iterations and lots of testing to refine this algorithm until it made sense. A car expert helped me analyse the results and refine the algorithm, I don’t think she will ever read this but I wanted to say thanks anyway!
After the algorithm was working fine, it wasn’t too difficult to expand it to other models and brands, it just required a bit of tweaking depending on types of car, e.g. you can’t use the same weighing for the engine power of tiny city cars and SUVs.
Then I worked on the wireframes of the results page and the project was ready for a developer to grab!
What I learned from this project
- To own a project.
- An essential role UXers need to play in a project is to translate user needs, stakeholder expectations and development requirements into the same language so everyone understands. You can always find a common ground between what users can tell you and what you need to know from them.
- Writing an algorithm is not as difficult as you would think.
- Majority of people could use a much smaller and greener car!
