How we build confidence in a feature @ carwow

carwow is an automotive marketplace platform that has over 44 million annual visitors and 12 million registered customers. Therefore, adapting our core flows comes with a risk to our customer experience, core business metrics (such as sign-up), and revenue. Due to this, we like to make sure we have built as much confidence as possible in anything new we surface to improve our customer experience. Now let me introduce you to the context and current thinking for how we build confidence in a feature at carwow.

In 2020, carwow’s focus was to help our customers find the right car based on their preferences. To improve this flow for our customers, we have explored how we could improve our onboarding and preference collection techniques throughout the site.

Introducing the “configurator”

Our current configurator — a flow for customers to select and build a car — is an incredibly efficient way of serving people who know exactly what they want in their next car such as the make, model, trim and even engine. However, we know from customer data we have collected that many still do not know what they want. The above screenshots show how complicated the user flow was for buyers to find offers for new cars and didn’t know the exact specification of the car they wanted to purchase. For buyers to receive accurate offers from a dealer, there are far more specifications and options that they need to select other than just the fuel type and gearbox of a car.

To improve how users can find their perfect car based on their preferences, we needed to think about how we might capture user preferences to find models all the way down to a detailed specification (we call these derivatives) that were better suited to them…and so…the “discovery flow” was launched!

The discovery flow

Initially, we started with just five questions (capturing body type, budget, fuel, gearbox, and manufacturer) to help customers hone in on a few models that matched their preferences.

We thought we had nailed it!

….but sadly, our customers were still finding it difficult to select their perfect car. We saw some frequent themes popping up across our funnel flow analysis in the configurator and specific insights from our customer calls that suggested we needed additional support to help our customers.

So, what was the problem?

The “discovery flow” had helped customers find a model that suited their preferences. In contrast, the configurator helped customers who had an exact specification in mind. However, what we had not addressed was bridging the gap between the two, to help the customers who had been matched with a model but still could not narrow down the options to an exact specification to fit their needs.

Customers were finding it difficult in both flows to narrow down to a single derivative that included all the must-have equipment that met their needs. How did we identify this? We saw the largest drop-offs in our configurator when users were presented with a lot of choice. 20% of people would give up at the point of selecting a trim, and a further 23% would give up when selecting additional options for their car. We also gathered some useful feedback and qualitative insights from our customer calls to validate what we had seen in this data.

What did we learn from insights like this?

We found that customers often have key features or must-have equipment preferences, but tend to be quite flexible with additional options due to their impact on the final price. At carwow, we know our customers are looking for the best deal from their offers. Therefore, showing them what features are available in each trim will hopefully provide enough confidence they are getting the best value for their money. In summary, the highlights were:

• Customers usually have in mind their “must-have” features
• Features help the customer make a decision
• Trims are hard to match with preferences

So you might think, “just add equipment options to the flow…”

For those not aware of the car industry, it is abundant with different options, extras, equipment, and manufacturing terms within each of these categories that make it particularly complex.

We couldn’t just add them all. So we decided to build confidence in our new equipment feature first to create the most value for our customers.

We began our approach with some competitor analysis to see how other products similar to ours built a feature around this. Each product we reviewed varied, Cazoo used very simple terms to select and filter by, while DriveK has small explanations to better describe the add-on. One similarity is that all of these products have mapped their own car data from manufacturer marketing names to show results that best suit their own customers.

Next, we wanted to find out exactly what our customers at carwow wanted, launching our own site survey with an open text field to collect key requirements from any terms they were familiar with.

This snapshot of results helped us understand the different preferences across our markets (notice the popularity of a trailer hitch in Germany is not reflected in the UK). But we weren’t confident that this should represent the final list; we often see from our results that survey entries are primarily from particularly engaged customers who are more willing to provide feedback. Therefore they may not provide us with insights reflective of our whole customer base. So, in order for us to build greater confidence in the equipment selections we would surface to our customers, we set up a fake door test.

A fake door test is about rapidly validating an idea (it can be a product, a service or a feature): we show the users an option that does not actually exist. Link here

We took the equipment items from the survey and added a fake question in the preference collection flow allowing users to make selections without actually affecting the results shown. This selection data was tracked in Amplitude and allowed us to map out what was popular with our customers in each market.

Taking the results and simplifying equipment terms from the car manufacturer industry required a technical investigation into our internal data, and was important for how we would group and show matches to our customers. We needed to align these to generic real-world terms in order to help all of our customers.

Here is an example of how varied marketing approaches can be for one type of add-on. We took all of the instances applicable for these terms and created different data types as filters in our back-end product.

How does it all look?

Translated and rolled out with subtle differences across our three markets, we have created a consolidated equipment feature to help our customers in the UK, Germany and Spain.

When our customer has completed their preferences in the discovery flow they are able to adapt their results to show more (or fewer) results. From the original customer insights, we know there are some equipment preferences they can’t live without and others that are more of a nice-to-have. The ability to change these filters will open up different results for our customers’ budgets and give them a greater perspective of how they impact the final price.

Customers can now use the equipment types selected to compare the different derivatives available for each model, enabling them to find the right trim for their needs and budget.

Take a look for yourself and see what matches you come up with! If you have any feedback we will be delighted to hear your thoughts 💭 (In my quest for the right car, I came out with an automatic Peugeot 208 Allure 🚙)

Thanks for reading! If you enjoyed this blog post, and want to join the carwow team, why not take a look at our jobs page?