In a digital business, it makes little sense to make the distinction between technology and “the business”. We have multi-disciplinary teams with product, design and developers working together around business priorities. We’ve experimented with different ways of organising our teams. We initially adapted “tribes and squads” from Spotify but found shortcomings with the concept of squads arranged around specific functional areas. For example, not all customer feedback sat comfortably with a named squad; rather than being more agile this slowed us down as changes needed to be managed across multiple squad roadmaps. Rather than making us go faster, the squad approach often slowed us down. So, we changed, we’ve kept tribes, but now we are more fluid with our teams. They swarm on problems and opportunities as and when required.

Good decisions require data, to understand and diagnose the current situation, measure impact, and validate the changes you make. We have a data engineering team building a data platform, and data scientists use Databricks, looking for customer trends and insights in the underlying data. But we can’t rely on them to answer every question; we are attempting to democratise data. Increasingly, our developers will start with building the dashboard, so we can measure impact in a visible and accessible way from the start. We use BigQuery with Looker to create dashboards that are accessible across the business. For everyday self-serve analytics, everybody has access to Google analytics.

When you can continuously deliver change—at Auto Trader we average 570 releases a week— it’s an obvious next step to want to test and optimise the platform. Significant improvements or innovation to the customer experience are initially validated either remotely or in our labs through usability testing. Every week we have customers in our offices providing feedback on what we have in production or are thinking about developing. Do we need to change the feature, or is there a better way of helping the customer and their job to be done?

We’ll remove as much uncertainty as possible through the design thinking process. We’ll then move to production, releasing to increasing volumes of users, looking at the difference between test and control. We have evolved our approach to testing; it’s easy to become beholden to the A/B test and feel the need to test everything. To get statistical validity takes time, and with multiple tests running at the same time, it is easy to build-in confounding errors. Sometimes it’s better to just go straight to production. Trust again; very few decisions are irreversible, so trust the team to back out changes if they see an adverse impact.

At Auto Trader, we are always learning, seeking to be better and reinventing ourselves. We did that by evolving from a magazine to a digital business. We have the privileged position of being the market leader, but to continue to grow we need to think and act like a start-up. This means not only rapidly responding to customer feedback on what we do today, but testing—and failing—new ideas and innovation at a speed that makes car finding as easy and car buying as frictionless as possible.

Check out: Top Customer Experience Solution Companies