Concept to cash in five months, and the insight that came from watching a stopwatch
A wholesale vehicle marketplace went from an idea in January to live in May, with $1B through it in nine months. The product worked. The business came from an analytics number nobody was looking for.
One January, the wholesale auto marketplace I led product and engineering for had an idea: let a dealer sell a wholesale vehicle from their own lot, through their phone, instead of trucking it to a physical auction. The industry's biggest trade show was in March. We wanted something real to put in dealers' hands by then.
What follows is the version of the story I tell new product leaders, because the launch is the part everyone remembers and the stopwatch is the part that mattered.
Eight weeks to a prototype people could hold
The first version for the show had a working front end and a mocked back end. Dealers could walk through listing a car and see what buyers would see. It was not connected to anything, and we said so. The point was to get a hundred reactions from the people who would use it, not to pretend it was finished.
The reactions were good enough that when we got back, I brought 22 teams together to build the real thing. Operations, sales, customer care and the auctions all got daily builds, so by the time the application was ready in mid-May they had already been using it for weeks. Training was not a phase; it was a side effect of the daily build.
We went to market ten days early. Three weeks after that, with sales being driven by dealers who wanted to sell to dealers in other states, we turned it on across six states.
The number nobody asked for
In mid-June I was reviewing the usage analytics with the operations leader. Listing a vehicle was taking users eight to twelve minutes per car.
That was not a defect. The flow was clean. It was just how long it takes to photograph a car properly, walk around it, record the damage and the options, and fill in the details that make a wholesale buyer trust a listing from someone they have never met.
But eight to twelve minutes per car, for a dealer with thirty cars to move, is four to six hours. Big dealers were not going to do that themselves. So the operations leader and I asked a different question: what if we did it for them?
We tested the idea with a handful of operations people and product people, without building anything. Someone would go to the dealer's lot and list the cars for a small fee. Dealers said yes faster than we expected. By the first two weeks of July we had a specific version of the application for that "Concierge" user, and the service became a large revenue line of its own.
The product launch was the thing we planned. The Concierge business was the thing we found, and we found it because somebody was looking at the time-per-vehicle number with a question in mind rather than a dashboard to fill.
What I took from it
Three things, and I have applied them at every company since.
Ship the mocked version to real users before the real version exists. The feedback from a hundred dealers at the show was worth more than three months of internal debate, and it cost eight weeks.
Give every operating team the daily build. Training becomes free, launch day becomes boring, and the operations people start finding product ideas because they are living in the product.
Read the analytics as a question, not a report. "Why does this take twelve minutes?" produced a business. "Listing completion rate is 94%" would have produced a slide.
The product had more than 4,000 users within two months and $1B transacted through it in its first nine months. I am proud of that. I am prouder that the number I remember from it is twelve minutes.