I assumed that running lean in the business world boiled down to streamlining manufacturing operations (see my earlier post, “What is lean manufacturing?“). I was wrong. My view of lean concepts, not to mention my pre-existing evaluation criteria for media projects, was shaken up not long after the supply chain management class at MIT ended. That fall, I saw a flyer on campus for a special lecture by Eric Ries, a young entrepreneur who wanted to talk about starting tech ventures using lean methods.
The lecture was sparsely attended, but was utterly fascinating. Ries, who later released a book titled The Lean Startup, had some very interesting ideas about product development based on his experience as a software developer and founder of several tech companies. In the mid-2000s, Ries cofounded a company called IMVU that intended to capitalize on the instant messaging craze as well as interest in video games and virtual worlds. As CTO, Ries was responsible for overseeing development of a product that let people create 3D characters who could hang out in a computer-generated environment. Critical to IMVU’s strategy was building an add-on application that connected the virtual world to instant messaging applications such as AOL Instant Messenger and Yahoo Messenger. The idea was to let users invite their friends to join IMVU.
Ries and his partners thought that IMVU served a market need—a social platform for the gaming generation. They saw the IM add-on as clever way to build viral growth. A new user could easily invite dozens of friends from their buddy lists. It didn’t matter if the user was on AOL or Yahoo Messenger—IMVU could work with them all. They assumed users would love this feature.
So they got to work. The first version of IMVU ended up taking six months to develop. There were lots of features, from the IM add-on to customization tools for users’ avatars. When the product finally launched, it was an utter flop. Practically no one downloaded the software. Only after interviewing test users did they realize most of their initial assumptions about their target audiences were wrong. Testers liked the avatar concept, but they hated the IM connections. Users did not necessarily want to invite their friends—they wanted to use IMVU to make new friends!
Realizing that they had misjudged the needs of their prospective customers, Ries and his team went back to the drawing board. Developers ended up throwing away thousands of lines of code to completely rework the product. IMVU eventually settled on a paid virtual world model without IM connections. Growth finally kicked in, and the company was able to scale up to a real business with more than $X million in monthly revenue.
Looking back at the experience, Ries acknowledged that IMVU’s initial top-down product development process was flawed. The team made a huge leap of faith about the market. They ended up wasting months creating a product that customers did not want. It was a huge mistake that almost killed the company.
IMVU was not alone. “Most of the software projects that are undertaken are never used by anyone,” Ries said during his lecture. It wasn’t just naive startup teams with little market experience wasting time making things that no one wanted to use. Big corporations made this type of mistake, too, he added.
Lean Startup at IMVU
After its initial failure, IMVU took a different approach that ended with success. The company started to test its assumptions before engaging in monumental product development projects. This entailed creating hypotheses about the business, features, and other aspects of IMVU. They then tested their hypotheses, measured the results, and quickly incorporated changes based on what they learned. Then they started with a new set of hypotheses, learned from the results, and iterated again. Ries called this approach validated learning.
Ries urged everyone attending his talk to consider validated learning and other aspects of his Lean Startup framework for our own startups. His 2012 book articulated the framework and gave more examples from IMVU and other companies.
The book also looked back to some of the same lean manufacturing concepts that I had learned about in my supply chain management class, such as the elimination of waste and focusing on speed. While heavy industry and software development may seem to have little in common, Ries believed that some of the lean methods used by companies such as Toyota could be applied to technology startups. He appreciated the focus on small batches and continuous improvement, as well as the internal feedback cycles used in the kanban production control system to regulate parts procurement and other production processes.
But Ries made one critical change: Instead of feedback taking place internally, he brought the feedback cycle out to the marketplace itself. Ries insisted that the team needed to validate hypotheses about the product and its customers. This means placing pre-release versions of the product—or representations of the product, such as a demonstration video—in front of test customers, and doing so in a way that allows the team to empirically measure the results and achieve validated learning, which can then be brought to the next iteration. This build-measure-learn cycle is central to the Lean Startup framework.
The hypotheses can involve grand assumptions about the marketplace (“people are interested in buying self-driving cars”) or smaller hypotheses about features (“customers need a manual brake pedal to override the car’s computer”), design (“North American customers prefer blue cars”), or marketing (“customers are more likely to sign themselves up to the waiting list if a 2% rebate is offered.”)
For instance, if the team wants to test the rebate hypotheses, it could create two versions of a website landing page and show it to 200 prospective customers. Half of the customers would be shown the first version, which includes a button that says “click here to get on the waiting list.” The other half would see the second version, which features the same button, along with a sentence that promises a 2% rebate to anyone who puts themselves on the waiting list and ends up buying a vehicle. By comparing the results from the A/B test, the team will better understand appeal of the rebate. If the results are similar, there is no reason to continue the rebate as prospective customers will sign up at the same rate even if no rebate is offered. If, however, the rebate option has a signup rate that is three times greater, then the team has validated the hypothesis. It makes sense to incorporate the rebate offer into the marketing plan.
Coming soon: Lean Startup, MVP, and finding product-market fit