Startups - Ideas You Can Test

     In my last post, I provided a broad overview of how lean manufacturing principles are being applied to the process of innovation.  Lean thinking is ultimately about eliminating waste from a process and being as efficient as possible.  In the context of manufacturing eliminating waste means producing Toyota's with fewer and fewer defects.  In the early stages of a new business, eliminating waste means finding a sustainable business model quickly and efficiently.  All startups begin with an idea.  Here are a few examples:

• People will pay more for self-serve frozen yogurt sold by the ounce
• Let's make a digital version of a Sony Walkman that plays MP3s
• Let's make a digital Walkman, make it super simple to use and connect it to a digital music store
• We should gamify the job search process

These ideas may be good or bad or somewhere in between.  The goal of a lean startup is to quickly discard the bad ideas, optimize and scale the good ones, and figure out how to move the so-so ideas into the good or bad categories.

The FroYo Test


     Let's take the frozen yogurt example to see how lean thinking works.  The traditional way to startup a a new frozen yogurt shop would be use the "Field of Dreams" method.  In other words, build it and hope that customers will come.  You may be thinking, "well only inexperienced or quixotic dreamers would do that."  A better way might be to do some market research:  talk to potential customers, conduct surveys, and figure out price points, etc...  Unfortunately this type of research isn't always reliable.  For example, people tend to overstate their willingness to pay for something in surveys.  Furthermore in the case of the frozen yogurt idea, no one will tell you that they'd be willing to pay more for the privilege of serving themselves.  The self-serve idea relies not only on giving customers control their serving size but also more critically on fact that customers will misjudge of how much yogurt actually weighs and end up paying more than they thought they would.  So, with that research in hand, you could build your store and it might turn a profit and it might not.  If it doesn't, then you've spent and lot of time and money doing research and building your store.

    However, by applying lean thinking, an entrepreneur should be able to find a quicker and more resource efficient way to test her idea.  Market research and surveys are fine.  But nothing beats the feedback you can get by putting a real product in front of real customers and seeing how they react in words and with their wallets.  The entrepreneur could try the idea by going out to a park with a rented yogurt machine on several consecutive weekends and see if the number of people buying yogurt and the amount they spent differed between self-serve and full-serve and between charging by the serving size and charging by the ounce.  Approaching these tests scientifically, the entrepreneur could quickly and systematically quantifying the result of these statistical experiments.  And certainly, these experiments would require less time and money than building an store. Thus taking a lean approach, we are able use a minimum viable product to test the most important parts of the business model (self-serve and by weight pricing) while deferring the more costly but less critical parts (design and build a store).

Data-Driven Design is Good, Data-Driven Design is Bad
     Though the FroYo example, we can see how powerful the use of minimum viable products and hypothesis testing can be.  However, there are dangers to be had with this approach as well.  Designer Doug Bowman famously described these dangers in a farewell to Google blog post in 2009, "data eventually becomes a crutch for every decision, paralyzing the company and preventing it from making any daring design decisions."  Bowman alludes to the larger problem of local optimization where a solution is optimized within neighboring solutions but is sub-optimal overall.  In the gaming industry, Zynga is known for using data-driven design principles to drive viral adoption and in-game purchases better than anyone else.  While this has been great for Zynga's bottom line, rightly or wrongly, they have also become known alternatively as copycats and packagers of additive game mechanics with little or no game play.  The success of these challenge-free games prompted Georgia Tech professor Ian Begost to create a Farmville parody game called Cow Clicker.

     Data-driven decision making taken to extremes can lead to paralysis,stifling of innovation, local optimization, and soulless cash generating machines.  However, using data from hypothesis testing to assist in decision making does not need to lead to these problems.  Organizations need to remember their original hypotheses came from their collective creativity, experience, insight and judgement.  Even though testing might produce surprising and even counter intuitive results, the team must use these same qualities to interpret the results, and to decide whether or not to continue on their current path of investigation or pivot to a new one.