Monthly Archives: November 2015

Start with MVS or MAS, not MVP

Ash Maurya of Austin, TX based “Lean Stack”, describes lean startup as occurring in three stages. The first is customer discovery where you are trying for problem/solution fit. “Is this a problem? Is this a solution for the problem?” The key question there is “Can you get people to act towards a suggested solution around a properly defined problem?” I think sign ups and click throughs and qualitative measures of interest are good indicators or metrics for this question.

The second is customer validation, in which you build a product and try to get customers to validate that that product actually brings the solution to the problem. This is the first time the product shows up. Product/Market fit.

The third is creating new customers. You’ve found the balance, now you have to go turn more people into customers and grow.

So, really the most important first step is to determine whether or not you’ve found an actual problem. There could be a lot of solutions for that. So, at this stage, it’s not all that important to pick the “right” solution. Really anything that realistically suggests it can solve the problem will do. Once you’ve been able to clearly define the problem and have some traction, then you can start tweaking the product and find the “best” solution.

So, if this is the case, then it makes sense to start out with the cheapest solution. Minimum Viable Product, at this point, doesn’t actually do the concept justice. It should be Minimum Viable Solution, or, to put it in terms of a bootstrapping company, Most Affordable Solution. MVP puts too much focus on product design, and not enough focus on problem discovery.

A Business Plan is not a Treasure Map

If you’re not up-to-date on some of the current thinkings on business plans and the like, let me run through it really quick for you. In the old days, even up until I graduated Uni in 2011, a business plan was a multi-page document (30-40 pages) that you would use to plan out your business in extremely fine detail. You would analyze the product you wanted to build. You would analyze the market you were trying to penetrate. Your customers. The business climate. Etc. It was used both as a roadmap, but also as a way to try to convince banks and financiers to loan you the money to start a venture.

This doesn’t fly with startups, the term startup here to include any business venture whose proposed value is based on introducing some new feature to the market, whether that’s artificial intelligence or just basing a well-known industry around better customer service.

Business owners and entrepreneurs like Steve Blankenship and his student Eric Ries, among others, have come to understand that the key job of a startup is to test whether or not an idea has value. If you’ve ever heard the term “Minimum Viable Product” or MVP, that’s what this comes from. What’s the least expensive thing you could build or create that would demonstrate that your liquor popsicle/adult diapers/fantasy chess league is valuable and interesting enough that people will pay for it.

This solves a big problem that a lot of large companies in the past dealt with. They would take a business plan and borrow a bunch of money and execute it AND THEN go to market. Only when they arrived at the store with millions of dollars in debt would they discover that their Ford Pinto was a godawful car. The entrepreneurs of today posit that if you can discover that the Ford Pinto is a terrible car for a million or two instead of hundreds of millions, you’re actually saving time and money.

This leads me to a confession, whether by nature or nuture, I somehow, knowing everything above, still tend to think about the MVP as if it were a Treasure Map. To be more precise, when I sit down to do my one page business models/plans, I treat them as if it’s important that I deduce the direct route to the treasure. The problem with this, of course, is THERE IS NO PATH.

No matter how clever I am, I can’t decide the twentieth step in the plan by extrapolating from the first or second, because the entropy of clear information is too great. By the time I get to the twentieth step in my head, ANYTHING I DESIGN WILL BE BASED ON A REALITY THAT DOESN’T EXIST. The most insidious part of this kind of thinking, is that I tend to wait to do things until I’m sure I’m right, which, as you can guess, means nothing gets done.

This is my challenge in the weeks ahead. I must treat business plans as nothing more than a baseline on which to measure the truth of my most immediate ideas. I must remember it’s a HYPOTHESIS for the here and now, and not a unified theory of everything. You can’t fight Ronda what’s-her-face by guessing when the fight’s going to end, or she’ll punch you in the freaking mouth. Survey the landscape ahead, for sure. Know the terrain of the battlefield. But, fight the battle in front of you, not the one in your head.

Value Potential Graphs via Machine Learning

After watching a few videos on machine learning, I feel like there’s a very, very large opportunity in using it for what might be known as “job placement”. I put that in quotes, because I think in the future that concept as we know it now might be pretty foreign, because I don’t think it will be divorced from education, and I don’t think it will be as structured as it is now.

It seems that whenever we, and this is potentially me projecting here, discuss something like job placement it has a very structured, top down feel. A central body evaluates you and points you somewhere else. However, if there were instead a machine learning network in place, you could perhaps do monthly little learning boot camps and provide feedback on how you enjoyed it, and be served feedback on how quickly you picked it up. That would turn an aspect of your skillset into labelled data allowing your self to be further sorted towards value activities of best match.

Machine learning isn’t wholly a top down assignment. Semi-supervised machine learning is internally assigned values, spot-checked by participants, surfacing patterns in the data and giving you a better idea of where you can offer the most value to the world.

What if “higher education” was simply an iterative process of labeling your progress in an internally consistent machine learning network.

Then, imagine this in a world where many companies had their hierarchy mostly distributed. That means no HR department. That means a “Value Potential” graph introduces you to someone at a company and you make a contract with that person to do work for their company. This kind of completely distributed corporate hierarchy exists, and I can only imagine it will grow in popularity.

Everyone’s afraid of AI these days it seems, but what if our entire concept of what AI can do for us is wholly tainted by top-down way our societies have been managed for hundreds of years. It’s hard for us to think of anything else. But, absent of centralized state control, what can AI order us to do? Really, in that scenario, general AI isn’t a commander. It’s a tool for us to find our local maximum. To gain perspective on where we are in the mix. To turn each of us into the most informed decision makers we can be.