Business Model Innovation Algorithms: Smarter search for New Business Models — Part 2
This is the second post in a series about Search Algorithms for Business Model Innovation
In the previous post, I stated that to be more effective as a startup in finding your next new business model, you need to be better at search. In this post, I’ll explore what that means.
Being better at search
For most startups the search for a working business model is like stumbling around blindfolded trying to hit a piñata. Without even being sure there even is a piñata in the first place. Some startups try to do better and use some sort of methodology to look for the piñata in a more or less organized fashion, but I haven’t seen one yet that uses a really structured type of search strategy.
A trip to the Gurus of Search: Computer Scientists
Let’s start by learning about search. Do you know where search and search strategies have been studied the most? In fields like artificial intelligence and computer science. And those studies are not just used to make Google Search work, or to make sure that your route planner gets you from A to B. Search is used by computer scientists for almost every decision making process, for almost any problem solving task.
To a computer scientist, problem solving at the core is nothing more than searching for the right solution.
The language of search
Concepts and language from computer science that can help to understand searching for business model options better are: the concept of search space, the fitness landscape, the fitness function, local and global maxima, and hill climbing, and heuristics.
To discuss improving search strategies later, it is helpful to internalise these concepts as a way of looking at business model search. For each concept, I also give an example of how it can show up in a typical innovation workshop.
Search Space
Imagine you went to the beach, and after a great sunset you want to go home, only to find out you lost the car keys in the sand. Where can they be? The mental picture you create, that directs where you want to look for the keys, is your search space.
The broadest search space you could imagine for this search is the entire beach. But that doesn’t usually make sense. Mentally, you will immediately make the search space smaller, excluding the places where you definitely did not go.
To make the search for the keys go even faster, it makes sense to start at the places where you think the likelihood of finding the keys is greatest: near where you spent the most time, or where you changed into your swimming gear. This narrowing and prioritising of the search are examples of heuristics, strategies that help you make search more effective.
For a startup, the search space is much more multi-faceted and messy than is the case with a lost key. For a startup, search for a good business model includes finding the right customers, the right value proposition, the right sales channels, and lots of other factors, most of which are not precisely defined or quantified.
The ideal business model candidate is hiding somewhere in the search space, and can be thought of as an optimal combination of these factors.
So, one of the first things you can do when you start a search for a new business model is to define the search space you’re going to search in. What are combinations that you can discard right away? How will you make those decisions?
In practice, in innovation workshops, I help teams define their (initial) search space through defining the design criteria for their business, for instance using this Design Criteria canvas.
Fitness and Fitness Function
The term fitness comes from evolutionary biology. There, the fitness of a certain organism is used to describe its chances to survive in a world with given circumstances, expressed as a single value. In relation to business models, a fitness score is a measure of how well that business model performs, according to criteria you have defined, such as how much revenue the model generates. But a fitness score does not only have to be about money, it could be also defined by how well it adheres to aspects of your vision. If your vision is to be 100% renewable, business model options that are more renewable would have a higher score. You get the picture.
To work towards an algorithm, we will need a way to evaluate and compare business model options, to see which one is better. That means, we need some kind of (objective) fitness function that assigns a score to each business model option on the same scale. Keep in mind that your fitness function will have a huge impact on the outcome of your search, so make sure you design it carefully.
In practice, in innovation workshops I use a combination of rating the design criteria, back-of-the-napkin estimates of cost and revenue, and a ‘difficulty’ rating to calculate fitness values for a lot of models quickly in excel. More background on how to do that can be found here.
Fitness Landscape
Every point in the search space defines a unique business model option. That option will have a fitness value. That means, that every point in the search space has a fitness value as well.
Think of the fitness landscape as a landscape with rolling hills. The altitude of the hills describes the fitness score, and each position in the landscape some set of features in the search space that makes up a business model option. In reality this is would be a multi-dimensional landscape, but since that is hard to picture, let’s stick with a simplified version for the example.
Hill Climbing
To identify the best solutions in this search space, we need a way to identify the peaks. Preferably the tallest ones.
From a random starting point in the landscape, ‘feel’ your way up the hill by always moving uphill, from business model option to the next. Going uphill ensures, using your fitness function, that each new combination is better than the last. In this way, you are guaranteed to find a peak, and therefore a potential candidate.
In practice, in innovation workshops, this means taking ideas for business model options and seeing how they can be tweaked to get a better score — of course always keeping in mind the design criteria defined for the business.
Local and Global Maxima
Now, there could be trouble on the horizon. The hill climbing strategy is great in that it will find an optimal solution, but it is only a locally optimal solution. So, that peak you just climbed may have some bigger brothers further out in the search space that you don’t know about.
That risk is something you can mitigate by first probing the search space. Instead of starting from one random position and moving uphill, we’ll start from a number of positions. That should give us a higher chance of locating the tallest hills.
In practice, this means starting out with a number of different business model options. Some of which are better than others, and they all serve as a probe in the search space.
The next step: creating an algorithm
Now that we have a clear picture of the concepts involved in search, and how they are used in innovation workshops, when doing it ‘by hand’, the next step is to piece them together into an algorithm and seeing what machines can help us with. In the next post, I’ll show an example step-by-step algorithm to search for business models.
