Newborn wildebeest in Africa have better odds of survival than corporate innovations.
Corporate innovations are often encouraged, funded and grown one year, only to be frustrated, de-funded and shutdown by their parent companies a few years later. We call this “corporate infanticide” – when companies kill their own projects for internal reasons. Sometimes this is for the better, but a lot of times it’s for the worse.
I felt this up close and personal after working on a corporate innovation that was shut down just as it was on the cusp of success. Worse yet, our shutdown happened for reasons that had little (if anything) to do with the business itself. A broader strategic shift happened high up the corporate masthead and the ripple effects were enough to capsize us.
When this happens, bystanders are quick to point out how obvious and inevitable it all was. It may be obvious in hindsight, but it’s a heck of a lot harder to predict in advance. So in 2007 I started working on the idea of building predictive models to predict which innovation projects companies will kill, and which they’ll continue to support. Really? Yep. Why not?
More than a year of migraines later, we had an algorithm that looked for clues about how aligned or misaligned a project was with its parent company and predicted corporate infanticide with over 80% accuracy (99% statistical confidence). These models are different than the other predictive models Growth Science is known for because these focus on the internal workings of corporate bureaucracies, whereas our other models focus on competitive responses and market behavior. In other words, corporate infanticide models aren’t about a startup’s odds of survival, but the odds of an innovation project within the four walls of a giant corporation. In this context, we can now say corporate infanticide is more predictable and, most importantly, with the right steps it can often be avoided. While our algorithm isn’t perfect, we keep working on it and it’s at least “good enough” to give businesses a better grip on an important issue. Sometimes even a small statistical edge can create big advantages; just ask a casino.
Corporate infanticide was an especially tough nut to crack because the topic is so mushy. When a business gets killed by its parent company there are usually official reasons, such as “we’re refocusing on our core,” or “market conditions changed and we can no longer justify the NPV with sufficient confidence.” There are also plenty of unofficial theories such as “the general manager was a complete idiot,” “an executive was doing a power grab and stole the budget,” or “this place can’t get its head out of its own @#$!”. I needed a way to deconstruct corporate infanticide into discrete, objective, testable variables that could form the basis of a predictive model.
I couldn’t, for the life of me, find a workable approach. Thankfully an insight by mentor Clayton Christensen finally led to the big “aha”. Christensen is well versed in Edgar Schein’s work at MIT on corporate culture, which led to this idea (I’m paraphrasing here): in trying to predict corporate infanticide, maybe what we’re really trying to predict is corporate culture. When an innovation is launched from within a parent company, is it something the parent’s culture will embrace or reject? If we could somehow predict alignment with a parent company’s culture, perhaps we could also predict infanticide.
So… how do you predict what a business’s culture will do?
Deconstructing corporate cultures into discrete, testable variables was made possible by the notion that what people call “culture” is just the sum total of many little behavioral rules of thumb. Tribal do’s and don’ts. Cultural heuristics. These cultural rules don’t show up in the employee handbook, but anyone who’s been around for a while can whisper them to you. For example, how much future revenue must you promise if you want any chance of getting a corporate innovation funded where you work? There’s always a secret “hurdle number” at each company – and you’d better have a spreadsheet that gets you there. That’s one example of a cultural heuristic in business. There are also cultural landmines such as “beware of cannibalizing sales from the parent’s core business” and “don’t make the parent’s top customers mad”. You get the idea.
Using this framework, we came up with a long list of cultural heuristics shared by most businesses. I built a database of corporate innovations and populated their heuristic information. Then, after a nauseating amount of statistical analysis, a predictive model emerged.
Proud as I may be of this algorithm, this isn’t about promoting or defending any specific model. Let’s not make this about me. The more interesting topic is about how relatively basic data science and a few key insights can lead to meaningful tools for corporate innovators.
What about you? Have you seen the same corporate innovation tragedies play out over and over in your business? Do cliché internal dramas arise one generation after another? If so, take a moment to realize that what’s happening: you’re detecting patterns. The word “algorithm” may seem exotic, but it’s just a fancy-pants version of pattern recognition. If you’re a corporate innovator, you probably see patterns all the time. The only difference between your intuitive pattern recognition and data science is, the latter takes the extra step of making those patterns explicit and testing them with data. You can do it to!
Yes, it’s a jungle out there. Sometimes shuttering an innovation project is the right thing to do. There could even be times when it’s best to do the opposite of what an algorithm recommends. The goal isn’t to over-conscript the world with overly rigid boundaries. The goal is to make the best possible decisions with the best possible information; and sometimes algorithms can help. Everything counts when it comes to surviving in the wild.