Thread by @oliverbeige

I realize I owe a few tweetstorms, so let me just write one about an application of stochastic multi-stage filtering to corporate innovation, typically called "intrapreneurship" or venture studio" or somesuch. The idea is to replicate a startup ecosystem inside a company.

https://twitter.com/oliverbeige/status/1087312501073682432

Most of these programs are ambitious attempts by companies to rejuvenate or "self-disrupt" their in-house R&D, typically set up as vertically integrated "idea to world domination" funnels. And therein lies the crux.

The starting point is usually a call-out within the organization to submit ideas with a modicum of feasibility research. This is usually the successful part of the venture, at least initially. There are many, many untapped ideas inside a big corporation.

The first decision is usually about whether such a program intends to be an HR program (think Kafka's Chinese Wall) to motivate employees, and as such a cost center, or if the impetus is to challenge the traditional R&D funnel, as a profit center. Here I assume the latter.

Such a model doesn't only challenge its own R&D, it also challenges the Silicon Valley model by trying to vertically integrate the whole process, and by sidestepping the SV "parent shopping" advantage: the same startup can approach multiple VCs but needs only one deal.

You won't be surprised to learn that most companies are in over their heads with these programs.

Some ballpark numbers. In an internal call-out, an intrapreneurship project might collect some 500 ideas of varying quality and elaboration. As a decent rule of thumb, 90% of startups don't survive, 90% of surviving startups don't scale, and 99% of ideas never become startups.

The typical funnel is set up in 3 or 4 steps. Out of 500 ideas, ~50 get selected into a local 2-day session, 10-15 to an "accelerator" type 3-month program, and 1-3 of these to a 2-3 "internal startup".

Since this is a wonderful opportunity to showcase their innovation chops, expect top management to show up at all times for the public decision making events. The shlepp is relegated to the sherpas.

What a scaling startup is is always a matter of dispute, but lets say a 1,000 employee operation that could be sold at a significant multiple of initial investment, if there were such a market. Inside companies there usually isn't. A scaling startup is *not yet* profitable.

With the ballpark numbers I posted, each submitted idea (vetted by multiple experts) has a probability of 0.01 * 0.1 * 0.1 = 0.01% of becoming a scaling venture. With 500 ideas per batch, that's a 5% chance per round to pick a "value ball" from the urn.

If the program is somewhat efficiently run, we get 2 rounds per year, so each year there's a ten percent chance to find a scaling one. These are *not* conservative numbers btw. A typical budget is to the tune of 3 million $/€ per year.

If we assume away learning curve effects, a scaling startup should then be expected to pop up in year five. Typically it happens later bc these programs start out with massively unrealistic assumptions. Even SV VCs have a hard time expressing expectations in probabilities.

If it also takes five years for the startup to demonstrate feasibility at scale (still *not yet* profitability), then we are ten years away from the initial launch, executive buy-in has waned significantly, and the controllers are having a close look at expenses. 🧐

Ten years into the cycle we are also down $30M in program expenses plus whatever incubating the candidate startups costs. And no profitability yet to show for it.

In Valleyspeak, this is called "innovation theater".

[tbc with some design flaws and how to fix them]

@threadreaderapp unroll part one plz, good friend!