To innovate, first deprecate.

April 7, 2019. Filed under infrastructure 33 systems-thinking 6

Building on Magnitudes of exploration, which discusses an approach to balancing platform standardization and exploring for new, superior platforms, I wanted to also look into this problem from a systems thinking perspective, similar to my previous examination of Why limiting work-in-progress works.


Supporting Innovation notebook is on Github and Binder.

Standardization

Starting our we model a company that doesn't adopt new platforms, instead choosing to place their entire investment budget into their existing platforms.

PlatformInvestment(100)
Platforms(10)
[Potential] > Platforms @ 0
[Potential] > PlatformInvestment @ 5
[Potential] > Progress @ PlatformInvestment / Platforms

As their platforms over time, they get faster and faster over time.

System diagram.

It's not aflashy path to success, but it is a path to success.

Exploring a bit

Changing things up, let's consider an organization that adds more platforms over time, splitting their investment across those platforms.

PlatformInvestment(100)
Platforms(10)
[Potential] > Platforms @ 1
[Potential] > PlatformInvestment @ 5
[Potential] > Progress @ PlatformInvestment / Platforms

Hmm, I have a bad feeling about this one.

System diagram.

Comparing this approach with the standardization one above, things don't work out very well for us, with this naive approach to exploration causing us to spread our investment further and further over time.

System diagram.

Certainly this experience might resonates to some folks, but it's also a bit unrealistic. Who keeps every platform they've ever considered using?

Deprecation

So let's factor in deprecation, getting rid of old platforms as we add in new ones.

PlatformInvestment(100)
Platforms(10)
[Potential] > Platforms @ 1
Platforms > Deprecated @ 1
[Potential] > PlatformInvestment @ 5
[Potential] > Progress @ PlatformInvestment / Platforms

Now we're exactly back to the results we got from standardization, albeit with a bit more complexity.

System diagram.

That said, we hopefully not adopting platforms which are in every way equivalent to our previous platforms.

Innovation

Now let's factor in the idea of innovation, which is to say that we're being careful to only adoption new platforms that are actually better than our existing platforms in some way

PlatformInvestment(100)
Platforms(10)
InvestmentRate(5)
[Potential] > Platforms @ 1
Platforms > Deprecated @ 1
[Potential] > InvestmentRate @ 1
[Potential] > PlatformInvestment @ InvestmentRate
[Potential] > Progress @ PlatformInvestment / Platforms

and things are starting to look good for us

System diagram.

If we compare these approaches, one which focuses on standardization against another which innovates–adding better platforms over time–the winning approach is clear.

System diagram.

However, note that we're being very rigorous about deprecating existing platforms at the rate that we're adopting them, to avoid spreading our investment across more tools.

Hmm, that's a bit optimistic.

Deprecation rate

What if, as is often the case in companies with a decentralized approach to adopting new technologies, we find ourselves adopting better technologies–we're still innovating–but get less rigorous about deprecation?

Let's compare two models here, the first which has us introducing and deprecating at the same rate:

PlatformInvestment(150)
Platforms(10)
InvestmentRate(15)
[Potential] > Platforms @ 3
Platforms > Deprecated @ 3
[Potential] > InvestmentRate @ 1
[Potential] > PlatformInvestment @ InvestmentRate
[Potential] > Progress @ PlatformInvestment / Platforms

The second model is identical but only deprecates two platform each round, instead of quite matching the three new introduced platforms.

System diagram.

We can see that deprecating early is essential to gaining the largest upside of explorations, even if every tool you adopt is superior to your previous selections. Even more damningly, let's compare standardization against slow deprecation.

System diagram.

I think this is a startling result! If we aren't mature about deprecating old technologies when we adopt new ones, then it's better to skip the whole exploration bit and standardize on what we already have.

Back in reality

Looking at these results, note that these are results over one hundred rounds, and assuming we can explore and deprecate technologies over a single round. For most companies, that would mean a round is at least six months, and potentially much longer.

There is still something important to learn from this exercise: : if you want to become a truly innovative company, the two fundamental skills are determining if a new technology is significantly superior to existing ones, and deprecating old technology.