Monday, March 03, 2014

What is right and wrong with Christensen's Disruptive Innovation?

tl;dr: There's nothing wrong with the principle but there are two forms of 'disruptive innovation'. One is highly unpredictable and difficult to defend against. The other is highly predictable, can be defended against and therefore shouldn't disrupt.

Two of my pet dislikes are the phrase 'Cloud Computing is an example of disruptive innovation' and how 'Christensen was wrong on the iPhone'. At the heart of this is everything I like and dislike about disruptive innovation theory.

To explain this, I need to go back to basics. First, as covered many times in this blog - components of systems tend to evolve through supply and demand competition. This can be measured over two axis - one of ubiquity and the other of certainty.  The certainty axis was actually derived from a combination of the Stacey matrix and modelling in i-Space (another post, for another day). The upshot of this is that when you look at any activity (or practice or data) the type of publications around it evolve through four basic types (see figure 1).

Figure 1 - Certainty and Publication type

By measuring this change in certainty and by examining ubiquity of an component, it was possible to derive an evolution curve through different distinct phases (see figure 2).

Figure 2 - Evolution
Now, when I normally examine any system I do so over two axis - the value chain (from user need to invisible sub components) versus the state of evolution.  Any and all components (activity, practices and data) within a system evolve from one extreme (the highly uncertain and rare) uncharted space to the more industrialised.

Q. Can't we use diffusion here - from early adopters to laggards? 

A. The process of evolution takes place through the constant appearance and diffusion of maturing instances of the act, practice or data due to competition.  However, diffusion is measured over adoption vs time and both the total applicable market and the timespan may vary with each instance of the activity.

For example, take an activity A with different evolved states - A[1] to A[5] - e.g. maturing versions of telephones. Then if you examine the diffusion curves of each instance of the act, then the total applicable market and the time each instance takes to diffuse from early adopters to laggards can be different between the diffusion curves.

Because of this variability then you can't effectively map over diffusion and in any case, why would you want to?  Diffusion tells you about how things spread, evolution tells you about how things change which is what we really want.

To emphasise this, for the same activity (A) through its different evolved states - A[1] to A[5] - then figure 3 provides an example diffusion curves for each instance, figure 4 provides the evolution curve and figure 5 provides a map. NB figures 3 to 7 are purely illustrative and based upon a fictitious example.

Figure 3 - Example diffusion curves A[1] to A[5] (illustrative)

Figure 4 - Evolution curve A[1] to A[5] (illustrative)

Figure 5 - Map A[1] to A[5] (illustrative)

Q. What has this got to do with disruption?

A. Now, this is where we get to the fun part (sort of).  As components evolve their characteristics change from the uncharted (rare, constantly deviating and highly uncertain) to the industrialised (common, predictable, standard). Furthermore the component itself itself can represent a range of underlying subsystems bundled together.  Hence activity - A - might represent a range of component sub systems designed to meet a specific user need. For example, a telephone contains many subsystems from the physical shape of the receiver to the electronics within it.  Hence each instance can represent an entire value chain of components (see figure 6).

Figure 6 - Each instance can represent a value chain of components (illustrative)

When we talk of evolution of a product such as the substitution of A[2] with A[3] then such substitution tends to be based upon sustaining change i.e. an improvement.  However, you can also get a change in the associated value network. So for example, let us assume that the shift from A[3] to A[4] is a consequence of a change in the underlying components in order to meet some new need through some new property. An example would be physical size becoming important with disk storage. This type of change is extremely difficult to predict because the change is new i.e uncertain. (see figure 7).

Figure 7 - Change in the Value Network (illustrative)

Combined with inertia caused by pre-existing business models (e.g. success of A[3]) then such changes can be highly disruptive and difficult to protect against especially if they first appear in novel markets (where the new property is important) and then develop in that space until the performance characteristics are such that it substitutes the traditional market. This is the classic example of Christensen's Disruptive Innovation whether you're talking about hard drive formats (physical size) to hydraulic vs cable excavators. The reason why it's so difficult to protect against is the change in value network is unpredictable and hence we can't see it coming, there often isn't time to deal with inertia and manage a smooth transition for an existing vendor.

The outcome of RIM vs Apple was an example of such a highly unpredictable change and why the 'Christensen was wrong on the iPhone' is somewhat farcical ... it's almost impossible to predict, it could have gone either way. To be clear, this form of 'disruptive innovation' is extremely hard to see coming, there's little warning that a storm is on its way.

However, the shift from product to utility (i.e. from A[4] to A[5] in the above diagram) is highly predictable. Even the consequences of this from co-evolution of practice to potential reduction of barriers to entry into other value chains can be determined. Naturally we suffer from inertia but we normally have a considerable amount of time to prepare (in the case of cloud computing we've had since 1966 and Parkhill's challenge of the computer utility) and there are numerous weak signals we can use to identify that the change is upon us. If that's the case why do companies get disrupted by it? Well, despite the fact that this storm can be seen from far away, companies still fail to see it because few look at the competitive environment.

Q. So, what's right and wrong with Christensen's Disruptive Innovation? 

Well, there's nothing wrong with at all, it's an excellent piece of work. 

However, the problem is in use.  We tend to describe both the genesis of something (e.g. A[1]), product changes (e.g. A[3] to A[4]) and shifts from product to utility (e.g. A[4] to A[5]) all as 'innovations' when they're not the same.

Product substitution due to an uncertain change in the related value network is highly disruptive because it is incredibly difficult to predict. It doesn't matter whether this is a change in the underlying components or the act becoming a component of something else. Christensen could no more predict the iPhone's success than RIM could and the success of the iPhone was not guaranteed. The only defence against this is a highly adaptable culture.

However, product to utility substitution is a highly predictable consequence of competition and there is no reason for a company to be caught unawares by such a change and disrupted by it. In this case disruption occurs because of poor situational awareness and a poor understanding of the basics of economic change.

So is Christensen's work right? Well, it's certainly a plausible hypothesis and well supported by examples. 

Q. But surely Christensen should have been able to predict iPhone's success? 

A. Absolutely not. It's a highly unpredictable change in the value network which disrupted many due to the inertia those companies had developed through past successful business models.

Q. Is cloud computing an example of disruptive innovation? 

A. Absolutely not. It's a highly predictable change which has disrupted many due to the inertia those companies had developed through past successful business models. There was no reason for these companies to be disrupted other than incredibly poor situational awareness verging on blindness. The use of the phrase 'cloud computing is a disruptive innovation' is more synonymous with 'we've been utterly outplayed by something obvious and we need someone else to blame' than it has to do with classic examples of disruptive innovation.

--- 24th June 2014

Overall, I happen to support Christensen's hypothesis of disruptive innovation with the exceptions that the word 'innovation' has become a catch all and is fairly meaningless and ditto 'disruption'. Not all things classified as 'disruptive innovation' are equal e.g. RIM vs APPL is not the same as Cloud. With the former disruption is reasonable, with the latter disruption should not occur. Some refinement is therefore needed in my opinion but alas, as with many things, 'disruptive innovation' is in danger of becoming a catch all phrase and losing its meaning.

--- 4th Sept 2014

Tidied up some of the language in the post above.