Lessons From Clayton Christensen: The Software Innovator's Dilemma

Written By: Dharmesh Shah May 10, 2006

I am a very big fan of Clayton Christensen’s work.  He is the author of “The Innovator’s Dilemma” and “The Innovator’s Solution” both of which are on my recommended reading list for startup entrepreneurs.  He is also a professor at the Harvard Business School.  In my opinion, Professor Christensen is one of the leading management strategy thinkers of our time and there is a lot to be learned from his thoughts and ideas.

 

I had the distinct honor of attending four private classes taught by Professor Christensen to the current class of MIT Sloan Fellows, of which I am a member.  This was a “closed session” with about 50 students from around the world.  Christensen gave a series of lectures focusing on both his prior work and some new research and analysis that he is currently working on.  My sincere thanks to Professor Christensen for taking the time out of his exceptionally busy schedule to present to us.  This was immensely helpful.

 

Disclaimer:  Since I’ve read the books, attended the presentations above and debated/discussed the topics so much,  I think that some of his concepts are now “hard-wired” into my brain.  As such, I you should assume that all the brilliant stuff in this article are his ideas and the feeble attempts at describing some of the concepts and applying them to software startups are mine.  This is a pretty safe bet.

 

Credit also goes to Brian Halligan, who is a good friend and co-founder of my latest venture.  Brian and I have discussed the concepts here extensively and he essentially co-authored this article with me.

 

The Concept Of Continuous Innovation

 

In any product category and any industry, there is a “performance” curve that is based on some primary dimension of performance.  Said more simply, there is some attribute of the product that customer’s value and the offering from companies tends to move “up the curve” based on that attribute.  In the car industry, one performance curve might be fuel efficiency.  In the enterprise software industry, one possible performance dimension could be features/functionality (this is a common one) or scalability.

 

Lets take a look at an example.  If you are Siebel (now owned by Oracle), and it’s the 1990s, your primary performance curve for your leading CRM product could be “features/functionality”.  One of these features is the customization that can be done for each individual customer.  So, as Siebel evolves its product “up” that performance curve, there are more and more features added.  It’s important to note that Siebel management doesn’t wake up every morning and think:  “What kind of features can we add to the product today?”.  More likely, the features are being added at the request of Siebel’s customers.  Christensen would say that it is good management that listens to its best customers and improves the product in ways that customers care about.  This also happens to move the company towards better profit margin business.  There is noting inherently wrong with this behavior.  Listening to customers is a good thing.  These types of improvements, along the existing “performance dimension” is what Christensen calls “continuous innovation”.  Its basically moving the offering “up” the existing curve.  So, when Siebel adds a new feature or capability to its product, it is conducting “continuous innovation”.  

 

The Problem With Continuous Innovation

 

A simplified description of the “dilemma” that management faces is that as it continues to listen to its best customers (and improves the product on the selected performance dimension), it eventually exceeds the average customer’s needs (and likely the needs of the market overall).  Over time, the product does too well along the primary performance curve.  In the example of Siebel, it is likely that most of Siebel’s customers didn’t need all those features or capabilities.

 

This is where startups come in (and where things start to get interesting for you, the software entrepreneur).  One way to “disrupt” the existing game is to focus on discontinuous innovation.  Basically, this means that you pick a different performance dimension (and hence a different curve) than the one the current players have picked.  To extend our Siebel example further, enter Salesforce.com.  Whereas Siebel was competing on features, scalability and customizability, Salesforce.com played a different game.  It picked “simplicity and immediacy” for its performance dimensions – not features.  Instead of trying to beat Siebel at its own game (at which it would have likely lost), it decided to play a different game – and start a different curve.

 

The results of this kind of behavior have a set of patterns:  When picking a new performance dimension (and ignoring the existing performance dimension that the market already knows about), chances are it will be a different set of customers that value the new attribute.  Said differently, whereas Siebel’s best customers valued features, scalability and customizability, the early customers of Salesforce.com didn’t.  At least, not as much.  One effect of this is that the existing Siebel client-base likely couldn’t care less about the Salesforce.com offering – it was inferior based on their needs and based on the performance attributes they cared about most.  As such, Siebel continued (for the most part) to do what it did.  Listen to its best customers, focus on delivering to their needs and leaving the “low margin business” off smaller customers to the new entrant (Salesforce.com).  This “low margin” business (customers that value something other than what you’re good at) can often be a large pool of under-served (or non-served) customers.  

 

Now, a funny thing starts to happen as the new entrant starts delivering value to its new customers (which historically, were likely not Siebel customers anyways).  Over time, the product improves on the newly selected dimension of performance (simplicity and immediacy) but then also starts to match the needs of the average customer on the old measure of performance.  So, as Salesforce.com evolves its product and gets new customers (that Siebel didn’t really care about), the product eventually improves to the degree that some of Siebel’s “lower end” customers that were being grossly “over-served” by Siebel’s product become interested.  The new attributes might be interesting to them (simplicity), and there might be other areas of value (lower price).  So, these customers begin to look at the new market alternative and start shifting to Salesforce.com.  Not all at once, but slowly. 

 

This story continues whereby as Salesforce.com moves further and further up its performance curve, it starts to incidentally be meeting the needs of some of the old Siebel customers.  Eventually, even some of Siebel’s best customers are compelled to look at the alternative because not only is the product delivering most of the features/functionality they’re looking for, but there are other attributes that make the switch compelling (lower cost of ownership, better usability, etc.)

 

But, the story does not end here.  Now, as Salesforce.com continues up its own performance curve and succeeds, it eventually becomes the next Siebel.  Now it starts listening to its “best” customers and focusing on business that gives it the best revenues and profits.  Why shouldn’t it?  That’s good management.  Hence, the innovator’s dilemma.  The same phenomenon that allows Salesforce.com to capture Siebel’s market begins to make it vulnerable to the next startup that comes along and finds yet another performance dimension that a new class of customers care about.

 

Summary of the lesson:  It is generally a dangerous strategy to play the same game as the existing players by focusing on the same performance dimension that they have already gotten very good at.  As a startup, if you are to succeed, you have to play a different game by finding a different performance dimension that a different group of potential customers cares about.  If you’re doing it right, your offering should be demonstrably inferior in the primary performance dimension that the current market already understands and values.

 

In the next installment, we’ll look at ways that these kinds of innovations can be approached in the software industry for startups.  Stay tuned…

 

 

Related Posts