Disruptive Technology and Unlearning

(Originally published in November, 2000)

As a philosophical bent, it has been my policy to share mainly personal thoughts in this column. But it is time for an exception. I am reading a really terrific book entitled The Weightless Society by Charles Leadbeater. And I want to start out this column with an extended quote from the book –

Learning will be useless unless we can match it with a capacity for unlearning. [my emphasis added]

 

People generally like to learn, if it is enjoyable and relevant to their ambitions and jobs. What holds people and companies back, for example when they are innovating, is their inability to unlearn. Learning is usually seen as an additive process, in which new learning builds on old. This is true most of the time when learning and innovation is incremental. But when the environment suddenly changes, new competitors and products come onto the scene. Then companies and individuals have to adopt new skills and outlooks very rapidly. The key is not to build on past learning but in this case to dispense with it because the old rules of thumb, practices, routines and assumptions which might have worked for the old economy will no longer work when you face new competitors, technologies, and demands from [customers].

 

But unlearning can be extremely painful; it is far, far harder than learning. People like acquiring new skills; they hate feeling that old skills, which a source of their self-confidence and identity, are no longer needed.  For years, I carried around with me all of my notes and essays from my university days. I could not bear the thought that all that learning and effort was redundant. But eventually, I threw them all out, and I have never missed them. Companies face even greater difficulties because for them unlearning means challenging people’s ingrained assumptions, job descriptions, roles, and sense of identity. As well as investing so heavily in learning, companies should set aside resources to invest in unlearning.

 

The process of unlearning inside of companies prepares them to engage in the development of disruptive technologies. And this is something that Finisar (and now Small World Group) keeps trying to do¦ and it is fun. So, the rest of this essay will talk about a couple of these times and what unlearning was required along the way.

The first time Finisar introduced disruptive technology was when we built Gb/s transmitters, receivers and transceivers using low cost 850 nm lasers. Our first products were used in equipment that demonstrated high bandwidth delivery of movies at the Consumer Electronics Show (CES) in January of 1991. These modules were coupled to Gazelle Hot Rod chips and a high speed DRAM array. Later that same year we began selling the transceiver function for about $600 … competitive technology in the telecom space was selling for $5000, so we were well below that cost.

In that time frame, I actually received a call from an early potential customer. As we ended our discussion of the technical aspects of the product, he asked, how much? I replied $600 and he said – yes but how much for just one – and I had the pleasure of saying that was the unit price. If he wanted 1000s the price would be more like $200 for the devices.

Over the next couple of years we worked with two early system customers who designed these devices into back planes of SONET switches. One system used 128 modules per system! This type of use was clearly impossible given devices that cost 10-50x as much as our devices did in those volumes. So, our products caused an unlearning by system designers in these two companies. They decided to employ optical links between freestanding frames and even different shelves in the same frame of the systems.

Before this time, optical links were not thought to be competitive with parallel copper buses of the same total data capacity. A competing technology was the HIPPI standard that had a bulky set of twisted pairs in a quasi ribbon cable format. It could carry a Gb/s but at the price that was high and made worse when the weight, size and overall bulk of the cables were considered.

In contrast, Finisar offered the overall system at competitive cost and much higher convenience with the optical solution. So low cost, Gb/s, multimode optics found wide deployment inside these early SONET switches. And eventually became the workhorse link for Fibre Channel and gigabit Ethernet standards.

Another example where Finisar introduced disruptive technology is in the design of our early test equipment. When we built those first links, we received many complaints from early adopters that the links did not work. This was horrible …

But it turned out that the Gb/s chip sets behind our modules were generally the culprits. So for our own purposes we built some basic test boards with sockets for our modules and sockets for the Gb/s chip sets. We were able to show early systems adopters of this technology that our modules were working fine and that some of the chips sets were either weak or not functioning properly.

These boards could generate Gb/s traffic, perform bit error rate measurements on a link (BERT) and snoop the Gb/s traffic and capture it into buffers for examination and analysis. Simple but really helpful. So these customers asked us would we sell this aid to them.

Finisar entered the test equipment business!

At that time, Gb/s BERTs sold for between 50K and 100K. We offered our first systems for about $15K. It again proved revolutionary. The tool was critical to the development of each of the major Gb/s standards.

But it started another even more interesting trend. Today, when we introduce a new physical link technology … say 2.5 Gb/s or 10 Gb/s low cost links. We NEVER develop the link alone. We develop the test equipment to support this link simultaneously with the link. And we normally introduce both of these products into our markets at nearly the same time.

This happened about 6 months ago when we introduced the double rate Fibre Channel transceivers and our test equipment for the emerging double rate standard in about the same 1-month period.

And it should happen again early next year at 10 Gb/s! (and we expect to whack the 10 Gb/s test equipment standard costs which today cost upwards of $500K and have delivery times of more than 26 weeks … so stay tuned)

By operating this way, that is by introducing physical layer components and the supporting test equipment to BERT and snoop the links we are introducing, we support an emerging standard and so that its development can be timely, interoperability can be tested, system bottlenecks can be identified and overall, the design cycle can be significantly shortened.

We have learned that this overall approach allows us to more frequently introduce disruptive technology to our markets and to have them unlearn lessons that no longer apply.

And that is fun for us all.

Posted in Essays, StartUp Ideas

  1. Allen Taylor says:

    Nice writing. You are on my RSS reader now so I can read more from you down the road.

    Allen Taylor

Leave a Reply