The Diffusion of Innovation, Rogers (1962)

Rohit Khare khare@alumni.caltech.edu
Fri, 27 Dec 2002 13:21:15 -0800


(A brief posting on a very influential set of ideas -- thanks to J.=20
Tigue for bashing me over the head with it repeatedly :-)

This is a slide from an early KnowNow pitch in May 2000:

> KnowNow is a _Viral_ Platform
> --> We are bringing publish-and-subscribe to the Web
>
> Everett Rogers=92 Diffusion of Innovation model (1962):
> * Relative Advantage
>     vs. single-notifier solutions
> * Compatibility
>     vs. novel and/or proprietary protocols
> * Complexity
>     simple extensions, complex router
> * Trialability
>     zero-install scripts for client, server
> * Observability
>     instant (wireless) gratification

The book is at: http://www.amazon.com/exec/obidos/ASIN/0029266718

This is a classic in sociology. Ever wondered where "early adopter" and=20=

the humped adoption lifecycle model originated? Or the=20
"boundary-crossers" beloved of the small-worlds network aficionados? Or=20=

what came before Christensen's "disruptive innovation" model? Or just=20
wanted to break out of the narrow mold of computer-industry innovation=20=

cases? This book has all of that, generously illustrated with examples=20=

from the Green Revolution, medicine, even politics IIRC.

I recently wrote to another journalist trying to recap why the DoI=20
model is so much more important to the success or failure of startups=20
than absolute scientific advantage. In it, I tried to recap how KnowNow=20=

emphasized those factors:

* Relative Advantage. How is this innovation better? How much better?=20
[Company X] seems to say, 'we're 10x more efficient w.r.t network=20
bandwidth', but I'd counter that the biggest obstacle our nascent=20
segment faces is retraining programmers to think in an event-based=20
style -- efficiency w.r.t. software development costs. [Rather than=20
provide yet more] libraries to compile against, KN chose that as the=20
weak point. We allow any program that can send or receive network=20
messages to play -- no KN APIs required.

* Compatibility. How different is this innovation from the one it is=20
replacing? [Company X]'s solution is far from a drop-in replacement for=20=

accelerating TIBCO or IBM MQSeries -- and that's wise, frankly. Again,=20=

my own bet is that it is more important to be compatible with the way=20
developers think about events (KN appears to masquerade as Java JMS=20
events, JavaScript mouse events, Excel spreadsheet change events, and=20
so on) than to be compatible with any one middleware product.

* Complexity. How difficult is it to understand how this innovation=20
works? The "enterprise-class" pub/sub engines are multi-million dollar=20=

behemoths frankly *designed* to be complex (to justify those=20
pricetags!) I know I didn't do a great job of this on the phone this=20
morning, but to another Web programmer, what KN does is very simple (we=20=

turn the Web into a two-way connection that lets you watch as other Web=20=

pages/resources change in real-time).

* Trialability. How much will it cost me to try this innovation out?=20
Not just in the "free 30-day trial" monetary sense, but as a=20
hassle-factor. KN's low-end, for example, is an open-source, free to=20
use plugin CGI script that quickly modifies any Web server to become a=20=

two-way server (mod_pubsub, available on SourceForge). Installing=20
hardware routers and dedicated servers is clearly at the other extreme.

* Observability. Can I tell that this innovation is working? The=20
earliest research Rogers derived his model from was the Green=20
Revolution, tracing the propagation of new seed varieties. Actually=20
seeing greener, taller, and more fruitful plants is an easier sell,=20
than say, vitamin-fortified rice, whose benefits are long-term and=20
subtle. In the computer industry, this is called "demoability" and KN=20
chose to develop several very demo-able products on top of our=20
platform. When I mail you an ordinary Excel spreadsheet or Web page=20
URL, and you can see it changing in real-time, you can observe the=20
difference of KN's event router. [Company X], for example, makes=20
certain solutions faster, better, cheaper, which is harder to=20
communicate than simple feasibility ("our product makes the impossible=20=

possible -- nay, easy!")

Anyway, this is less an essay about Rogers than a bookmark, and a=20
suggestion that y'all go out and have a look at the original.=20
Recommended.

Best,
   Rohit Khare