I’m sure you heard of Winston W. Royce, prominent scholar of software engineering , the author of influential article on managing development of large software systems. The article is widely recognized as the ‘bible’ for the waterfall model for software development, despite the fact that it doesn’t mention ‘waterfall’ at all, nor does it describe anything close to what is known as the waterfall model. Unfortunately, this is exactly the impression first two pages of the article make, and, arguably, most readers couldn’t make it past the second page.
However, there is another example of the same problem which, perhaps, has even wider impact on the industry. I’m talking about the notion of hype cycle conceived by Jackie Fenn and productised by Gartner.
On the surface it looks deceivingly simple. Firstly, analysts asses the media visibility of a particular technology. Then, the level of visibility is mapped onto intuitively simple model of hype curve — like this one. As a result, the ‘position’ of the technology on the horizontal axis tell you how mature the technology is. The only thing left, allegedly, depending on your (or your executive’s) appetite for risk, is to decide whether you’re going to adopt the technology, or wait for others to explore the uncharted waters first. At least this is what Gartner’s introduction into Hype Cycle research methodology tells us.
The funny thing is that most people who take the position of a particular technology on the Hype cycle into account when making investment decisions never read past that description. Not surprising, because for 29-pages in-depth overview of the research methodology Gartner charges US$995 — enough to suppress occasional spark of curiosity.
The following is a compilation of my own observations with regard to making sense of Gartner’s hype cycles, backed by the fact that I did read the paper on the methodology. Hope you’ll find it useful.
First and foremost, the ‘hype curve’ phenomenon that could be measured and plotted as a chart does not exist as such. By that I mean that one cannot find the evidence of the fact that a technology has entered, say, the Trough of Disillusionment phase merely by checking the volume of publications or similar accessible measure. It is always an analyst’s call. Hence, it is not an input into an analysis, it’s an output that goes alongside with the narrative.
Secondly, Gartner itself is inconsistent in applying the research method, especially around the presentation aspect. In one publication horizontal axis could be labelled as ‘time’ and in another publication is becomes ‘maturity’, which is clearly not the same thing. Again, it’s not an input, it’s part of the message.
Thirdly, the model implies that the technology is well defined and consistent construct which has a lifecycle. Whilst it could be true in certain cases, namely, when the technology is closely related to a particular market niche serviced by group of vendors, it’s not something to take for granted in general. Consider ‘NoSQL’ theme, or more recent ‘Big Data’ one. This is where it becomes fuzzy and dangerous.
Finally, Hype cycle model is not falsifiable in Popper’s sense. By that I mean one cannot draw any ‘interesting’ predictions that could be proven right or wrong. In particular, no one can reliably predict how long would it take a particular technology to mature by merely looking at some publication figures. Where the model works well is in explaining the current situation and recent events, which, again, means it is a plot device, not an input.
All those observations are fairly easy to explain if we consider the underpinnings of the phenomenon of hype cycle. As Jackie Fenn explains in the methodology paper, it is, in fact, a combination of two more or less independent tendencies. Firstly, any emerging technology gets good initial media coverage, which tends to trail off at the point when it stops being noteworthy for majority of the audience. Secondly, any widely used ‘commoditized’ technology creates an ecosystem around itself consisting of people using it, selling it, teaching it, and looking for talent or advice. Combining these two trends together with some normalization one can demonstrate the emergence of the curve similar to one normally depicted on Hype cycle diagrams (by the way, demonstrating it is a very interesting exercise in system dynamics).
And this is where, I guess, main takeaway point lies. It is counter-intuitive and, hence, useful. By the nature of the underlying trends, technology evolution doesn’t stop as the lowest point of the Hype cycle. In fact, it is the point where it matures as fast as never before. One looking for investment opportunities into technologies, should closely look at the ‘saddle’ of the chart for some fresh ideas.