Hype Cycle History on Predictive Analytics

So, it is about that time again. For the past few years, Gartner has released its Hype Cycle for Emerging Technologies in July. I’ve enjoyed watching this report document the changes in this market. I also like to see where the terms that I hear being overused appear on the Hype Cycle: Are they at the “Peak of Inflated Expectations?” Have they fallen into the “Trough of Disillusionment?” Have they fallen off altogether?

I’m not a gambler, but it sure would be fun setting odds on the probability that certain terms will be on the cycle this year; and which ones have fallen off.

I see Emerging Technologies first hand every day. A big part of my job is making sure that we sell and implement solutions that will be successful for customer – we must rise above the hype! I hear people throw around the terms Big Data, Blockchain, IoT, Machine Learning, Data Science and Wearables daily. In this blog, I thought that I’d look at what has happened to the term Predictive Analytics on the Hype Cycle over the past few years.

First a Little History:

Back around 2002, I was sitting in a marketing meeting at SPSS, where the executives announced that we were creating a new category in the market: Predictive Analytics. I recall them saying that it resonated well with Gartner. My coworkers and I looked around smugly realizing that the terms we were using “Data Mining”, “Machine Learning”, and “Statistics” just didn’t sound sexy or mainstream enough. We were worried that what we were doing would be watered down – it wasn’t. It would, however, take about 8 years for the term Predictive Analytics to appear on the Hype Cycle.

Predictive Analytics at the End of its Hype:

After making its début in 2010, Predictive Analytics sat in the Slope of Enlightenment and Plateau of Productivity with an expected “less than 2 years until mainstream adoption” from 2010 – 2013. Not coincidentally, IBM acquired SPSS back in the 2nd half of 2009, in time for the 2010 appearance (see: https://www-03.ibm.com/press/us/en/pressrelease/27936.wss. )

See: http://blogs.gartner.com/hypecyclebook/2010/09/07/2010-emerging-tec...

In 2014, Predictive Analytics no longer appeared, while Prescriptive Analytics, which first appeared in 2013, and Data Science were now moving up the Innovation Trigger Stage toward the Peak of Inflated Expectations. See: http://www.gartner.com/newsroom/id/2819918.

The Rise in Data Science:

Around 2013, a few of my peers that previously called themselves “Data Miners” and “Statisticians,” suddenly fell in line and began calling themselves “Data Scientists.” While the skill sets that are required to be a “Data Scientist” were and are all over the place, it has become a common job title, with many colleges and universities offering degrees and certificates. Note that “Data Science” was first deemed to be a sexy career choice in late 2012 ( https://hbr.org/2012/10/data-scientist-the-sexiest-job-of-the-21st-...). A great primer on the emergence of this term, it seems to have appeared first in 2001, can be found at: https://www.forbes.com/sites/gilpress/2013/05/28/a-very-short-histo...

Why did Predictive Analytics Lose Dominance?

The irony of the term Prescriptive Analytics, defined by Wikipedia as “extends beyond predictive analytics by specifying both the actions necessary to achieve predicted outcomes, and the interrelated effects of each decision,” is that to use Predictive Analytics appropriately, you should always be looking for the best outcome. For example:

To predict default risk, one must weigh the risks of default to the missed opportunity of giving a loan
To deliver an offer to a customer, you should weigh the trade-off of other offers, versus giving no offer
When developing a forecast, decision criteria needs to be set in whether over or under forecasting is preferable

So, while Predictive Analytics fell off the Hype Cycle, it was likely not really describing what good analysts were doing. Without prescription, prediction just tells you what will happen, not what to do about it!

Thinking Machines:

2014 also saw the growth in prevalence and excitement around “Thinking Machines” – technologies that embed machine learning capabilities to solve particular problems including:

Natural Language Question Answering, which had been on the Hype Cycle since Watson’s win on Jeopardy in 2011, had moved very predictably from Innovation Trigger to the Peak of Inflated Expectations (it is moving down the Trough of Disillusionment, as of 2016).
Autonomous Vehicles took a similar path, appearing first in 2012 and more rapidly hitting the Peak of Inflated Expectations by 2014 (it has maintained this position through 2016).
Smart Advisors and Smart Robots first appear on the scene.

Data Science for the Masses

What happened in 2015, was even more interesting – the world starting thinking about these capabilities in 3 ways:

Analysis for the masses - Data Science was replaced by Citizen Data Science(moving up towards the Peak of Inflated Expectations) and Prescriptive Analyticsappears to have morphed into Advanced Analytics with Self-Service Delivery (at the Peak of Inflated Expectations). These offerings were purported to provide technology easy enough for the common person to use. I’m unclear what the difference between the 2 of these are, but this was around the time that IBM started heavily promoting Watson Analytics and SAS began marketing SAS Visual Analytics and Visual Statistics.
Analysis for “Data Scientists.” Machine Learning was on its way down from the Peak into the Trough. By the way, this term was coined in 1952 by an IBM employee (see: http://acityofpearls.tumblr.com/post/57427420436/machine-learning-w... ) -- not exactly new! It was very surprising to me that this was Machine Learning's first appearance on the Hype Cycle.
Growth of “Thinking Machines.” These systems are purpose built to help humans better interact with the world, leveraging predictive, prescriptive and cognitive capabilities.

Rise in Cognitive

Finally, we make it to 2016, Machine Learning has moved backwards to the Peak of Inflated Expectations. Gartner introduces a Hype Cycle just for Data Science (see: https://thomaswdinsmore.com/2017/02/14/spark-is-the-future-of-analy... ). My how far we have come!

New terms again have popped into the Emerging Technology Hype Cycle, some likely replaced the prior years’ terms, while others are simply more sophisticated iterations of prior capabilities.

Bottom Line: We are no longer are asking humans to interact with the tools, but rather the tools are being designed to interact with the human. See: http://www.gartner.com/newsroom/id/3412017

Smart Data Discovery (moving toward the Peak of Inflated Expectations) is probably a replacement for Citizen Data Science and Advanced Analytics with Self-Service Delivery, where the system does the discovering.
My generic umbrella term of “Thinking Machines” becomes General Purpose Machine Intelligence (at a very early Innovation Trigger Phase, with more than 10 years until mainstream adoption).
Cognitive Systems that leverage Machine Learning begin to dominate the Hype Cycle. This includes: Smart Workspace, Neuromorphic Hardware, Gesture Control Devices, Cognitive Equipment Advisors, Virtual Personal Assistants.

Bets for 2017

I’m excited see what the 2017 Hype Cycle has on it with respect to the Predictive/Machine Learning space. My top bets are: