The Rise of Analytics

As my operations-research readers know, analytics has become the word en vogue in the community - the INFORMS Practice conference was recently renamed INFORMS Conference on Business Analytics and Operations Research to reflect this trend, as Mike Trick pointed out in this blog post of his. My non-operations-research readers will be left thinking: what exactly is operations research anyway? Research on operations? That is partly true - OR (as we call it, which makes for some interesting Google queries since the web service mistakes it for "or" (as in "either/or") emerged from the need to improve military logistics during World War II, but has become much broader than that, now representing the broad field of quantitative decision-making.

According to this Wikipedia page, "operations research is an interdisciplinary mathematical science that focuses on the effective use of technology by organizations"; this certainly is the most awful definition of OR that I've ever seen. The second paragraph is more accurate, stating: "Employing techniques from other mathematical sciences --- such as mathematical modeling, statistical analysis, and mathematical optimization --- operations research arrives at optimal or near-optimal solutions to complex decision-making problems."

Few non-OR trained people will naturally come to the same conclusion when first faced with operations research, and the issue of how to call what we are doing is one that we have all struggled with, whenever anyone asks us about our profession. (I stick to: "I do mathematical models for business.") In contrast, analytics has become a much more accepted term in the business community, where books by Thomas Davenport and Jeanne Harris have emerged as market leaders: "Competing on Analytics: The New Science of Winning" (2007) is a landmark book in that respect, and was followed earlier this year by "Analytics at Work: Smarter Decisions, Better Results". These books are published by Harvard Business Press, which certainly added to their legitimacy.

Mike Trick recently posted on his Twitter feed a graph, using a new Google Labs tool called n-grams, showing the incidence of words like "operations research" and "analytics" in books; when this morning a friend and reader of this blog sent me a more complete graph including "industrial engineering" and "systems engineering", I figured it was time for a blog post. Here is the graph, courtesy of Andrew Ross. You will notice that the use of "operations research" abruptly rose in the 1950s and peaked in the early 1960s, to undergo a fast-paced decline ever after. This is not good news for our profession, as "operations research" is part of the brand we communicate to the media and potential business collaborators. Analytics, on the other hand, is currently the most popular of the terms by far, and the trend does not seem to be slowing down by far.

 Will analytics (as a word, not a discipline) turn out to be only a fad? The term is marginally clearer than "operations research", but a lot remains to be desired. For instance, Wikipedia defines it as (and we all know not to trust Wikipedia, but it remains the first site people looking for definitions go to): "Analytics is the application of computer technology and statistics to solve problems in business and industry"; in other words, it makes no mention of optimization. In addition, the rise of Google and its own Google Analytics will indoubtedly reinforce the idea that analytics are about analyzing numbers such as web traffic data, without recognizing the need for mathematical models and optimization software. From the n-gram, laypeople could easily conclude that "operations research" was the fad, and that the bubble burst thirty years ago; however, we have all continued to practice operations research. Maybe it became less interesting to the general public, after the novelty effect wore off - but our field is there to stay, under one name or another.

I've always been interested in promoting operations research and related ideas (hence, this blog), but the issue of raising general awareness of operations research has taken a new relevance for me, as I am the incoming chairwoman of the Public Information Committee at INFORMS. You can read more about that committee here. (If you have ideas on how to raise awareness and understanding of operations research, management science and analytics in the general public and fight misconceptions, feel free to drop me a note.) Given the rise of the word "analytics" and its higher name recognition in the media, it is up to the INFORMS community to make sure the term continues to reflect what we do and that its meaning is not affected by web powerhouses counting page views. I've created a new category for my blog - Analytics - to support the mission of the Public Information Committee and this represents the inaugural post under the new label.

Comments

[Irvin Lustig]

I coauthored a paper with some IBM colleagues, including former INFORMS President Brenda Dietrich, for Analytics magazine that talks about the different kinds of analytics. Operations Research fits into what we call prescriptive analytics, and at IBM, we're using that term to promote our optimization products. See http://analytics-magazine.com/?tag=prescriptive-analytics

[Aurelie C. Thiele]

Prescriptive analytics - I like that! Thanks a lot for the link, Irvin. I enjoyed learning about the descriptive, predictive and prescriptive kinds of analytics. That article deserves to be widely read. Thanks again!

[Paul Rubin]

The article by Irv et al. is a good step in defining "analytics" in an inclusive way. The one thing I don't see (and don't usually see in definitions of OR, for that matter) is what I think is sometimes called "decision analysis" or "decision theory" -- the process of making decisions. In any case, while the descriptive/predictive/prescriptive classification appeals, I'm worried that for most people "analytics" will become more or less synonymous with data mining.

[Fbahr]

I'm sorry that I've to say that I'd rather oppose the idea of sth. like "prescriptive" analytics. While I'm perfectly fine with the definitions of descriptive and predictive analytics, the prescriptive "branch" of the given definition approach doesn't make sense to me at all. Analytics in its broadest sense is commonly defined as "the science of analysis" (kind of) - where analysis derives insights from given structures or data. To me, and with all respect, the term "analytics" is over-hyped these days (and will ultimately follow the fate of all hypes, cf. http://twitter.com/fbahr/status/15706213656301568). That's not to say that analytics as defined in terms of statistical analysis, data mining and "predictive science" isn't a big deal these days - yes, it is, for sure: but not everything has to be redefined in terms of a "analytics" discipline (especially when it's the opposite of analytics: planning!).

[Aurelie C. Thiele]

While the word "analytics" has its flaws (its close association to data mining is one of them), it gives people a better idea of what we do than "operations research". Of course, we don't only analyze things, we improve them too. I like "optimization" but it makes laypeople think we're working on pure math topics. My favorite description of my field, when I don't want to use "mathematical models in business", is "quantitative decision-making", but that is a bit long.

Maybe we should make up a new word.

[Industrialengineertools.blogspot.com]

I added a few more terms the the Ngram Google trend tool.

http://ngrams.googlelabs.com/graph?content=operations+research%2Cmanagement+science%2Canalytics%2Cindustrial+engineering%2Csystems+engineering%2Cdata+mining%2Capplied+mathematics%2Cbusiness+analysis&year_start=1800&year_end=2008&corpus=0&smoothing=3

Notice how much growth in the term "data mining". That definitely accounts for something.

I loved that article from Analytics Magazine on Prescriptive Analytics. That is such a great term to describe Operations Research with relevance to today's buzz.

[Aurelie C. Thiele]

Wow! Thanks a lot for the graph. An eye-opener! I wonder what the recent little bump in the popularity of data-mining means.

Here's another one, with optimization added in (and a couple of words removed) http://bit.ly/e20BFP

Nice to see "optimization" do so well as a keyword.