individual-entry-BA-Blog

« Don't be adversely selected against - use EDM! | Main | Using business rules for Anti-Money Laundering »

Integrating Analytics into Business Processes

In The Four Styles of Integrating Analytics Into Business Processes Gareth Herschel and Betsy Burton of Gartner discuss the four kinds of analytic integration with business process – intra- or inter-application and explicit or implicit. Comparing this to my model of enterprise decision management it is clear that I am talking about analytic integration with an implicit calling mechanism. For the kind of analytic integration I am discussing the inter-/intra- distinction is less important and can be considered a technical decision. Here's how they describe implicit invocation:

"Analysis that is implicitly called by the business application may be invisible to end users, constraining their choices or displaying relevant data as they follow the business activity."

This is how analytics are used in EDM applications – the analytics drive rules or decisions or are part of how a decision is made. They are typically not requested by a user. As far as the user is concerned they are requesting a decision, it is the design of the application that causes the analytic to be embedded. This is not to say that explicit embedding of analytic is not useful – far from it – it is just typically not what I mean when I talk about analytics in EDM.

So let’s consider the two types that seem most relevant when thinking about EDM – these are what Gartner calls “Style 1: Intra-application and implicit” and “Style 3: Inter-application and Implicit” In both cases the analytic services are automatically performed without explicit user prompting but in the first case they are part of the primary application, in the second part of a different application (probably delivered through a service-oriented approach).

The key impacts of the various types are discussed in the paper so I will just make a couple of points that seem to be key for EDM:

  • “Business users see the results of the analysis or rule-driven recommendations as a seamless part of their business applications.”
    This is the core premise of EDM – the role of analytics in EDM is to improve the quality of decision, not to make business users (or customers self-serving etc) decide they need to know something. The attraction of this approach is clear – the need for the user to have analytic skills or the time to apply them is elilminated. They can, in Gartner’s words, “focus on the rapid use of the analysis rather than its creation”. In particular this offers users of the application analytic insight regardless of the user's skill in data analysis and this can be key when considering self-service applications and those aimed at low-level/high-turnover staff such as those in call centers.
  • Analysts working on developing these kinds of analysis have an additional responsibility. They can no-longer say “I built a great model now it’s someone else’s problem”. Instead it is their job to make sure that they understand the context in which the analysis will be used and that the implementation of their model is reasonable and cost-effective. A good model that is actually running in a real system beats a great model no-one can implement.This is one of the ways in which analytics in an EDM sense is different from some traditional data mining/analytic approaches. It really matters how easy and quick and accurate the process of turning the analysis into "code" is.
  • For the IT folks, the model development process often involves huge amounts of data with all that implies for performance. However this should impose only modest constraints on IT as the number of analysts is small and the process of building the models is batch-oriented even if done often. The operational impact of executing the analytics in the system should be fairly small as only rules and executable “equations” will need to be run with high performance and throughput. Again this is a key feature of EDM analytics - the complexity is in the development of the analytic, not its execution. A predictive scorecard is a classic example - lots of data must be processed, complex algorithms applied and sophisticated techniques used but the end result is wonderfully simple to execute.

Companies that have implemented a business rules platform to make decision automation easier will have no trouble adding analytic insight to those decisions using this implicit invocation approach. Companies trying to embed analytic insight into traditional code have only themselves to blame…

First time on the EDM blog?
Subscribe to the EDM blog feed or check out some other recent posts:

TrackBack

TrackBack URL for this entry:
https://www.typepad.com/services/trackback/6a00d83451629b69e200d834cfca9869e2

Listed below are links to weblogs that reference Integrating Analytics into Business Processes:

Comments

Alberto

Jim,
1. "This is the core premise of EDM – the role of analytics in EDM is to improve the quality of decision, not to make business users (or customers self-serving etc) decide they need to know something." - You failed to distinguish between a supervised model and an unsupervised model. The unsupervised model should also let the end user decide that they need to know omething that may not be apparent cognitively.
2. "Analysts working on developing these kinds of analysis have an additional responsibility. They can no-longer say “I built a great model now it’s someone else’s problem”. Instead it is their job to make sure that they understand the context in which the analysis will be used and that the implementation of their model is reasonable and cost-effective. A good model that is actually running in a real system beats a great model no-one can implement.This is one of the ways in which analytics in an EDM sense is different from some traditional data mining/analytic approaches. It really matters how easy and quick and accurate the process of turning the analysis into "code" is." - You are right!!! and that is the difference between the business side being responsible for a data mining project vs. the IT folks. Most businesses delegate the business function of an application to the IT folks...a big mistake.
3. "the complexity is in the development of the analytic, not its execution."---AMEN!! It could not have been said better

One more point: there will be some additional complexity to the IT folks as data mining joins with the human experience (cognitive querying) to create an artificial intelligence.

The comments to this entry are closed.

Search Site


  • dmblog.fico.com

Subscribe

  • enter your email

Upcoming Events

  • FICO Tools & Analytics User Forum 2012
    BERLIN: September 11-12, 2012 LONDON: September 18-19, 2012 Gain new insights for improving business performance through advanced analytics and decision management tools.