Enterprise Decision Management - a Weblog

« Shameless Commerce - Come to InterACT! | Main | Data Mining Survey »

April 26, 2007

Tom Davenport and the business analytics concours

Some colleagues recently attended The Business Analytics Concours Research Summit at the Babson Executive Conference Center - Wellesley, MA. I bring this up both because I think it is an interesting program and because Tom Davenport and Jeanne Harris were talking about their research project, "Managing Business Processes Analytically". Tom shared results of research he and Jeanne have done on automated analysis and decision-making. One of the primary ways to leverage analytics in business processes is to embed analytical decisions into the process flow itself. I have reviewed their excellent book "Competing on Analytics", which discussed this some, and a previous paper they wrote "Automated Decision Making Comes of Age". While this use of analytics, to automate decisions in processes, is not the only one Tom and Jeanne discuss, it is my favorite (of course).

Tom defined automated decision-making and contrasted this "industrialized" model to the old "craft" method of interrupting a process while a decision-maker communicates with an analyst who gathers and processes the data and returns the results for the decision-maker to use. He then described the "why" of automated decisions including better, faster, and more consistent decisions with less dependence on scarce and often costly human expertise. And he provided an array of examples of "what" decisions have been automated across a variety of industries before concluding with discussion of the success factors in both process and technology management, including the integration of analytical and transactional systems.

After his presentation the members of the concours discussed it and these key points came up:

• The danger of inertia
• Once decisions have been fully automated, they may be forgotten about and turn into bad decisions as business conditions change.
• Analytics need to be reviewed and refreshed as a regular part of process management.
• Embedding analytics typically leads to more branched process flows
• Some decisions fully automated
• Some automated with human review and possible override
• Some performed by humans with the benefit of "guided analysis"
• Some left fully up to the process performer.
• Automation of decisions can avoid a human tendency to play hunches
• Hunches are often based on incomplete data and analysis as one organization found with currency positions
• Partial automation of decisions can help decision-makers
• They can be helped to see more possibilities and avoid unnecessary errors, as one health care organization did for the process of prescribing drugs and ordering tests.
• Embedding a lot of analytics and outsourcing a lot of analytical work can reduce the organization’s opportunity to develop analytical skills

All of these were good points. The danger of inertia is what makes Adaptive Controlso important. The danger of hunches and the power of analytics to counteract them was something I brought up when reviewing Blinkand in David Ullman's book Making Robust Decisions.Sometimes the right approach is partial automation of a decision to support someone but, regardless, you need to treat decisions as a corporate assetandplan the use of decision technology to act as a platform for bringing analytics into processes . For full automation you need to think micro not macro when it comes to automating decisions. The analytic skills point made me wonder, again, if analytics should be centralized or distributed.

Wish I had been there...

Technorati Tags: adaptive control, analytic application, automated decision making, competing on analytics, EDM, enterprise decision management, micro decisions, predictive analytics, business analytics concours, Tom Davenport