Saturday, January 31, 2015

From Big Data to Decision Management

The full data-performance life cycle includes data --> analysis --> decisions --> performance.

Businesses enter the digital era of Big Data, with the business dynamic of velocity, uncertainty, complexity and ambiguity, how to make the right decision by the right people at the right timing also becomes a strategic imperative, because it will directly make an impact on business’s short term bottom line and long time prosperity.

Decision Management is still an emerging discipline. The whole purpose of analytics is to make better decisions based on data (big/small). You can call it with any name like decision management/theory/science/technology/engineering. The typical challenge seen with the traditional analytics approach is to arrive at insights, but not necessarily affect actual business decisions, and not always in a timely manner. Decision management /engineering approach embeds analytics in actual business decision scenario- rather than leaving it to the receiver of insights to use.

The full data-performance life cycle includes data --> analysis --> decisions --> performance. Analytics is means to the end, not the end. However, in reality, there is not enough focus on decisions. So a lot of people get a bit caught up on the analysis as if this is the end of the process: data --> analysis --> conformance. No, if the analysis doesn't lead to performance, it's rubbish irrespective of the apparent eloquence. This actually represents a problematic situation if a statistical approach cannot deliver the expected return; people might start to question the competence of the researcher rather than the suitability of the approach. Also, keep in mind that if a statistical approach is indeed the solution, at some point the researcher is not required; and great many organizations would like this to be the case as it justifies the capital expenditure. Keep in mind any decision-model runs the risk of creating a false sense of precision and confidence.

The role of a decision model is to systematize one's preferences and beliefs and identify their consequences (as specified); thus allowing critical comparison of one's holistic view to the consequences of the formally specified one. If the formal specification is reasonably close to the truth, this critical comparison is very helpful, because whenever you find a difference, you have the opportunity to improve either the intuitions (= an insight) or the model (= fix a bug or improve the logic). When the two points of view are reconciled, both are improved, the model corresponds to the gut feel, and it identifies a choice with a rationale that works.

Framing the analysis in terms of the span from worst to best on each criterion as a decision management practice. Without that, the analysis is working in the realm of tangible measures rather than the preferences it is intended to embody. The “SMART” (Simple Multi-Attribute Rating Technique) decision technique solves this problem by using weights based on moving from the worst to the best in various criteria. There's an art to using it, but it works well, even for a combination of rational and emotional criteria. Another way to look at it is not as two criteria to be balanced, but a single criterion to be maximized, specifically the expected value of the market size. You just multiply to get this expected value. So multi-criterion decision analysis doesn't even apply here.

Improving decision quality is about reducing the uncertainties of the most variable elements. The process of working with decision-makers to support their thinking through is subjective, though, how they judge tradeoffs between choice criteria is more influential on decision quality than marginal improvements in the choice of multi-factor attribute analysis methods. Secondly, presenting forecasts of outcomes in value distribution terms contributes to creating a proper awareness of the reality that in many decisions, good decision making merely reduces the risk of errors, in the face of an uncertain future environment.

Still, analytics is just a tool, like any type of management, decision management is both art and science; thinking fast and slow; it has to well combine the analytics and intuition; information and experience, management and engineering, and manage decision life cycle with effectiveness and agility.


Thank you for providing this blog really appreciate the efforts you have taken into curating this article if you want you can check out data science course in bangalore they have a lot to offer with regards to data science in terms of training and live projects.

Post a Comment