Friday, August 30, 2013

Five Key Factors beyond Software for Analytics Success

Analytics as a trouble shooting turnaround tool for business insights and decisions need to be positioned aggressively for the right kind of ramifications. Besides the effective software tool, what are key factors for analytics success?

  • The whole concept of Analytic depends upon the DATA. The more the data and specific to particular domain helps lot in testing the analytic product and deriving results from that. Without the right data, software can't do much. 
  • Without PROCESS knowledge, models and data are likely to be useless. Model how customers respond to configurable product attributes. Collect data about the product attribute levels and the responses, one needs measurement scales for these fields (these are called "data domains"). 
  • Identifying the right metrics and applying sound cross-validation procedures is a very sophisticated art than a science.Getting the right data, metrics, and cross-validation with the right data are important to analytics success. Fundamental inter-dependence between the variables and linkages of macro-factors with the micro-variables of a given business process are the key decision elements to address business problems. 
  • Analytics Talent: All sort of data talent such as data scientist/analytics, data artist, data solutionary are key success factor, as it takes a human to move beyond the obvious statements that data makes, and provide the subtle analysis that can only be uncovered using intuition and insight, to solve business issues, which means beginning with the end, the end is business value that analytics can deliver.  
  • Define the perspectives on prediction:
    1). Identify several variables that have influences and probably "groups of influences".
    2). These influences could be either direct or indirect or a combination of both in a judicious ratio implying that certain "traits"; so the variables have varying degrees of influences on the outcome variable as well as different levels of inter-dependence with the other predictors.
    3). In the absence of a definitive approach to unearthing these influences on a precise domain, the prediction quality would be of a less than average status.
    4) The traits of each chosen variable need to be outlined and then grouped to arrive at both shared and divergent attribute                                                                                                                     5) Doing well’ in a prediction is not the same thing as doing well in an explanatory model. Answering the thorny business questions is focal point analytics can make difference. 
There are so many choices of techniques and tools to address a given problem,  that the next big game changer in business analytics will possibly not come from another single "breakthrough" technique but a well coordinated data integration, selection of technology, modeling process, defined metrics as well as the qualified analytics talent.


Post a Comment