Analytics Framework- A pocket guide to your analytics strategy

"The goal is to turn data into information, and information into insight"- Carly Fiorina

This quote on analytics caught my attention when I was proposing an analytics framework for a customer of mine in the US. Though defined for an organization, this framework can be scaled to other organizations as well.

This article is a walk through of the defined analytics framework.

Background,

As part of their improvements initiatives ,this customer had setup a core metrics team across the organization, the team had setup very good process to capture and update IT metrics but these metrics were not leading into any control plan for improvements.

I proposed Analytics, to start converting the metrics numbers into information and then start targeted improvement activities, the very next reaction from the customer was if they should be using big data and all possible technologies available for analytics.Well, if you ask me,conceptually analytics is meant to deep dive into data, observe it, study it and the take necessary decision accordingly. If a basic excel sheet can provide that then we should get started with it while we move into more advanced levels.

The Framework

The analytics framework starts from basic analysis to advanced prescriptive level.

The heart of the analytics framework is the four types of Analytics plotted on two axis, along X axis is the degree of value proposition which is increasing to the right and along Y axis is the degree of complexity of the type of analytics increasing vertically.

Data is the key input to analytics, the data can be structured, unstructured, the source of data could be in simple excel sheets to complex database in different source systems. 

Analytics need to have an objective or a problem statement and hence it cuts across the blocks. 

Descriptive Analytics

Analytics starts with Descriptive Analytics,Descriptive Analytics gives opportunities to understand the data, the patterns in the data and make meaningful observations and interpretations.You can do further analysis on correlations of data sets and arrive at interpretations.Here you develop different hypothesis on the problem statement you are addressing and make use of the data to test the hypothesis. 

Diagnostic Analytics 

The next block of analytics is diagnostic analytics, in this block you do diagnostics on why the problem occurred, you can do this by applying different techniques like RCA, 5 whys etc , You also come up with action plans to address the problems 

Predictive Analytics 

The next block which is more complex and which gives more value proposition is the predictive analytics, Here you predict the future outcomes or predict a particular behavior.You can use basic simple linear regression equation to complex ML algorithms.This can be implemented on more complex data sets like unstructured data. For example NLP ( Natural Language Processing) uses unstructured data for analysis.

Other type of ML alogrithms are Naïve bayes alogorithm etc

There are different technologies that can be used like R, Scala etc 

Prescriptive Analytics 

The final block of the analytics is the prescriptive analytics which is very similar to predictive analytics but it gives a prescriptive angle to it.

For example if you have a model to predict a system failure, prescriptive analytics should provide possible solutions to fix the failure.

Framework Implementation 

There are three main components in any kind of implementations, People, Process & Technology.The strategy should in detail define each of the component.

This framework can be implemented on a whole or as individual blocks by using four phases. You start with 

• Define- Defining the problem statement
• Gather- Gather the data, cleanse and process the data
• Implement- Implement the proposed block, for eg- if you are doing predictive analytics, you build the prediction model and implement it
• Control- Control the outcome and the resolution of the problem 

Conclusion

In summary this framework will provide a high level input to an organization's analytics strategy. The strategy should be further supported by a good tactical and operational plan.