What is Analytics?
Analytics is the entire process which leads to the discovery of meaningful relationships amongst data.
A lot of firms use “analytics” term for their BI initiatives. However, analytics is a lot more than plain vanilla reporting which falls in the realm of descriptive analytics. It involves the application of statistics, mathematics and computer science to derive meaningful relationships amongst data.
In today’s world, with almost every organization using IT to enable their business processes and for collaboration purposes, we have a rich depository of both structured as well as unstructured data residing in the relational databases and content management systems of the organizations. These datasets when leveraged properly have immense business value. For long organizations, have been taking decisions based on the gut feeling or hunch of individuals which is based on the knowledge and experience they have acquired over a period of time. However, the intent of analytics is not to replace the human cognitive abilities but to aid it in decision making.
Types of Analytics
This requires telling the user the “what” part of their business data. It takes its name from descriptive statistics in which data is represented (described) in various forms – bar chart, pie chart, line graphs etc. This is mostly used to report the data in various forms. Business Intelligence related initiatives fall under this category.
This requires identifying the “why” part of the data i.e. if something happened then what were the factors that led to that particular thing happening. For example, descriptive analytics can show that the sales of a particular company went up in a particular period. However, the company would want to know what the factors that drove that sales up were. The factors driving the sales can be identified by performing analytics on the sales data. For example, a company might have carried out a sales promotions exercise which might have driven its sales up, or its sales might have gone up because of the festivals which happen during a particular part of the year. The sales data, when modeled along with the factors driving the sales, can give the sales manager a clear idea regarding which factors influence the sales at different times of the year.
This can be considered to be a form of supervised machine learning. Here, we use historical data to create a statistical model and then use that model to predict the future. For example, predictive
analytics can be used to model the shopping behavior of a consumer and give him/her a shopping propensity score. Based on this propensity score relevant offers can be sent to those consumers whose shopping propensity score is above a certain threshold.
Any problem in the realm of business analytics will require three sets of skills:
- The ability to identify the business problem
- The ability to identify the data which is relevant to solving the business problem at hand
- Identifying the right algorithm which needs to be used to solve that particular business problem. In most of the cases, the algorithms preexist, the person just needs to identify which algorithms can be used to solve a particular problem. However, in some of the cases, the person might be required to write an algorithm to solve a particular problem.
Getting the right data
Getting the right set of data at one place is one of the prerequisites for leveraging data analytics completely as much depends on the availability and quality of the data that we have.
Given that this is a fairly new field and customers normally focus on return on investment for any new initiative, adoption would depend on how the project sponsors/organizations view it. There are two ways in which the project sponsor/organization can view it. One way the organizations can see it is that they view it as a necessary investment as their competitors are also investing in analytics capabilities (can be in-house or sourcing it from a vendor) and therefore they do not want to miss out on any tools/capability which helps them compete with their competitors. The other way the organization might see it is that they expect a certain return on their investment from the analytics capability and that return has to be greater than their cost of capital. The adoption of analytics depends much on the executive sponsorship for such initiatives.
Getting the people with right skills
It is very difficult to find the set of skills required for performing data analytics on one person. So, normally, any team trying to solve a particular business problem in business analytics space will need a guy who has a sound understanding of the business problem he is trying to solve and he will specify the data that needs to be fetched for performing analytics, a technical guy who can query the data required to perform analytics and has sound understanding of various algorithms which are pre-existing and has the ability to transform it to a form on which algorithms/statistical techniques can be applied.
Also, a number of advanced analytics tools are available in the market. The analytic needs of the end-users drive whether we need to use those tools. Alternatively, open source language like R can also be used to performing analytics on client data to aid the client in decision making.