What is Analytics?
Analytics is the professional field of discovering insightful relationships amongst data. The insights can be used to control or predict outcomes of processes or systems of any type based on what has happened in the past. When the data is simple, we can figure out causes and effects intuitively. However, in real world situations with numerous potential influences on an outcome, such relationships or correlations are less obvious or easy to determine. This is when analytics becomes useful.
You may ask how analytics is different from basic statistics. Statistics tends to give us information about one or two parameters at a time, e.g. median household income or the whether seat belt usage is correlated with the number of highway fatalities. These individual statistical measures do not account for what is causing median income to be what it is or whether the effect of seat belt usage is eclipsed by airbags and hidden by higher traffic volumes.
Unlike statistics, advanced analytics tools can find trends, causes, and predict outcomes irrespective (in theory) of the number of parameters considered. Respectively, the methods that do these are classified as Descriptive Analytics, Diagnostic Analytics, and Predictive Analytics. Descriptive Analytics, also known as Business Intelligence (BI) is the calculation and presentation of metrics (from raw data) in a dashboard format. Analytics is dependent upon the amount of history we have to analyze, and so the limiting factors are data storage, retrieval, and computing power.
Recently these factors have reached a critical mass of speed and affordability. Organizations are continuously creating or recording a myriad of data from social media activity to weather conditions to manufacturing operations metrics. Hence, the buzzword: “Big Data”. With Big Data, the great possibilities and power of analytics for practical use have emerged.
What can Analytics do for business?
It’s a misconception to think that analytics is only in the realm of large corporations with deep pockets. Analytics can be used at various levels and to varying degrees to create value and competitive advantage.
If you have a knowledge gap with respect to a process or system, then analytics may very well be able to help. You may want to know what factors are the biggest contributors to your operating cost; or which customers are most likely to buy from you again. Closing such gaps will enable the making of better decisions, like which overheads to prioritize on controlling or which customer to cultivate better relationships with. All this to the end of a more favorable outcome at the top line, bottom line, or both.
Even, if you don’t have a specific pain point to address, you may be collecting data within various functions or departments. If that is the case, you may able to find opportunities to exploit after seeing how all that data is connected. However this does not happen automatically - there is a well defined process which can help you get there.
The Analytics Process
First, you have to extract and ‘clean’ your data. This can be as simple as compiling or consolidating some MS Excel files or as complex as automating access to a data warehouse with open source tools like Hadoop. ‘Cleaning’ means to ensure there are no formatting errors or accounting for missing elements. The data is run through an analytics tool or process that is chosen based upon the objective or intent of the work done. For a basic BI application, the outcome can be a set of insights such as Key Performance Indicators (KPIs) shown on a dashboard. The advantage of this is that you can monitor status in real time as your measured data updates as well as track how your KPIs are changing. However, you could go further to build a model, based on your data’s history and into which you can put hypothetical conditions to observe possible KPI outcomes. Otherwise, you can extrapolate current conditions based on trends to see where your KPIs are headed. The great advantage of this, is that with newer more recent data, the model may be updated, so it never becomes obsolete.
In the next article of this series, we will outline and explore what to expect when you apply this well-defined process to your data.