Data to Decision 3 Stages to Analytics LifeCycle
Analytics has become a buzzword in the recent half a decade or so with many enterprises trying to make sense of their data through data ingestion, data pre-processing and munging to uncover hidden patterns and insights in orderto make favorable business decisions.
What is meant by the term Analytics?
What is being termed as ‘Analytics’ today has possibly existed in various forms and shapes over the past two decades, in the form of basic statistics, calculation engines and algorithms albeit in a more dispersedway.
What makes the execution of this very analytics possible today is the advancement in computing machinery (RAM, processing speed, performance), availability of ready-to-go ‘packaged’ libraries and toolboxes, open-sourcecoding languages like R, Python etc. and most importantly availability of large volumes of historical datasets hosted in the local databases of organizations.
What is the need of Analytics?
For a long time, whereas the data was being collected and stored in enterprise databases, the need for buildinganalytics to drive business decisions was not very strong. Only in recent times, this need has become stronger through visibly favorable business outcomes arisingfromthe churning of data through advanced algorithms.
Analytics in various domains
Example, the fields of Retail, E-Commerce have advanced their revenues by leaps and bounds by purely studying the customer buying patterns and recommending items at discounted prices. The fields of Banking and Healthcare have also benefited through the data collection and analytics through credit risk analytics, revenue forecasting, patient segmentation so on and so forth. The list of analytics across each domain today is endless, with every domain using specific jargon to increase the saleability of its analytics offering.
Not only have data-rich domains like Banking, Retail been benefited by the analytics era – traditional domains like Energy & Utilities (E&U) also seem to be on the upswing in analytics adoption. Hardware products dovetailed with analytics solutions have become quite popular in boosting theoverall product sales. A major part of the data in the E&U sector is collected online using a distributed network of sensors. The sheer richness in the datasets and theintricacy of the domainmake the analytics development in the E&U sector quiteinteresting.
Role of domains in Analytics life-cycle:
Domain plays a critical role in the analytics life-cycle. The technical know-how in a given domaincoupled with the proficiencyin algorithms is the right mix of desirable skill-setsin the analytics arena. Data alone might not be able to solve the problem unless backed by a sound knowledge of the domain. Right from the point of identifying business-specific KPIs to buildinga data-driven model to model validation and interpretability, understanding of the domain is key.Further, solving an analytics problem in a given domain requires continuous validation with the Subject Matter Experts towards achieving a holistic and realistic solution.
Stages of Data Analytics Life-Cycle:
The Data Analytics Life-Cycle consists of multiple stages that can be largely classified into 3 key buckets.
- The Business Problem Identification& Data Preparation
The first and foremosttaskinvolves identifying pressing businessproblems within an industry of choice that can perhaps be solved through digitization. Once a realistic and economically viable problem is identified in conjunction with the Domain Experts, it is important to understand its technical feasibility and the kind of data that is available in line with the analytics requirement.For this a quick probeinto the data stored in the enterprise databases may berequired, to see if the data can further be used towards mathematical modeling. In the absence of immediate data availability, it may be essentialto collect or acquire data for the required parameters over a period of time depending on the feasibility of data acquisition. In the case of partial data availability, surrogate models can be built based on correlations amongst the variables in the input space, and the target variable of importance, whose values we are attempting to predict.For this, an Exploratory Data Analysis (EDA) can be performed to understand different independent and dependent variables, their data types and correlations within the data sets.
2. Analytics Developme
The secondstage into the Data Analytics pipeline includes analytics development. In a given domain, different types of Analytics like Descriptive, Prescriptive and Predictive Analytics can be built. The objective of designing anddeveloping a certain type of analytic should be made clear through continuous reviews with the Domain Experts, so as toachievethe desired business outcome or KPI. Example, in the Retail domain; some of the KPIs can includerevenue increase, customer retention, cost reduction etc. and analytics can be built to address these outcomes as target variables based on other available factors as the input variables.3. Output Visualization &Key Recommendations
Once the analytics is built,validated and interpreted in line with the business requirements; it is equally imperative to provide a visually appealing user experience that can translate the outcome in a simplistic yet comprehensive fashion. The development of user-interfaces requires a fair level of expertise in certain front-end application compatible languages like AngularJS, Django etc. In parallel, readily available tools like Tableau, QlikView can be adopted for a quickerand comprehensive display of data through dashboards. Another outcome from analytics can be recommendationsresulting from the study of patterns or insights in the data that can help the enterprises or customers drive better business decisions going forward.
Analytics as a discipline spanning multiple domains has an encouraging future. A step forward into this fieldcertainly requires an analytical leaning to be able to gain insights from the data. Technical know-how of the domain is an add-on to deliver more precise and promising analytics solutions. To conclude, an analytics recipe should have a delectable mix of the ingredients of data, domain,and dashboards to drive key business decisions.