Data Analytics - where it is heading?

What is ‘Data Analytics’? In most simple terms, it’s about gaining insight from data. World, is facing an explosion of data from more disparate and diverse data sources than ever before. Enterprises are increasing shifting towards intelligent and automated data-driven decisions. An Organization needs to adopt an analytics strategy that is adaptable to changing business environment. It should consider organization’s infrastructure, current analytics maturity, manage wider sources of data (internal, public, social etc.) and keep pace with the velocity of data.

Primarily, there are three type of analytics strategies –

Descriptive - ‘What Happened?’ Descriptive analytics is the oldest and associated with consolidating the data. This is synonymous to adhoc reporting, Dashboards, Trends, Pre-defined KPIs etc. It is primarily mining past data and understands reasons for success or failure. Descriptive analytics identify relationships and groups the data.

It’s about comparing what happened in past to what is happening now. Descriptive analytics is not only an important source to understand past but also a stage for preparing and exploring data for various predictive models. Typical examples could be demographics (age, Gender, region etc.), Transaction history, sales trends, top five devices using a particular app, Top three topics discussed on social media etc.

Predictive - ‘What will happen?’ Uses various modeling methods; identify risks and opportunities. It not only suggest ‘what will happen’ but also help understanding ‘why it will happen’. It’s like analyzing historical data combined with customer sentiments; define correlations and patterns between data sets, to determine likelihood of probable future outcome.

A well-known example of predictive analytics is log analysis. Predict Hot spots, memory bottlenecks, and threshold violations over live network feeds helps in decreasing system down-time, application failures. Implementing such analytics not only increases customer experience but also bring in cost reduction caused by abnormal system outages. Another very common industry application is predicting a Fraud by correlating a card and owner’s Geo location; Predict sales of certain product based on customer sentiments and usage pattern; Analyzing an individual’s social data and comparing it against data such as past individual claims, bank transactions and more can create a unified view of an individual that predicts and reduces fraudulent activity

This utilizes techniques such as Clustering, Logistic, Decision trees, neural networks etc. Output to predictive gives an intelligent insight to prospective changes that may be required in an organization business process.

Prescriptive - ‘How we can make it happen?’ Suggest Next Best action (NBA) and implication of each action (Pros and Cons). This is the final stage in an analytics maturity model. We can also call this an ‘optimization’ stage. We believe, it is still some time for organizations to reach this stage where they can utilize predictions to drive future market conditions.

This uses techniques such as Machine learning, simulation modeling and statistical modeling etc. Prescriptive creates an automated self-learning model and optimizes its output by applying the leanings to the predictive stage. It fine-tunes and predict the outcome of the various predictive models by checking the impact of market decisions on the model prepared. It helps mitigating the future risk. The learning of each outcome is then given as an input to the predictive stage. We should understand that there is a constant feedback and learning applied between predictive and prescriptive stages.

Google self-driving car is a very good example. Other examples are smart electricity grids with real time demand-based network optimization. This can also help plan pricing considering usage, weather, energy-loss and other economic trends. Considering market conditions and availability - prescribing a sell / hold-on advice, should the product is not sold for next 3 weeks. Data received from wearable devices such as Disney’s magic Band (tracking every move, preferences, choice of rides), credit card, social media, demographics, economic data can help plan new facilities or scale up of existing infrastructure etc. Individual social media data, combined with data showing the consumer’s movement over the web, can also be used to enhance marketing with personalized campaigns that can reduce spend on television and other media channels.

While, we are looking at various aspects of analytics, it is also important to understand, organization’s approach towards data analysis. This gives an idea on most of the different type of analytics projects carried across the organization.

Few industry use cases that have a larger focus and ROI are

• BFSI - Detecting Fraud Patterns, Claims management
• Manufacturing - Inventory Management, Product analysis
• Healthcare - Smart wearable devices, Personalized Medicine
• Retail / CPG - Demand analysis, Brand Perception, Store location analytics
• Telecommunication - Customer Churn analysis, Network Performance and optimization

The fact that such large volumes of data can be processed real-time, has created powerful opportunities by enabling real-time control of market shifts and stay ahead of competitors.