Big data Analytics in Retail

All the industry leaders like  Wal-Mart, Axa, Citibank, Humana, GE and several others are exploring how Big Data analytics can be used to better understand customer needs, pinpoint risk, improve marketing, enhance the customer experience, combat fraud, and drive profitability.  Companies are seeking ways to rebuild their customer relationships in this time of extremely high customer expectations.

Retail industry is among the early adopters and innovative users of big data. But they have the challenge of tackling the huge data since 1970s when barcodes were first introduced to scan the products at POS.  All sorts of supply chain data came into effect later in 1980-90s while RFID and other sources such as surveillance video cameras started sending humongous data to data centers recently.  These have challenged Retailers to capture, store, cleanse & analyze all the data they collect.

Further to flood the data centers are consumer’s interaction with social media & internet which is generating billions of data points that can be measured via clicks, page views, time spent on per page and path traversed from landing to conversion.

Big data analytics is helping retailers to collect and analyze this fine grained shopper visit data and optimize page designs, placements and tailor promotional messages.
McKinsey report say that using big data analytics can raise the operating margins by as much as 60%

Some of the questions Retailers have are:
·        How to drive critical decision around market segmentation, personalization & merchandizing?
·        How to avoid lost revenues due to stock outs, lower online sales per visit, lower visit to buy ratios?

Here is a glimpse of what retailers can do in big data analytics:

Customer:

·        Enhancing customer experience across all the channels such as calls, emails, campaigns, catalogs, mobile offers, brick & mortar stores
·        Customer sentiment analysis to know the market pulse and market dynamics
·        Call center data analysis for customer feedback
·        Build loyalty programs based on purchase data & customer segmentation
·        Staffing optimization based on weather forecasts & promotional campaigns for better customer experience

Merchandizing:

·        Optimizing the product placements and layouts based on video data
·        Price optimization based on competitor pricing
·        Market basket analysis for revenue growth
·        Optimizing seasonal markdowns
·        Store analysis for best location & better effectiveness
·        Improve in-store sales by leveraging past data with current economic, weather & season/holiday data

Marketing:

·        Consumer segmentation, cross selling
·        Campaign analytics to channelize advertising dollars in optimal medium for highest ROI
·        Sentiment analysis from social media, call centers, surveys, blogs, product reviews
·        Identify new products, service & market opportunities by real time monitoring of these customer sentiments
·        Location based personalized offers on smart phones, tablets
·        Web log analytics for customer behavior analysis & next best offer

Supply Chain:

·        Inventory optimization to avoid stock outs
·        Demand driven forecasting fueled by structured and unstructured data
·        Route optimization for cost reductions
·        Warehouse space optimization
·        Vendor performance analysis for better competitive prices

Ultimately, the goal of big data analytics is to develop an effective Omni-channel experience that integrates many different factors of supply chain including supplier effectiveness, warehouse optimization, and inventory / logistics optimization for real-time customer engagement.

Big data analytics provides the required ammunition & tools to accelerate growth, boost profits, control risks and meet regulatory & competitive demands.

 

Enhancing Customer experience with Big Data Analytics

The era of Big Data is upon us.

Big data by definition is:

·         combination of all the internal structured business data (CRM, ERP, POS and all the internal system data) and external unstructured data ( Social media data, feedback surveys, Audios, Videos, streaming data, Call center data, images)

·         unmanageable volumes by traditional databases

·         enormous speed at which it comes into the organization.

Today, most organizations are still struggling to unlock the full value of this Big data that is available to them. From internet to mobile and social, the amount of customer data is continuously growing.

Company’s ability to extract value from big data through smart analytics will be the key to their business success.

Channel based marketing is a least priority now. The increased amount of data available at individual customer level has allowed companies to do a personal marketing.  But all this customer data out there is worthless if you can’t process it & turn it into actionable intelligence.

With Big data platforms helping in  collection, integration, and transformation of large volumes of data, companies can conduct complex and varied analysis on much larger datasets and reduce the time to action and reaction to customer needs.

Organizations can now impact the entire customer life cycle and every interaction by being well prepared for each interaction, shaping the interaction in real time as it happens and driving the huge improvements across all the channels for next interaction.

By listening to the data as a signal from customers and working to personalize the experience for the customer, creates the value for the customer as well as business.

Some examples of enhancing the customer experience using Big data Analytics:

·         Retail giants are using Big data to personalize the offers to enhance customer experience

·         Healthcare companies are using it for improving the patient care in hospitals

·         Banks are using it for upselling and bringing the new products to market

·         Insurance companies are making tailor made policies for their customers in real time

·         Manufacturing companies are using Big data to improve their products, predict the failures in their product lines ahead of time to make every customer interaction very smooth

Using Big data to address customer inputs before they become problems is extremely important to ensure they stay loyal and more profitable to the company.

Customers expect to have the best possible experience from their vendors/service providers. They want to be recognized as individuals and not as a part of a segment.

Big data Analytics – A disruptive technology !!

Big data is the most talked term these days in the analytics world. It will have big transformative impact on all the aspects of the business.

Most of the companies now have realized that there is a huge competitive advantage in analyzing the humongous data quickly & effectively for future insights.

Big data analytics is the disruptive technology bringing the 4^th aspect of Value to the already published TDWI’s 3Vs – Volume, Velocity & Variety.

• It enables business users to process every granular bit of data in quicker way removing the traditional need for sampling & then applying the models
• It encourages an investigative approach in users for data analysis since they get access to whole data
• It can reveal insights hidden in the data, which were previously too costly due to large data movements
• As per Gartner report, Big data is priority of SMB & it will drive $232 billion in spending through 2016.

Some of the technology platforms which are used for big data:

• Distributed file based: Hadoop-MapReduce (Cloudera, Hortonworks, MapR)
• Appliance based: Greenplum, IBM Puredata(Netezza), Oracle, Teradata
• Columnar databases: HP Vertica,  ParAccel, 1010data
• In-memory databases/tools: SAP Hana, Qlikview, Tableau
• Non relational/NoSQL: Cassandra, MongoDB, Splunk, Hbase

Hadoop is on the top of the list of technologies used for dealing with Big data due to its ultra high scalability & low cost compared to other platforms.  It is a suite of products linked together, which breaks up the large datasets into smaller chunks on commodity servers, and data processing is done in a distributed cluster environment to quickly return the results.

Some of the probable Big data use case in various industries:

Insurance: Collecting data from monitoring devices fixed in cars & providing the personalized insurance policies based on driving habits, Underwriting price optimization for insurance products, Claims fraud with social network link analysis.

Retail: Market basket analysis for entire merchandising instead of sample data, Sentiment analysis based on social media for improving brand perception, customer service, competition analysis, Customer & market segmentation, Weblog analysis for customer behavior.

Banking & Finance: Fraud detection with entire history data for better detection, Trade surveillance in capital markets, More accurate risk score to customers, Text mining on call center data.

Healthcare: Improve patient care with analyzing electronic Health Records (EHR) & reduce insurance payer costs, Reduce hospital readmission rates by analyzing information from discharged cards.

Manufacturing: Forecasting warranty costs & detecting issues in spare parts of finished goods, text mining to understand the complaints from customers for product improvements.

Because of nascent stage immaturity of Big data initiatives, there are many views of what is it & how it can be applied.  Organizations need to focus on Big data processing, while avoiding the movement of large volumes of data which is very costly.

Big data help make better decisions – faster, more efficiently with higher quality.

so what is prescriptive analytics?

Today every business is surfing in the ever expanding sea of data & using analytics for getting the edge over their competition.

With the explosion of unstructured data on social media, audio-video steams companies are rushing to use this for big insights.

There are mainly 3 types of analytics & it is based on the company’s maturity on analytics as to which one to adopt….Descriptive, Predictive & Prescriptive analytics.

Let me explain with examples.

	Descriptive Analytics	Predictive Analytics	Prescriptive Analytics
What questions are answered?	What happened? How many customers? Where revenue is less? Why it is so?	What will happen next? What trends will continue? What if we change pricing?	What is the best course of action for given situation? What is the impact of seasonality?
How it is done?	Use of KPIs, dashboards, charts	Use of statistical methods to understand the relationships in input data & predict the outcomes. Use of data mining, forecasting, predictive modeling.	Use of advanced statistical optimization & simulation techniques with inputs & constraints to recommend what actions to be taken.
General examples	How many customers have churned? Why did they churn?	How many customers will churn in next few months?	What actions to be taken to retain these predicted churners?
Some of the Industry examples	Netflix uses data mining to find out correlations between different movies that subscribers rent & then recommend the one which you are most likely to watch	ING using personalized campaign offers in real time by predicting who will respond, to increase 30-40% response rates & reduce direct marketing costs by 35% per year.	Amazon.com using price optimization based on demand to increase the online shopping revenues.

  How various industries are using prescriptive analytics?

·         Consumer product companies are using it to maximize the marketing dollar spend

·         Transportation & Logistics companies are using it to find the best route for their deliveries & backhaul

·         Healthcare service providers are using it to decide how many beds they should increase in the hospitals

·         Manufacturing giants are using it for inventory optimization to decide how much safety stock they should keep of each item, where to stock it based on the demand

·         Telecom business is using it for providing on the spot offers to customers when they call the customer service centers

One daily life example - Imagine you are driving a car with a built in GPS, which analyses all the data it collects from the satellite about traffic, accidents, weather etc. It tells you which routes will have heavy traffic (prediction), but also recommends you the alternate routes (prescription) with less traffic.

So Prescriptive analytics is where you know what the future is, but also know what to do with it, with alternatives of best outcomes J

Fraud detection in Medicaid / Medicare

Medicaid is a USA government run healthcare program for the poor, elderly and disabled, which is jointly funded by the states and the federal government

Medicaid has been the top priority on many state budgets. Considering the huge investments from both government as well as states, there are whole lot of malpractices to grab this money.

Experts estimate that abuses of Medicaid eat up at least 10 percent of the program’s total cost all over USA which is a waste of more than a $80 billion a year.

Let us see what the types of Medicaid Fraud are:

·   Doctors billing for over 24 hours per day of procedures

·   Use of single patient ID  to generate billing across  multiple providers

·   Fake companies invoicing for services which were not provided

·   Altering the claims forms or misusing the codes to receive higher payment amounts

·   Pharmacists filling prescriptions for dead patients

·   Home health-care companies demanding payment for treating clients actually in the hospital

·   Home health care, visiting nurses billing additional amount

·   Patient transportation services claiming charges for patients who are not even moved to and from hospitals/home

One industry example mentioned in the reports: In one brash scheme, immigrants set up a network of fraudulent medical-supply stores in the Southwest, hoping to cheat Medicaid and Medicare. The gang hired recruiters to bring them innocent patients eligible for Medicaid or Medicare. They then paid off local doctors to prescribe motorized wheelchairs worth $7,500 but instead gave them motor scooters worth just $1,500, pocketing the difference. Investigators shut down the scheme after noticing billings for wheelchairs in Arizona, Texas, and other states scaling into the hundreds of millions of dollars.

How analytics can help in Fraud detection/prevention:

·   Detecting the patterns of fraud in the bills provided by doctors, hospitals, nurses

·   Profile & segment claimants to identify those who are likely to commit fraud

·   Detecting overpayments due to provider’s incorrect billing

·   Use of Big data platforms to analyze huge volumes of data for fraud detection

·   Identify connections of fraudsters via social network analysis

·   Apply analytics with combination of methods of anomaly detection, business rules, predictive modeling & network analysis

·   Advanced text analytics to analyze unstructured data to reveal fraudulent activities

There are some steps, which individuals can take to prevent fraud:

·   Review your Medicaid bill for each service. Are the dates correct?

·   Only give your Medicaid number to those needing it. (Doctors, hospitals, clinics, etc.)

·   Don't lend your Medicaid card to anyone.

·   Never request medical services or equipment you don't need.

·   Don't sign blank forms for medical services or equipment.

·   Request and retain copies of anything you sign.

Warranty Analytics - Increase Product quality, Customer satisfaction & Brand perception

As a customer, when you buy any home appliances like TV, AC, Refrigerator, Home Theatre or a brand new car, you get a company warranty along with it. This is the commitment from the manufacturer that if any problem arises in the product or spare parts within the warranty period, then company will repair or replace it free of cost. 

Industry numbers shows that warranty costs range from 2% to 6% of the company’s revenues. Predicting these warranty costs is an important step for successfully managing the business.  If manufacturers reserve too much money, then they lose opportunities to grow the business because they end up with less cash.  If they set aside too little money, then they lose opportunities because they have to keep adding to the warranty reserves funds. 

Let us see some quick definitions of warranty:

• Base Warranty – original warranty coverage provided by manufacturer at no extra cost, since it is included in the product price.
• Extended warranty – this comes into effect after the base warranty expires.
• Warranty reserves – amount of money set aside by manufacturer for the purpose of servicing the warranty claims. This is based on the forecasted warranty costs.

In automotive industry, warranty generally guaranties free repairs or replacements subject to both age of the car & mileage.

Warranty Analytics is integration of warranty claims data with customer, product, sales and geographic information, so companies can accelerate detection of failures and reduce time to correction.  

It can help in significantly improve the early warnings of parts failures based on customer complaints and failure patterns, combining structured data with un-structured data (such as call center records) to give alerts and information about developing trends that would have gone unnoticed earlier.

By identifying warranty-related issues early, companies can save thousands of dollars in both repair costs and customer retention because issues are proactively addressed before they become significant, costly problems.

Root cause identification of parts failures is the biggest challenge in the industry today.  70% of annual warranty expenses are consumed by repetitive and chronic problems.  Prioritization of these root causes helps companies calculate how much it will cost if nothing is done. This allows them to determine the best course of action and associated costs, as well as any potential effect on customer satisfaction.

Managing warranty costs is an enterprise wide challenge, impacting multiple departments, including quality, product engineering, customer service, finance & purchase.

Typical areas of applying Analytics on Warranty data involves:

• Data mining to Identify the patterns of claims
• Text mining to identify problem areas and fixing them,  instead of technicians trying to select from hundreds of warranty categories
• Predicting the expected number of claims or cost of claims
• Predicting fraudulent claims
• Investigating the association between different types of claims
• Identifying issues before they become showstoppers
• What-if analysis such as if we increase the mileage what will be impact on warranty costs

Some of the warranty analytics benefits:

• Increased customer satisfaction, product quality & brand reputation
• Tremendous impact on bottom line due to early issues identification
• Huge reductions in total manual claims processing costs
• Prevention of fraud on warranty claims
• Optimized warranty policies for maximum financial performance
• Increase efficiency of support logistics such as optimum stocking of replacement parts or deployment of technicians

It helps answers the questions like:

• Our competitors just raised their product warranty from 3 years to 6. If we do adopt the same, how much more warranty costs we will incur? If we don't, how much revenue we will we lose from reduced market share?
• Given a new product with no historical data, should we play it safe and offer only a one year warranty, or can we offer a three year warranty to improve our brand perception?