Rd big data & analytics v1.0
Download as PPTX, PDF1 like939 views
The document discusses analytics with big data, describing how businesses are using analytics to gain insights from large datasets. It provides examples of common business questions and the types of analytics that can help answer them, such as forecasting, recommendations, and predictive modeling. The document also introduces Robust Designs, a software company that specializes in business intelligence solutions using their CUBOT product.
![Recommendations:
Item Based Collaborative Filtering: based on Product Similarity [Cosine Rule Method]](https://image.slidesharecdn.com/rdbigdataanalyticsv1-150322112632-conversion-gate01/85/Rd-big-data-analytics-v1-0-19-320.jpg)
Rd big data & analytics v1.0
- 1. Analytics with Big Data Introduction: Robust Designs, Jan 2015
- 2. Contents Analytics Market: Questions our customers have asked us Market Evolution: Business Intelligence, Analytics, Big Data Our Credentials: Analytics Samples from our Work Enterprise Analytics Applications About Us: Robust Designs
- 3. Analytics Market: Questions From our Current Customers Analytics Market: Questions our customers have asked us
- 4. Telecom: Consumer Marketing ā¢ Tell me about my customersā behaviours ā¢ Who is moving out ā¢ Who is moving in ā¢ Who is spending on what ā¢ What can I cross sell ā¢ Tell me what can I offer them ā¢ Based on what others who are similar have bought ā¢ Micro segment and create personalized offer for a marketing campaign ā¢ Track the campaign and learn and refine
- 5. ā¢ How do I find out who are the distributors who are loyal to me ā¢ Design a loyalty management programme which is more likely to be successful ā¢ Which collection cases should I target today ā¢ What is the strategy for each case: ā¢ Should I write a letter, call ā¢ What figure I should settle for Financial Services: Sales, Loyalty, Collections
- 6. Healthcare Medical ā¢ Show me how dengue fever is spreading in a city ā¢ And predict how it is likely to spread ā¢ What are the chances of readmission given history of a patient and treatment received Financial ā¢ Predict billing trends by DRG ā¢ Predict likely defaulters
- 7. Retail / Manufacturing ā¢ What stocks make best basket of things to carry in my showroom ā¢ Which stocks I should carry how much keeping in mind the order trends
- 9. Data Evolution You asked questions, IT gave you answers You discovered answers by yourself Big data gives you answers ā any questions? Core system reports Business Intelligence Distributed Computing + Statistics Enablers Data Evolution Large scale computing power is available now for big data to have a justifiable ROI
- 10. Whatās Changing Business Intelligence ā¢ Analysis is top down, done by humans ā¢ Identifying root causes for behavior and variance to set goals is a main target ā¢ Action addresses the root causes identified Big Data / Analytics ā¢ Analysis is bottom up, done by machines ā¢ Identifying root causes is considered not necessary ā¢ Deriving consumer behavior, patterns is paramount ā¢ Action is not based on cause ā effect ā¢ Action is based on patterns and on the entire population - using Micro- segmentation, Personalization
- 11. Descriptive to Prescriptive Analytics Customer Delight ----- Prescriptive ā¢ Track, Segment, Recommend Action Better Business ---- Predictive, Diagnostic ā¢ Business KPIs, Insights, Analytics Efficient Operations ---- Descriptive ā¢ Dashboards, Reports, Data ā¢Identify clusters based on some variables Correlation ā¢Predict behaviour of this set at a future point in time Prediction ā¢Suggest best actions to meet a desirable outcome Prescription Analytics Model
- 12. Big Data Is Big Data About Largeness of Data? ā¢ Not necessarily ā¢ And it is not just about Twitter, Facebook posts ā¢ But big data scales linearly, and can work with large data sets not handled before by conventional technology like databases and data warehouses ā¢ Big data is deployed where there is big money at stake, regardless of the size of the data ā¢ E.g., an asset management company managing assets of thousands of high net worth individuals can take advantage of big data as much as a telecom company with millions of consumers ā¢ Big data applications are relevant for small businesses too, but ā¢ the cost of big data has not yet come down to that price point, and ā¢ the expertise to apply the technology and into a domain is scarce What is Big Data? ā¢ It is about the ability to ā¢ take in all the data in your context, ā¢ being able to process it fast enough for your business, ā¢ apply statistics, and ā¢ generate so many graphs that no human can read them all in reasonable time, let alone analyze ā¢ But big data machines analyze them, come up with correlations, predict behaviors ā¢ To the point of knowing if this customer is going to churn, if this product offer is going to be successful with what degree of certainty, and if the lady walking around your store is pregnant
- 14. Experience - Analytics with R Understand Data Enhanced Data Visualizations Quantile Based Estimates, Distribution & Probability Plots Derived Variables Custom Aesthetics Group āthingsā Clustering Euclidean Distance Based Dendrogram Comparing Cluster Features Find Similarities Association Analysis Interactive Association Rule Mining Mining for RHS & LHS Relationships (After vs Before) Plotting Associations Recommend Next Best Recommendation Engine Item Based Collaboration Filtering (Based on Product Similarity) - Generating Top x recommendations based on a Consumption Frequency Count Table Forecast Time Series Analysis Forecasting via Holt-Winters, adjusting for Seasonal Trends, Arima Social Data Mining Text Analytics & Connecting to Social Data Plotting Text Data in Wordclouds using various source formats (plain text, pdf, XML) Getting Associations between Text Data Connecting & searching Twitter Public Data
- 17. Clustering: Comparing Identified Clusters by Mean/Median Time Series Data
- 18. Market Basket Analysis: Hospital Bills ā What items are billed together?
- 19. Recommendations: Item Based Collaborative Filtering: based on Product Similarity [Cosine Rule Method]
- 21. Enterprise Analytics Applications Use Cases and Solution Methods Enterprise Analytics Applications
- 22. Analytics Uses Across Functions Strategy & Planning Marketing/Sales Customer Service Finance Operations
- 23. Customer Service Business Question Analytics Suggestions Are my customers who get in touch with the company happy? What kind of feedback is received, how has this changed over time, and what is likely to be expected in the future? Sentiment analysis: Text Analytics & Visualization Logistic regression forecasting, Hypothesis testing, Conditional probability employing methods What changes to customer service have impacted customer satisfaction the most? Correlation Analysis: Predictor vs response variable testing Which customers are most valuable ā and are these customers being served appropriately? Auto clustering Customer lifetime value through weighted scoring Where do new customers come from? Can we offer better services with less hassle for customers? Location analysis, clustering
- 24. Operations Business Question Analytics Suggestions What activities are most profitable? Pattern Identification through Data Mining Are resources being maximized? How can I get more out of my current resources (e.g. reduce system downtimes, improve service efficiencies)? Pattern Identification through Data Mining Can I make an improvement to logistics/ supply chain processes? Pattern Identification through Data Mining
- 25. Finance Business Question Analytics Suggestions What are the revenue and cost projections for next year? Forecasting Methods: Seasonal Adjustment Forecasting(Holt-Winters), ARIMA, Linear Regression Which (groups of) customers are likely to default payment? Logistic regression & response modelling What is the revenue breakdown? Quantile based estimates, box & whisker plots How can I cut finance approval times ? What items (are likely to) take the longest for approval? Pattern Identification and Data Mining Conditional probabilistic estimates Requests approvals vs rejected statistics, breakdown Better Visualizations
- 26. Marketing/Sales Business Question Analytics Suggestions What are my sales forecasts for 2015? Forecasting Methods: Seasonal Adjustment Forecasting(Holt-Winters), ARIMA, Linear Regression What are our customer segments? Auto- Clustering Derived variables/segments through weighted scoring What is the distribution of my annual sales (or numerical data) by price of item? What prices command the greatest proportion of sales? Quantile based diagrams Box & Whisker Plots What are the unusual trends in my data (customers, employees, sales, datesā¦)? Where are the exceptions and how can we explain these? Outlier analysis : Quantile based (wrt Interquartile range), Confidence Interval based (beyond 95% CI), Distribution (2 SDs) What are the influencers for sales? Correlation Analysis: Predictor vs Response Variable Testing Will this type of product pricing work for this category of brand? Forecasting Methods (see first row), Hypothesis Testing, Logistic Regression What are customers saying about my company /brands on social media? Social media connection (Twitter, Facebook), Sentiment/Text Analytics, Hashtag/keyword searches, Likes/Comments/Shares statistics
- 27. Strategy/Planning Business Question Analytics Suggestions Where should I invest more capital next year? ROI estimation & complete summary statistics based on departmental statistics Weighted scoring & sales forecast estimates What kind of people should I be hiring? Scoring of employees, department performance statistics & activity profitability estimates What can improve the companyās brand? Social media, CRM, Internal e-mail & chat text analytics Where are the gaps in operational efficiency? Downtime statistics, resource utilization %, time- related patterns How can I maximize company profitability? Pattern Identification through Data Mining
- 28. About Robust Designs About Us: Robust Designs
- 29. Robust Designs and CUBOT is a software company specializing in BI solutions ā¢ Operational since 2004 ā¢ Privately held ā¢ Offices: Singapore, Mumbai, Bangalore ā¢ 15 people is our BI product with over 40 customers in India, Singapore, Malaysia, Singapore, Vietnam & Netherlands ā¢ Developed with the vision: ā¢ Faster to Implement, Simpler to use
- 30. Experience with BI over the Years India, Singapore, Malaysia, Vietnam, Netherlands PASTCURRENT Stayed 4-6 years Stayed 2-4 years Stayed 1-2 years PAST CUSTOMERS APACINDIA PASTCURRENT
- 31. Last Slide