Big Data and Predictive Analytics Cameron Hall Vice President, Products ValueCentric, LLC
Agenda 1 What is Big Data? 2 Does your organization have Big Data? - Spoiler Alert: Yes! 3 What is Predictive Analytics? 4 Practical Applications of Big Data & Predictive Analytics
ValueCentric collects and manages Healthcare data and delivers actionable results right to your desktop, mobile device or inbox. Our cloud-based analytics platform, ValueTrak, transforms Big Data into meaningful intelligence.
852 EDI Files 867 EDI Files Chargeback data Orders Rx Data Coupon data Payer data Claims data Contract data GPOs Specialty data PHI Pedigree data 340B 3 rd Party data IMS data Partner data Manufacturer Trade Sales Ops Finance Marketing Brand Sales IT Managed Markets Commercial Operations Contracting Manufacturer Data Cloud-based data management and analytics platform Extracts, Data feeds, APIs, etc. Secondary Systems Customers, Partners, 3 rd Parties, etc.
What is Big Data?
What is Big Data? Academic Definition: Data capturing, processing, and storage that has outgrown a traditional relational SQL database; exponential growth of data For example: Facebook captures 500 terabytes of data each day, feeding complex advertising and sharing algorithms
What is Big Data? Purist Definition: Datasets that capture every aspect of a subject area a complete universe of data. For example: If a car was equipped with a comprehensive network of sensors, it would be possible to compare variables leading up to an accident against those measured during normal driving.
What is Big Data? Practical/Marketing Definition: Novel applications of large data sets; the creative application of our accumulated subject matter knowledge. For example: Clever applications of data we all have (or should have) access to today; Healthcare Supply Chain Data.
Five Qualities that Define Big Data? Volume Variety Vast amounts of data Value Different types and formats Velocity Speed of new data Putting it to work! Veracity Messiness/Quality
Big Data: Volume Business Application: Data size (breadth and depth), performance Example: Product Activity (852) Each 852 can contain over 1,000 SKU updates and up to 75 unique ZA code updates ValueCentric customers receive ~600M 852 product points-of-information daily Channel Data (852) By Manufacturer Volume 47% 10% 43% Files / day > 1000 100-1000 < 100
Big Data: Variety Business Application: Data management, completeness monitoring Example: Unique Channel Data Layouts ValueCentric customers have amassed over 2,500 unique data maps Approximately 50,000 mapped identifiers
Big Data: Veracity Business Application: Data quality, Curation of data Example: Variation in outlets For every outlet (pharmacy, hospital, clinic, etc.), there are an average of 9 unique, separate ways of reporting that outlet Master Data Management services are essential to analysis of disparate data
Big Data: Velocity Business Application: Growth estimates, future investment Example: Product Movement Activity On average, ValueCentric customer base receives approximately 18M downstream product handling updates every week (shipments, prescriptions, returns, etc.)
What is Predictive Analytics? Big Data is the novel application of large data sets; the creative application of our accumulated subject matter knowledge. Predictive analytics is the practice of extracting information from existing data sets in order to determine patterns and predict future outcomes and trends.
Five Qualities that Define Big Data? Volume Variety Vast amounts of data Value Different types and formats Velocity Speed of new data Predictive Analytics are what give value to Big Data Veracity Messiness/Quality
Levels of Predictive Analytics Prescriptive Analytics Let s make it happen Value Diagnostic Analytics Why it happened? Predictive Analytics What could happen? Descriptive Analytics What happened? Difficulty http://www.informationweek.com/big-data/big-data-analytics/big-data-analytics-descriptive-vs-predictive-vs-prescriptive/d/d-id/1113279 http://www.ibm.com/analytics/us/en/analytics-technology/index.html
Big Data: Organizational Emphasis Application of resources varies widely Developing Organization Mature Organization Most companies are still here
Big Data and Predictive Analytics, Practical Example: Service Level Monitoring
Service Level Monitoring Example Service Level Definition and Impact: Ability to fulfill customer demand for your product Percentage measurement of total sales / total demand Potential lost sales to a competitor Sign of frustration for your downstream partners and patients Product Activity Data (852) includes enough information to measure service levels, including sales, lost sales and adjustments to lost sales
Service Level Monitoring Example Stage 0 Manufacturer: No data management and reporting capabilities Unaware of any potential service level shortcomings Stage 0
Service Level Monitoring Example Stage 1: Descriptive Analytics We have a service level problem Weekly Service Level 100.0% 99.0% 98.0% 97.0% 96.0% 95.0% 94.0% 93.0% 92.0% 91.0% 90.0% 1 2 3 4 5 6 7 8 9 10 11 12 13 Weeks Stage 1
Service Level Monitoring Example Stage 2: Diagnostic Service level problem stems from wholesaler forecasting issue 180 160 140 120 100 80 60 40 20 0 Stage 2 Day Accumulated Forecast Cumulative Sales Lost Sales
Service Level Monitoring Example Stage 3: Predictive Leading indicator that service levels might suffer Stage 3 180 160 140 120 100 80 60 40 20 0 Leading indicator to Lost Sales: Small deviations in forecast vs. actual Day Accumulated Forecast Cumulative Sales Lost Sales
Service Level Monitoring Example Stage 4: Prescriptive Stage 4 Prevention of lost sales excellent service levels 180 160 140 120 100 80 60 40 20 0 Simple forecast adjustment leads to improved service levels Day Accumulated Forecast Cumulative Sales Lost Sales
Big Data and Predictive Analytics, Practical Example: Returns Forecasting
Returns Forecasting: Challenge Lack of visibility of pending returns causes: Accrual/Financial implications Depending on returns policies, could make or break quarterly/annual numbers Subsequent demand variability as returns inventory is replenished Resulting shortages & reduced service levels at: Manufacturer Wholesaler Pharmacy
Returns Forecasting: Challenge Products undergoing major lifecycle events are particularly sensitive Loss of Exclusivity Introduction of competitive products
Returns Forecasting: Assessing Available Data Leveraging unique data in creative ways: Historical Returns data Date of return Expiration date Quantity Rx and Sales data TRx Pharmacy Sales Pharmacy Survey Expiration date distribution
Returns Forecasting: Two approaches Statistical model Takes into account other variables and overall trend relative to the most recent months Strong for quickly adjusting to unexpected changes in trend Weak in identifying spikes or drops, and may overcompensate for spikes and drop in future weeks Weaker for long term forecasts 1200 1000 800 600 400 200 0 0 5 10 15 20 25 30 35 Weeks Increasing returns trend leading up to lot expiration, but otherwise random with spikes and valleys
13 19 28 46 Week 1 Week 3 Week 5 Week 7 Week 9 Week 11 Week 13 Week 15 Week 17 Week 19 Week 21 Week 23 Week 25 Week 27 Week 29 Cumulative Returns 144 283 355 426 471 527 538 573 588 606 627 627 630 639 650 Returns Forecasting: Two approaches Product exp date 5/12 700 600 500 400 300 200 100 0-6 -5-4 -3-2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 Return month relative to expiration date Expiration date model Estimates returns by expiration date, using recent returns and historical patterns to estimate future returns from those lots Strong in identifying spikes or drops due to expiration data distribution Strong for longer term forecasts Weaker for quickly adjusting to unexpected changes in trend
Returns Forecasting: Finding the Right Mixture Statistical model Expiration date model Combined model Accounts for strengths and weaknesses of each model to provide a better estimate 900 800 700 600 500 400 300 200 100 0-20 -15-10 -5 0 5 10 15 20 Weeks (past/future)
Big Data and Predictive Analytics, Practical Example: Specialty Products
Specialty Growth: Big Data Implications Specialty products are low volume, how does that fall under Big Data? Volume not just about depth; breadth of data can make it Big High patient touch and service leads to a wealth of new data assets Some ValueCentric customers receive SPP or HUB data with well over 100 reported data elements (patient demo, triage/status/pap, payer, copay, physician, etc ), updated in near-real time
Specialty Growth: Big Data Implications Rapid growth in Specialty products means more distribution/service partners in the market with varying degrees of data management maturity
Specialty Growth: Big Data Implications Predictive Analytics and Big Data require more effort than just integrating new data into your Big Data repository, it comes with a degree of data management Veracity The smaller your Big Data is, the more important that your systems be able to make sense of the anomalies Importance of a good partner that understands and practices reliable Data Curation techniques
Specialty Growth: Big Data Implications Data footprint will grow exponentially Eg. Additional patient metrics Reimbursement complexities Ensuring patients have access to the drug Physical (distribution channel) Financial (HUB, payer, PBM, Medicare, Medicaid, etc.)
Big Data and Predictive Analytics, Practical Example: Downstream Buying Patterns and Investment Buying
Downstream Investment Buying Wholesaler speculative buying has become a thing of the past in our industry: replaced by IMAs and fee for service agreements years ago But is the practice still prevalent, now further downstream?
Downstream Investment Buying Industry is still vulnerable to the practice due to the predictability of price changes nearly half of all price changes occur in January and July.
Downstream Investment Buying Working with a manufacturer partner, ValueCentric was able to identify arbitrage situations at over 30 downstream outlets
Downstream Investment Buying Sample outlet, purchasing >$20M of this manufacturer s product annually
Downstream Investment Buying ValueCentric identified over $100M in at-risk inventory across all outlets Post price-change revenue loss Risk of returns exposure Supply chain variability including reduced overall service levels Manufacturer then removed predictability from price change dates Immediately introduced stability in downstream buying patterns at outlets previously capitalizing on price changes
Conclusions Big Data opportunities exist in all organizations -- even yours! Predictive Analytics, done well, creates the value from Big Data Competitive advantages exist for those willing to make the investment in Predictive Analytics The ROI can be immediate and substantial!
Recommendations Ensure you factor in: Data Infrastructure: Invest in a platform that will scale to support the volume and velocity of your data Data Curation: Invest in managing the variety and veracity of your data Data Scientists: Invest in the human capital or partner with someone who provides that service to get value from the data No need to go it alone, there are vendors and partners who can help you get there
Questions? Cameron Hall cameron.hall@valuecentric.com