Driving Operational Results with Big Data Jen Sindelar, Ph.D. Engineer, FedEx Ground Gary Burns, Ph.D. Senior Manager, FedEx Ground
What is Big Data? Big data is a popular term used to describe the exponential growth and availability of data, both structured and unstructured SAS Big data is the electricity of the 21st century companies use data to model and control processes and run the business Oracle Billions of connected devices Internet users and machine-tomachine connections are contributing to a tsunami of data growth. Enterprises that harness this data can use it to differentiate themselves and outperform competitors Intel Slide: 2
Analytics professionals turn Big Data into Useful Information Data Mining Data Capture & Storage Decision Making Models Useful Information Analytics professionals include: Industrial engineers Mathematical modelers Computer scientists Business analysts Many others They enter the workforce with a technical toolbox Programming skills Background in statistical software Simulation modeling capabilities Experience with operations research models and tools Slide: 3
Analytics professionals face challenges beyond the technical aspects of analysis This talk is not about data analytics techniques. Rather, it focuses on the challenges faced when working with large datasets in an operational setting Working with less-than-perfect data Integrating large and disparate datasets Summarizing and communicating information Incorporating big data into operational decision-making processes We will discuss best practices for driving operational results Examples from multiple settings are used to illustrate tools and techniques for overcoming big data challenges faced by analytical practitioners at all levels Slide: 4
Topics for the Talk FedEx Background What does Big Data mean to FedEx Ground? Discuss challenges & successes when using Big Data to drive operational results Case Study 1: Designing and implementing a new performance metric Case Study 2: Creating a performance dashboard Case Study 3: Incorporating third-party software & data Takeaways for analysts at all levels Summarizing and communicating information Integrating large and disparate datasets Working with lessthan-perfect data Incorporating big data into operational decision-making processes Slide: 5
FedEx Corporation consists of four operating companies with multiple reporting segments The largest express transportation company in the world. Cost-effective, small-package shipping. Priority and economical less-than-truckload freight. Technology and services customized for your business needs. Slide: 6
FedEx Ground is the second largest operating company in the FedEx portfolio Cost-effective, small-package shipping. Specializes in day-definite small package shipping FedEx Home Delivery is the only service of its kind dedicated to residential customers Fastest and most automated hub network in the U.S. ground pkg shipping industry World s most advanced material handling and scanning technology Slide: 7
4 3 FedEx Ground Operations 2 1 5 7 8 Day 0 Pickup (Murfreesboro, TN) 4:01 PM 6 9 Station Sortation 7:06 PM Day 1 Day 2 Hub Area Linehaul 9:18 PM Mainline Linehaul 8:21 AM Hub Area Linehaul 1:41 AM Hub Sortation 11:58 PM Hub Sortation 10:03 PM Station Sortation 7:28 AM Delivery (Columbia, SC) 12:29 PM Slide: 8
Big Data Analytics opportunities span all facets of operations Package routing Service improvement opportunities Packages 6+ million pkgs per day Equipment 5,000+ tractors 3,000+ load doors Equipment utilization Fleet optimization People Facilities 61,000+ team members 33 hubs 500+ stations Performance metrics Decision rules Facility location Facility design Slide: 9
Three case studies focus on using large datasets to drive performance and operational improvements Case Studies 1.Designing and implementing a new performance metric 2.Creating a performance dashboard 3.Incorporating third-party software & data These case studies leverage operational package data An average package receives 15 scans while it is in the network This translates to billions of scans per year Slide: 10
Case Study #1 Designing and implementing a new performance metric Slide: 11
Case Study #1 Designing and implementing a new performance metric Opportunity Barcode & scanning technology provides the location of every package while in the network for tracking and tracing Trailer dispatches are monitored through a transportation management system to provide on-time service Can we integrate data from these systems to measure how effectively we are using space within trailers? Slide: 12
Case Study #1 Designing and implementing a new performance metric Systems Challenges Communication between systems Legacy systems issues Data cannot be accessed real-time Data is only available locally IT resources not available for enhancements Funding not available for improvements People Challenges Resistance to change/communication Systems issues make it difficult to incorporate good suggestions Truthing the data Summary reports vs. detailed drilldowns Data access Data vs. Information Slide: 13
Case Study #1 Designing and implementing a new performance metric Solution Part 1: Make data accessible Slide: 14
Case Study #1 Designing and implementing a new performance metric Solution Part 2: Take a phase approach Phase 1: Alpha Test Limited set of trusted partners Provide time to dig into the data Be responsive to suggestions Phase 2: Beta Test Give everyone a chance to work with the data before they are graded Incorporate feedback from Alpha Test Phase 3: Initial Rollout Communicate, communicate, communicate Use the results to show improvement, not just opportunities Incorporate feedback from Beta Test Phase 4: Continuous Improvement Have a mechanism in place to address concerns Show continuous wins made possible by the data Phase 5: Long-term Solution Get people excited about what is coming next Get them involved in working toward the next step Show positive movement toward the long-term solution Provide a realistic timeline for roll-out Real-Time Next Day Next Week Slide: 15
Case Study #1 Designing and implementing a new performance metric Results Metric has been incorporated into operations planning at the daily, weekly, monthly, and annual levels Hubs and automated stations utilize drilldown reports to work directly with operations to improve load quality Showed significant system-wide improvement in the first six months Metric is utilized in goals-setting and budgeting processes Ongoing research into new technologies for measuring trailer capacity With direct field involvement and participation Slide: 16
Case Study #2 Creating a performance dashboard Slide: 17
FedEx Ground Linehaul Operations Day 0 Pickup 4:01 PM Station Sortation 7:06 PM Day 1 Day 2 Hub Area Linehaul 9:18 PM Mainline Linehaul 8:21 AM Hub Area Linehaul 1:41 AM Hub Sortation 11:58 PM Hub Sortation 10:03 PM Station Sortation 7:28 AM Delivery 12:29 PM Slide: 18
Case Study #2 Creating a performance dashboard Situation Local Metrics vs. System Metrics Many existing rules of thumb and reports to identify problems Balance Report Crossed Empties Report Late Arrivals Report But how is the system doing as a whole? Detail available, but not summarized Identifying opportunity areas Allocating scarce resources Slide: 19
Case Study #2 Creating a performance dashboard Opportunity High-level view of network performance Goal: Create a system-level performance dashboard Summary reports Drilldown to Region, District, Hub levels Detailed reports for specific metrics Slide: 20
Case Study #2 Creating a performance dashboard System Challenges Performance/speed Achieving real-time data availability Disseminating data Drilldown capabilities People Challenges What are the right metrics? Ownership of the metric and driving improvements Relevance of the dashboard and individual metrics (vs. numerous other reports) Communicating/explaining metrics Slide: 21
Case Study #2 Creating a performance dashboard Solution Keep metrics simple Start at the top and work down Provide details (separately) whenever they are available Provide tools for front-line analysis Give examples of how to use metrics to drive improvement Provide a goal Slide: 22
Case Study #2 Creating a performance dashboard Results Dashboard is utilized for both corporate oversight as well as local management System view used to identify key drivers of linehaul cost Region, District, and Hub views allow detailed investigation into opportunity areas Local facilities are now asking for details to help improve operations Additional benefits achieved due to improved communication between linehaul and hub operations Better load planning Provides communication channel for new metrics Slide: 23
Case Study #3 Incorporating third-party software & data Slide: 24
Case Study #3 Incorporating third-party software & data Situation FedEx IS address data Pickups & deliveries Stations & hubs Problem Addresses are messy Knowing the correct address means On-time service Reduced operational cost Improved customer satisfaction Slide: 25
Case Study #3 Incorporating third-party software & data Situation Goal: Consistently cleanse address data to the correct address, every time Associate information with this address Geospatial information (e.g. latitude/longitude) Descriptive information (e.g. business or residence) Customer preferences (e.g. operating hours) Strategy: Use third party software & data to assist with address cleansing Slide: 26
Case Study #3 Incorporating third-party software & data Systems Challenges Licensing Access to underlying data Working with vendors to provide new features Upgrades/compatibility issues Reliance/dependency on vendors People Challenges Building trust in external systems Core expertise lies outside the company Building knowledge base Gaining buy-in for one system vs. another Managing scope Slide: 27
Case Study #3 Incorporating third-party software & data Solution Form partnerships with a small number of vendors Bring vendors into the R&D process Build joint roadmaps for new functionality Own the interface and keep vendor information behind-the scenes Don t rely solely on vendor documentation build your own knowledge base of their products Conduct extensive testing, even for minor changes & upgrades Slide: 28
Case Study #3 Incorporating third-party software & data Results The FedEx address cleansing system provides multiple services Address validation Business/residential classification Geocoding Storage of address-related information New functionality added as needed by working closely with vendors International expansion of address validation & geocoding services Address lookup services Address quality enhancements Slide: 29
Takeaways Driving Operational Results with Big Data Phased Approach Provide Tools Own the System Access to Data Give Examples Knowledge Base Communicate Slide: 30