I know we can't share examples from every industry and every type of company in every industry. So, I wanted to share a few extra examples that I thought might be the most inspirational to you. They include examples from the transportation industry, agriculture, education, conservation, and professional services. At UPS they wanted to improve driver safety and efficiency. So, they put telematics sensors in 46,000 vehicles that capture the speed, direction, braking, drive, train, RPM, oil pressure, idling time, seat belt use, and 200 other data points including the geographic and map data. This new system they created called Orion helps them to determine the trucks performance and condition and recommend driving adjustments. The solution saves them over eight million gallons of gasoline per year and reduces maintenance and accidents by cutting 85 million miles off daily routes and a 25 percent reduction and reversing trucks for a $50 million total annual savings. It also enables them to deliver 35 percent more packages and they've doubled driver wages as a result. One of the more interesting analytic solutions involves a bit of extortion. A system called Indooratlas, maps the interior of a building using the unique magnetic fingerprint of the structure caused by distortions of the earth's magnetic field. They're able to determine the location of internal structures and people to within six feet including which floor they're on. The company then sells this data to companies so they can see changes to competitors' layouts and foot traffic. If you don't want your map published you can pay them $99 a month to keep your buildings magnetic map private. This is a great example of public-private partnership. Conservationists wanted to understand and improve animal populations. So, the Smithsonian Conservation Institute partnered with United Airlines, their solution tags smaller animals with VHF radio transmitters that are too weak for satellites to pick up. So, network of receivers is placed on United Airlines 5,300 daily flights. The lower cost of these tags receivers means a greater number and variety of smaller animals can be tracked continuously. It also reduces the expensive dedicated flights by private planes or catch and release techniques. Ultimately, the system is able to determine when and where animals are migrating and dying. In the agriculture space, farming has become quite automated, improving farming productivity, growth, and margins. At a farm called Tom Farms in Indiana, soil is tested with electrical charges and mapped for precise fertilizer dosages, and then they're applied automatically. They also use drones equipped with infrared cameras to survey for flood, irrigation and crop stress. Combines take a variety of continuous readings, analyzing data in real-time, on moisture, yields, etc. This has given Tom Farms the ability to farm 20,000 acres up from 700 acres in the 1970s with only 25 employees. The farm has experienced a return on investment growth from 14 to 21 percent now and has eliminated the need for crop diversification, to hedge against weather disease and market conditions. One of the challenges with beer kegs is that they're opaque. You can't see in them to find out how much beer is left. This creates a challenge for restaurants and pubs that either waste beer or fail to replenish kegs in time. A company called Steady Serv created a system called iKeg that tracks the contents automatically via sensors. This reduces the amount of wasted beer from 15 percent to virtually zero, and it produces real-time insights to all supply chain stakeholders like the brewers, the distributors, retailers and brand managers. It helps to improve customer satisfaction and loyalty, and it helps the retailers run effective ad campaigns, measure conversion rates, and enhance brand engagement. A company called Numerai uses crowd-sourcing and combine algorithms for stock trading. They share encrypted trade data with a vast community of data scientists and programmers, who then compete to produce the best algorithms using provided data and any other data that they can obtain, and they're paid in Bitcoin based on the fund's results. The algorithms then are combined via stacking and ensemble machine-learning techniques. Over 7,500 data scientists have built half a million machine-learning models that drive 34 billion predictions for the fund, and prediction error rates drop every week. One of the challenges that they have in Iceland is that everybody's related to everybody else. It's a very homogeneous society. While this makes for very fun family reunions, it makes it difficult to find a mate who's not your cousin. So, some enterprising young developers created an application that uses a database of family lineage including 720,000 current and deceased Icelandic natives. So, the app now allows Icelanders to bump their phones together to see how closely they're related. This app not only has the potential to reduce birth defects, but also uncomfortable family gatherings. Now let's look at some university examples. Georgia State University was facing a student advisory crunch and wanted to improve graduation rates by optimizing available resources. Their solution analyses 2.5 million grade records over ten years, to create a list of factors that hurt one's chances for graduation. It then applied predictive analytics and built an early warning system which it calls GPS for Graduation and Progression Success. The system prompts tens of thousands of in-person meetings between students and advisers and graduation rates are up six points since 2013. Also, graduates are getting their degree an average of a half-semester sooner than before, saving an estimated $12 million in tuition. Wichita State University wanted to predict student or applicant success rates and suitability for courses better. For current students, it used job and achievement data to predict if students would face problems. For new applicants, their particulars were matched with course requirements and the overall curriculum to assess their suitability. This led to fuller courses and greater student success in turn creating higher revenues for the college. The accuracy of the system increased from 82 percent to 96 percent in identifying high-yield prospects, while also saving the cost of hiring external analysts. Last but not least, let's turn to two examples in the auditing world. Deloitte wanted to streamline the process of detecting and reporting fraudulent behavior. The system they created uses vast amounts of transactional data, millions of rows of data and risky transaction detection scenarios. It runs various types of analyses like time-series, aggregations, and counts, and then writes stories in English via technology called narrative science to help identify which buyers and risk flags are the most important. It's able to create automated on-demand reports that accurately identify fraudulent behavior in time and expense transactions. It has reduced the time spent creating one report, from a week down to a few seconds. Similarly, PWC wanted to solve the problem of complex audit challenges by utilizing a new generation of analytic tools. Their new system uses discovery- based audit capabilities, along with comprehensive insights where high volumes of transactions are involved. It has the ability to reproduce complex business logic through the visual creation of business rules. Results are now delivered up to 90 percent faster than traditional tools, which gives them a strong competitive advantage within the auditing and consulting market, and it's given them a 40 percent greater accuracy in audits.
