Why the Graph Model was the Hardest Thing but the Best Fit
2021 NEXUS ONLINE DATA ENLIGHTENING SERIES, (NODES) SEMINAR SERIES
Thursday, February 25, 2021
Online Webinar & Q&A
Livestream Link – YouTube: https://www.youtube.com/watch?v=mScZTePZTPg
Q&A – Sli.do, event code: V214
Abstract: Over the years we have developed many complex ways of measuring profitability in our business. One of the main sources of that information was a program that measured operating ratio (OR) or the ratio of cost over revenue. 10 years ago we realized that this program was outliving its usefulness as our business had grown and diversified and it could not keep up with the change. In 2013 we decided to develop tools to support the program and measure things it could not. That resulted in several complex relational database structures being created. Those structures served us well, but a large number of table joins, some of them cross-server join, proved to be unstable and unsustainable. In 2017, we proposed a move to a graph model and our business rejected the proposal on economic grounds. The proposal was revised, re-presented and accepted in 2018. After a failed attempt with one graph model package, we finally settled on Microsoft Graph in 2019. Since then, we have developed a model that allows us to image our network at any one time, and over time, at the very smallest quantum from a cost and revenue perspective. Hundreds of thousands of relationships have been established through algorithm development and building sustainable data sources. Today, we have a viable graph model that continuously images our transportation network, allowing us to understand sub-networks, customers, driver and asset performance, and many other layers of relationships from the simple establishment of nodes and edges. On this platform, we are building a dynamic pricing model.
Senior Director, Network Analytics and Development, Bison Transport, Inc.