Graph Theory – Punctuality & Trending

I used to have a theory that if I was going to be late for a meeting, so was everyone else, so I shouldn’t worry. And this theory made me laugh and it saved my reputation when I breezed into the conference room just before everyone else did, because they were late and they were feeling bad about it.

Fast-forward to today, twenty years later, and graph theory is not just an idea for mathematicians or data scientists, it’s beginning to shape the real-world process decisions and algorithms which ensure a balanced ecosystem (a technology empowered ecosystem).

So when I’m delayed it is in fact part of a system problem? Yes and no. But I say yes, it’s never my fault! Realistically though, many problems are part of a system problem but it’s hard to see it with the naked eye. It’s easy to blame myself for being late or something I did because I don’t have all the tools to do a deeper assessment and I need to provide answers to all those people who are upset that I showed up late to another meeting….. But am I really late if everybody else is too? Or is being late the new “on time?” That is all about trends. A trend of being late is a system problem.

An analysis of an ecosystem as a graph, and for the purpose of this discussion, an acyclical undirected graph, yields powerful insights about the real-time flow of energy through a system. Abstracting each person as a node and the transactions and exchange of trend data as a relationship, can allow us to view trends before or during their development process. It is already well known and established that trends start as outliers and begin to concentrate in clusters and then move outward.

So the things that were making me late were probably making other people late? Sort of. The real answer lays in a much more complex problem which was uncovered in a research article for air traffic delays called “Systematic Delay Propagation in the US Airport Network” which asserts that the delays are a compounding “schedule” problem which is constrained by the concepts best analysed using graph theory

Flight delays represent failures to meet constraints imposed by a daily schedule. Its propagation in the network is a paradigmatic example of the way in which a distributed transport system moves toward collapse

But something that is still not fully understood is how the trends behave once they become trends. This is often where the feeding frenzy stops. The retailers for example are seeing the trend after it has fully matured, they are getting there too late…. but so is everyone else…. So why is the late-to-trend failure so pronounced for retailers if they are just as late to the party as anyone else? Most likely due to the reaction and response time that they experience as a natural result of moving and preparing physical product for sale.

It’s arguably true that if trends could be spotted early or even developed preemptively, physical retail stores would have no problem selling their product across channels. This is where graph analysis will show the most promise going forward:

a. Developing trend-able products which have trend-friendly characteristics
b. Real-time trend analysis yields repeatable pathways
c. Early discovery
d. Faster response and delivery of trend-friendly products
e. Using the graph, have ready-made levers which act on the pathways
f. Becoming “micro-trend” aware
g. Becoming “micro-trend” ready
h. Develop flexible supply-chains which efficiently adopt and discard trends

The solution to being late-to-trend isn’t focusing on being “trendy,” it’s about being graph aware and micro-trend flexible.

Digital attribution, especially multi-touch and cross-device is no longer the ultimate answer to solving these trend issues, partly due to CCPA and data privacy rules (which all data engineers must adopt or face penalties), but also because it is another “look back” approach. We are now in an age where modeling based on graph theory is the only way to predict behavior. Welcome to a new era.

And this is the reason data can and will solve the retail woes. Hiring data engineers is part of it, but having smart marketing scientists is key as well.

Please contact me if you are looking for a data team, I am working with an org right now that understands these things and has the resources to work in-house as a consultancy to help you get micro-trend ready.