V Net Unleashes the Machine

By Sanjeev Balasubramaniam

We’re always trying to innovate here at V Net and a key set of emerging technologies that we’ve been looking at is Artificial Intelligence (AI) and Machine Learning (ML). The significance of these technologies for V Net is their ability to make forecasting more accurate.

For the uninitiated (which included me up until about 5 minutes ago), the simple difference between AI and Machine Learning is nicely outlined by Forbes[1], as follows:

~ Artificial Intelligence is the broader concept of machines being able to carry out tasks in a way that we would consider “smart”. An example of this is the Atlas robot shown in the links below.


~ Machine Learning is a current application of AI based around the idea that we should really just be able to give machines access to data and let them learn for themselves. An example of this would be a website that serves you up pages based it thinks you like based on your initial clicks.

As you can see both Machine Learning and AI start from the same base. However, for our purposes of inventory forecasting and management, Machine Learning is a great first step to achieving our goals of having a true learning algorithm that can move from forecasting into predictive ordering.

Before we all get too excited though, I’d like to bring your attention to the diagram below created by renowned consulting firm Gartner. They’re famous for their “Hype Cycle” series which illustrates the truth behind how practical these buzzword innovations truly are. As you can see Deep Learning (AI) and Machine learning are currently in the “Peak of Inflated Expectations”. We are yet to fall into the “Trough of Disillusionment” before rising into enlightenment and then productivity.

Gartner Hype Cycle [2]

So, hype aside, V Net is working on integrating Machine Learning into our forecasting process. We have recently completed a Proof of Concept involving three AI consulting firms in a “bake-off” to see if they could accurately create 13 week daily forecasts for hardware SKUs and a 9 month long horizon forecast for commercial and Accessories SKUs. As Gartner prophesised, the results were disappointing from an accuracy standpoint, not being as accurate as the results we can currently achieve.

However, there were some promising learnings from the exercise and we’ve realised that for ML to truly succeed it needs to take on board “domain knowledge”, which is the day-in-day out business acumen and category expertise that we have all built into V Net over the last seventeen years. In fact, we may find that ML is not as accurate as V Net or may be overkill in some categories.


Our next step will be to partner with the most promising firms from the Proof of Concept to work on a project that gives ML some domain knowledge and measures the results from there.

Over the long term, ML can also be integrated into pricing optimisation and AI can be focused on optimising orders and harnessing Big Data within V Net Insights. The sky (and our ability to filter through the hype) is the limit!


View Tony’s latest quarterly update

V Net