Noodle.ai's war on industrial supply chain waste - a lofty ambition or practical solution?

Profile picture for user gonzodaddy By Den Howlett February 26, 2020
Summary:
Noodle.ai saves money while helping reduce industrial waste. That's a net good. How does it work?

Stephen Pratt Noodle.ai CEO
Stephen Pratt, CEO Noodle.ai

Many years ago, I heard the CEO of a firm discuss solving demand planning issues through the use of algorithms that could iron out constraints and so lead to nirvana for demand planners. While compelling, it was a lofty ambition that turned out to be a pipedream. 

In the real world, there is no such thing as the perfect supply chain, but there are many elements inside those chains that can be optimized. The problem is that inefficiencies coupled with inaccurate or incomplete data, make a planner's job more an art than a science and where 'good enough' is shorthand for best guesses. Noodle.ai is tackling parts of that gnarly problem, focusing on areas where there is a significant waste.

To understand what this means, I recently held a conversation with Steve Pratt, CEO Noodle.ai. As background, Pratt has a long history of looking at problems that seem intractable and then going out and building solutions. Historians of the outsourcing world might remember that he established Infosys Consulting as a way of turning Infosys from being a technology provider into that of strategic partner, eventually building a multi-billion dollar business unit in the process. 

Consulting was an industry that was ripe for us to come along and make a differentiated play. We did it, and that's what we're looking to do with noodle.ai in manufacturing and supply chains. 

Pratt explained that according to some research, industrial waste constitutes over 90% of the waste in the world. Two data points, from supply chains alone: First, in 2018, Hackett Group's working capital study estimated that $443 billion is locked up in inventory. Second, and according to a report from 2017 by Matt Barlin

The (US) Environmental Defense Fund estimates 15–25% of truck miles are empty. Using the low end of this range, assuming carriers operate with a deadhead rate of 15% of their total miles, yields 41.9 billion empty miles for all trucks and 25.5 billion empty miles for combination vehicles.

These are huge numbers by any measure. So, where does Noodle.ai see the most pressing problems?

The big problem that most companies have right now is “I've made all this stuff. Where do I send it?” And you get simultaneous stock-outs and excess inventory. Current ERP systems fail in that process, mainly because ERPs represent a series of workflows, and those workflows are a series of actions and decisions. So, think of the actions as boxes and the decisions as diamonds. ERP is really, really good at the boxes but not so good at handling the diamonds.

By way of illustration, Pratt gave the example of a large consumer products company that has over a hundred thousand SKUs, operating across more than 50 countries, with distribution centers everywhere. They run their ERP system over the weekend and end up getting hundreds of thousands of inventory alerts because the system 'thinks' deterministically rather than in terms of probabilities. In essence, the ERP gives up, and even where it can process data, the fixed nature of the rules it uses results in suboptimal results. 

So, where do you start? After all, there are plenty of C-suite officers mandated to deliver higher revenue, margins, and, ultimately, improved profits. Here, Pratt believes that while consulting engagements can produce useful roadmaps, the way he sees that play out leads to failure. 

In the traditional world, consultants create a risk/reward-matrix at the end of which they'll likely say - ok, let's develop these dozen AI solutions. This is wrong. Worse still, and because we're at the beginning of this AI cycle, companies will typically try to build their own solutions, much as they tried to custom develop CRM systems early on. An AI strategy should be centered around commercially available AI applications like we have developed at Noodle.ai.  

What we say is that you have to upend the traditional software model by starting with the data and then letting the data write the code. This way, the applications algorithm quickly discovers whether the problem has a solution within the data set. Better still, instead of spending many millions of dollars to arrive at an answer, you can use data you already have  to quickly achieve a return on your investment - typically inside a year.

Another problem planners face comes in the shape of conflicting KPIs between staff members and the inherent biases that creep into those same KPIs. 

One of the most exciting aspects of what we do is that Noodle.ai applications pick up on those biases and compensates for them to arrive at an optimal outcome. But getting there means we need top-level buy-in because inevitably, the results Noodle.ai provides implies a degree of change management.

In listening to Pratt, I was wondering if  Noodle.ai products threaten the jobs of value planners.  The question brings an emphatic 'no.' In his experience, planners know things algorithms cannot.  Planners know where to look for answers and the algorithms do the precise math and generate optimal recommendations. 

As we closed out our conversation, I asked Pratt where the idea of ridding the world of waste fits in, given that past sustainability efforts often flounder on the altar of shareholder value. The answer is simple yet effective. 

Most customers don't come to us to save the world, but they do come to us to find savings or revenue. The areas we concentrate on, especially inefficient supply chains, have the benefit of not just saving money but contributing to a more sustainable businesses. It is no longer a choice between profits and the planet.  This breakthrough technology allows us to have both profits and the planet.

My take

The sheer scale of what Pratt describes is sufficiently attractive to garner the attention of any C-suite officer. The fact that Noodle.ai starts with existing data, applying prebuilt models means that customers can quickly see the value. And while the initial training of AI models requires data engineering and data science - intensive cycles, subsequent upkeep of algorithms is fractional. Over time, the adoption of AI applications will be commonplace, and, if Pratt is right, we will see these types of system routinely solving some of the most pressing problems in reducing operating inefficiencies and industrial waste.