The digitization of the EMS-OEM-ODM supply-chain

What, When & Where to start

The digitization of the electronics supply-chain is gathering pace, but what will actually happen in the future isn’t yet obvious to everyone. Here is my view.

All the possible outcomes for the future of the supply-chain are frightening to anyone working in the industry today:

  • change will be rapid;
  • disruption will take place;
  • many current jobs will completely disappear.

However (and it is a big, reassuring HOWEVER), it is my view that a fully digitized, end-to-end supply-chain is going to take much longer than most people think.

At the recent Arrow Electronics digital forum in France, we spoke to an audience of supply-chain pros about what will change, when we think the changes will come by understanding the underlying technology and finally, where manufacturers, suppliers and the rest of the chain can immediately start their digital journey so they don’t just survive, but thrive.

Will the digitized supply-chain resemble Minority Report?
Tom Cruise in Minority Report. © 2002 Twentieth Century Fox Film Corporation


Our model of the future looks like this:

  • Every single task that happens more than once every day will be automated;
  • Thanks to Artificial Intelligence, this automation will be contextual (taking into account market conditions, end-customer demand, supplier relations, etc);
  • Humans will interact with robots (Avnet robot, STMicro robot, your own robot) on a simple, Slack-like interface.

Why is this frightening? Because in this model, most current jobs become redundant.


It will take much longer than most people think. To figure it out, you must first understand the underlying technology which makes it possible.


Let’s start with a simple machine learning example: the image recognition technology which distinguishes between a cat and a dog. It seems like an easy task, but there is a lot of variability amongst cats and dogs in size, shape, breed, hair etc. Unlike a human who can easily understand with high probability, it is actually quite difficult for an algorithm.

Artificial Intelligence can recognise cats and dogs, but not electronic parts

This is where machine learning comes in. The algorithm trains itself on a set of solutions (images of identified cats, images of identified dogs), to be able to sort new cat or dog pictures. To be trained properly, the algorithm needs a HUGE amount of cat pictures and dog pictures.

This is often the most challenging part. The machine learning technology has been effective for many years, but it is still very difficult — if not impossible — to have enough clean data solutions to train the algorithm.

That is a job for reCaptcha.

reCaptcha is the anti-hacking tool you’ve probably seen on websites which ‘verifies’ you are human. It actually serves two functions, the second of which is that it learns to distinguish between the animals, based on feedback you provide verifying what is, and is not, a ‘cat’ or ‘dog’. Of course you may notice more road signs these days than cat pictures because Google, which owns reCaptcha, is now investing more resources on the autonomous car than on cat and dog recognition…

ReCaptcha is an effort to boost learning in artificial intelligenceNow, back to the supply chain. Think of the current data actually needed to create an end-to-end, digitized supply-chain. Do you currently have digitized access to all the relevant information within your supply-chain? Your contacts, suppliers, clients, all your ERPs, all your systems? Do you have access to the digitized, organized and clean emails of your procurement teams, your customers? All the phone calls, the escalation meetings, the planning decisions? These are relevant if you want to distinguish, for example, a high shortage risk from a low shortage risk. What’s more: is the data that is already digitized actually clean? Are your MPNs, your lead-times?

You can see my point here. If you want an autonomous car, first you must digitize the signal of each and every data point which the autonomous car needs to understand so it can drive. If you want an autonomous supply-chain, you need to digitize all relevant signals in the component procurement process. That is why, before we had the autonomous car, we had assisted driving. For the same reason, before the autonomous supply-chain, there will be the assisted supply-chain.

This is why I think it will be at least another 15 years before we see the fully digitized, end-to-end supply-chain. Because, unlike autonomous cars, there isn’t enough added value in digitizing it just yet.


That is simple. Clean your data. Set up processes and integrations to keep it clean and in-sync with all your systems and the entire market at all times. Start with Manufacturer Part Number (MPN). Then move onto Lead-Time. Trust me, there might be some work to begin with and some initial costs, but it will save you A LOT of time in the long-run, and even sooner, a lot of money.

* * *
PS: We want to create a tool, free and accessible to all, allowing you to upload simple data like MPN and Lead-Times to check cleanliness against external data sources. But we’re busy people, so if you want to see and use something like this, then contact us via twitteremail or LinkedIn and let us know. If even just a few people see value in such a thing, we will make it happen!

Want to see Precogs in action?

Contact us to find out how Precogs lowers costs
while improving purchase processing and efficiency.