Supply chain automation and the return of human intelligence

When it comes to supply chain optimisation, people tend to focus on automation. For years we’ve focused on efficacy, MRPs, and cost efficiency. And all enterprises will have an ERP automating tasks and spitting out orders – that’s a given.

However, given current market complexities, relying purely on automation comes with a risk. Even with the latest and best technology, automation and artificial intelligence isn’t quite sophisticated enough yet.

So while it might sound a little counter-intuitive, here at Avnet Silica we always add some human intelligence back into the mix via consensus planning.

Considering the context of demand signals

Say an OEM changes something, a demand signal/change comes in and our system uses MRP parameters (lead time, stock target, transportation times etc) to suggest changes further upstream. Based on this automated system we’d then say to the supplier, for example, ‘we don’t need these 10,000 pieces this week, we need them next week instead’.

That’s how the industry works.

For our customers, though, automated forecasts can swing hugely from week to week. And even the fanciest systems aren’t advanced enough to account for how the supplier will react. 

You can, however, have the MRP configured to also look at things like historical forecast accuracy. For example, if we know a customer is pretty consistent at only ever consuming about 80% of what they forecasted, it can take that into consideration in what we're asking for from the supplier.

The problem is, the market is volatile at the moment.  Our confidence in the future based on the past isn’t so high. We need to consider other factors before pushing orders through to suppliers.

That’s why we add the value of consensus planning.

Before we push orders through, we validate, analyse, apply some reasoning and mitigation.

Because there can be a significant impact, especially in the current state of play, to dropping your orders. If you need to pick it up again, you could find yourself at the back of the queue.

The system might do most of this, but we add in a programme manager (i.e., a human) into the automated system, to add in some context.

That’s how we account for the demand signalling side of things.

There’s also demand sensing.

Demand sensing: looking ahead

Demand sensing is softer than a demand signal. It’s not hard and fast, like a purchase order, for example.

It’s thinking about what’s going on in the broader context for the customer, and taking decisions to plan for supply ahead of the curve.

Take a new product integration, for example.

If we know a customer is working on a new project/platform and it’s going to use a new part number, typically they'll go through some rounds of prototyping. So they'll do sampling and then they'll go into prototype a, then b, and so on. Then they move into qualifications. 

Once everyone's happy the product actually works, they’ll move into pre series production which is very, very low volume. Say a car maker just wants a few infotainment systems so they can do some more extensive testing before they go into series production. If all goes to plan, production then typically ramps up over a five to seven year period.

An example of demand sensing would be if we know that the customer is in the prototype stage, we would think about when they might switch from pre series to series production? Systemically, that will happen overnight. So they will go from needing a small number of parts to needing hundreds of thousands very quickly.

Demand sensing is about looking outside of the usual signals. Where else can we add value? If we can see ahead to series production coming within six months, and we know lead times as they are right now, we can start to pre-position supply to ensure our customers get the parts they need.

One of the reasons it’s called ‘sensing’ is it’s a risky game. Because the customer could finish the prototype C build and have a quality issue, which could kill the project.

So we might have discussions with the engineering teams, for example, at the customer and at the supplier’s end to gauge confidence levels. And in some cases, the supplier might need us to do a pipeline fill to enable them to secure the capacity at their suppliers for these projects. So we can come in there as a distributor and pre position the supply.

For now, and the near future, automation is not the complete solution.

Of course, in our world automation is important. But, unfortunately, the current environment is unstable. It’s not unusual to see a sudden and sharp increase or decrease in orders, only to swing back again the following week. And if you drop your orders, you don’t want to find yourself at the back of a long queue. As it stands, relying on purely automated forecasts could hamper your ability to secure the supply you need. The technology just isn’t quite there yet. One day, no doubt!

Until that time, we highly recommend adding human intelligence back into your automated systems to optimise your supply chain.

We know that each change a business makes to its supply chain comes with risk. And we know there’s no one size fits all. So, if you'd like expert support with your supply chain challenges, please get in touch. We'd love to see if we can help.

Follow Avnet Silica on LinkedIn