The Autonomous Supply Chain – Can we Fly by Wire?

Autonomous Supply Chain Blog Article

Supply Chain – “it’s not Rocket Science”

Many years ago, on a train heading out of London, I was sitting next to a recruitment consultant having a very loud phone conversation about a search to find a demand planner. The client had rejected all the candidates. “It’s not rocket science” was the phrase that the whole packed carriage heard. Many times.

So, let’s talk about rocket science.

Back In 1999, NASA lost the Mars Climate Orbiter due to a Unit of Measure conversion issue. The wrong thrust was applied and the orbiter burnt up in the Martian atmosphere. John Pike, Space Policy Director at the Federation of American Scientists, said: “I can’t think of another example of this kind of large loss due to English-versus-metric confusion, It is going to be the cautionary tale until the end of time.”

A little bit later, I was doing some work for a UK based company. A customer was unhappy about a 24-tonne order not arriving. Every resource was diverted to make sure 24 boxes arrived the next day. At least the delivery arrived intact.

Never underestimate the importance of accurate master data.

Degrees of Autonomy

Back to Rocket science. When we launch a probe to Mars, it has to act autonomously for most of the journey, especially the most hazardous part; the landing. At this distance from Earth, there are significant time lags – up to 20 minutes in terms of light, and sometimes days in terms of visibility between the probe in orbit around Mars and the base station. The rate we can transmit data in the short windows of opportunity is lower than an old-school telephone modem. “Joysticking from Earth is no longer an option” In these cases, we transmit planning level instructions to the probe and allow it to schedule the details and report back. A degree of problem-solving is allowed with a safe mode option. This was the case for the Beagle2 probe to Mars. Beagle2 was not entirely successful. It landed safely but failed to transmit anything back to Earth. 

A mission that was successful was the Apollo 11 landing on the moon. Most of the operations were planned in advance and executed in line with the plan, but the unexpected nature of the terrain required manual intervention from Neil Armstrong who landed the lunar module safely.  

This is an extreme example where the planners’ role is in sorting out the exception rather than managing the day to day. It is no coincidence that one of the first computer games was a moon landing simulator.

In Aviation planes “fly by wire” most of the time with the pilots involved only in monitoring except for take-off, landing and alerts. Fly by wire is based on a “performance envelope” which represents a series of limits that onboard systems aim to stay within by adjusting the available controls. The pilots don’t need to know the details unless there is an issue.

The use of a performance envelope for supply chain would work on the same principles, but we will always need our pilot. We just need to shape the envelope to give the right alerts at the right time

Which butterfly in which jungle? – Shaping the envelope

Autonomous planning is not an on/off switch. It is a journey. If we look at driving a car, there are 6 levels of autonomous driving from level 0 where the driver has full control through to level 5 where the car has no steering wheel, and therefore no pilot. At each stage, the requirements for accurate and timely data increases

It is possible to see all of the elements of supply chain planning as part of one process, like driving, that uses a single envelope to generate piloting requests. In this case, we would not worry about forecast accuracy falling below the normal threshold if we have the capacity to manage it, and profits are not adversely impacted. This is the car with a steering wheel that pops up only when needed.

If the performance envelope is breached, we need to know what the cause is – A forecast, a breakdown, a supplier issue? So the ability to show potential issues is still critical. We have to find the right butterfly in the right jungle. This suggests chaos. In many senses this is true, we have deviations from plan everywhere. Many of our numbers are not fixed – Forecasts, capacities, run rates, set-ups, defects, lead times…

So, what is the value of this approach? The answer is not in the elimination of the planner roles, it is in the evolution to pilot roles. When we react to every change in the incoming data, we disrupt the plan and cause chaos. If we use a finely tuned performance envelope we can minimize disruption. In order to do this, we have to define our envelope and understand the influences. In defining our envelope, we will embed our strategy including environmental and ethical policies. We will also be able to take back control when we have turbulence and adjust the envelope if needed.

For us in the supply chain, the risk is low. We will not lose a billion-dollar probe on another planet and years of work. The car will not crash if we find a problem with the plan. We can communicate quickly and react in time. 

We have the tools to enable this, we just need to bring them together. We can use Machine learning for statistical and promotional elements of the forecast, not to mention the possibility to make decisions on current period consumption by sales. We can add the influence of external leading indicators too, for instance, the weather.

On the production side, we can use carefully targeted linear programming techniques to solve the more difficult issues and use machine learning to schedule the plan. Then there is the potential for using supplier capacity portals, blockchain processing, reliable news sources, and so on to guide us.

This is the autonomous supply chain, a place where technology serves us, but we remain in control. A place where we set the direction of travel and the level of risk. It is not a black box. It requires pilots who can set and manage the envelope.

Let’s hope it’s not rocket science.

LEAVE A REPLY

Please enter your comment!
Please enter your name here