Our companies operate according to plans. We draw up forecasts, draw up budgets, formalize a production program, and allocate our resources to execute this plan.

MRP2 logic is firmly rooted in the principle that the business plan, S&OP, master production schedule, capacity and constraints follow and cascade in a hierarchical and deterministic manner, with plans that are well aligned.

Of course, we know that in real life things won’t work out like that. We know that actual demand will differ from forecasts, that there will be unforeseen events in execution, and so on. To take this into account, we evaluate and readjust our plans at regular intervals. We review our budgets once a year, perhaps our financial projections once a quarter, our S&OP once a month, our MPS once a week, and on a day-to-day basis… come what may.

Our lives as supply chain managers are thus orchestrated around interlocking plans, measuring variance against these plans, and course corrections. We spend a great deal of time and energy drawing up these plans, debating their relevance, measuring and adjusting them – mind you, our next pre-s&op meeting is on Monday, and we’re not ready!

Is this actually worth the effort?

We are faced with this dilemma:

• We need to anticipate – otherwise, we won’t have the right resources available when we need them.
• Any deterministic anticipation will be wrong.
• We need to be able to respond reliably to real demand when it materializes.

We know that we’re dealing with stochastic phenomena – on both the demand and supply/production sides, everything is highly random. We try to respond in a deterministic way.

A stochastic model of response to random demand

What in Demand Driven jargon we call a “DDOM” – Demand Driven Operating Model – is a response model to a random demand. This response model itself is subject to uncertainty because we will have random supply events.  Rather than seeking to establish a deterministic plan, we define mechanisms that must respond robustly – agilely and resiliently – to demand.

This operating model must therefore be designed and prepared to respond to plausible eventualities – within a certain operating range. Our job as planners or supply chain managers is to test the operating ranges and, if necessary, intervene in the parameterization to adjust the bandwidth of events that our supply chain must be able to handle.

Stochastic instead of deterministic, what’s the difference?

It’s easy to have a single consensus-based sales forecast plan, from which you can deduce a production and supply plan. The sequence is rather simple to implement in our IT systems and is compatible with binary logic. But because it's based on a forecast we know it’s wrong!

Starting with sales hypotheses that include probabilities, applying them to an operating model that incorporates hazards, and deducing the adaptation measures to be taken to ensure that our model is sufficiently tolerant to enable an agile response to real demand and economically viable – this is another discipline, as much for our IT systems as for ourselves.

We might be tempted, if we put the problem like that, to try and respond to this uncertainty with a probabilistic, multi-dimensional, hyper-connected model – say, for example, creating a digital twin stuffed with artificial intelligence and on the lookout for weak signals. Something like that.

This would be to forget the essential point, which is that the best way to counter complexity is with simplicity. This is the strength of the Demand Driven model. By taking advantage of the principles of the Theory of Constraints (don’t optimize everything, focus on a few control points, take advantage of buffers, reason by operating range, and decide as late as possible) we create a control model that is understandable, visible, and adaptable.

There’s still work to be done….

However, taking these operating ranges into account is still an emerging practice in the Sales and Operations Planning process. It is often a challenge to support IT systems. No longer seeking consensus on a plan, running simulations to validate the limits of our model, planning the adjustment of stock, time, and capacity buffers – as an investment in agility and resilience – and drawing the consequences on financial scenarios – several companies have accpeted this challenge, but it is still in its infancy.

This means challenges for both our teams and our IT systems. As humans, we don’t like uncertainty. Preparing our company’s future by simulating behavior at the limits of a stochastic model is more uncomfortable than making one or more iterations of a deterministic plan.

It’s also a challenge for our software. We are enriching Intuiflow’s functionalities to provide better tools for this process, and to meet the expectations of our customers who are pioneers in this field, while keeping things simple. It’s exciting to be able to make this shift away from determinism a reality, but the work has only just begun, and much remains to be invented….

If you’d like to join the effort, don’t hesitate to contact us!