Operations

The planning software nobody dares to question

Mohit Saini · 2 Feb 2023 · 5 min read

The planning software nobody dares to question

Picture a plant head proudly showing a visitor the new scheduling system his company has just bought. The screen was impressive: colourful Gantt charts, every machine loaded to the minute, a button to reschedule the whole plant in seconds.

Then they walk onto the floor. The supervisor at the first machine is running a job that isn’t on the schedule. Asked why, he shrugs: “The schedule changes three times a day. I know what is really urgent.”

Nobody in the review meetings says the obvious: the plant is not being run by the system it paid for. The scene is common enough that it is worth asking how it keeps happening.

Decades of promises

From the 1980s, MRP II (and later the planning modules of ERP) promised to bring order to the shop floor. The logic was compelling: with a clean bill of materials, accurate routings, correct inventory and a realistic master schedule, a computer could calculate exactly what every work centre should make and when. If everyone followed the schedule, inventory would fall and deliveries would be on time.

The market for these systems grew into billions of dollars. And yet, alongside the growth, there was a steady stream of stories of companies that bought the software and never used it to run the floor, or used it and saw performance get worse.

How does a solution keep growing while so many users are disappointed?

“It failed because of the implementation”

Look at how failures were usually explained. For years, the dominant explanation was obstacles: lack of top management commitment, poor data accuracy, untrained people. If failure is blamed on obstacles, the solution itself is never questioned. The next company simply resolves to try harder.

Later, critics began to challenge the logic itself: MRP assumes unlimited capacity, and its capacity checks are rough at best. The answer to that was a new generation of software: advanced planning and scheduling, which uses sophisticated algorithms to produce a mathematically optimal schedule across many constrained resources, and can reschedule at the click of a button.

The problem was redefined twice: first as a problem of visibility, then as a problem of mathematics. Is either of these the real problem?

Why the new suit doesn’t fit either

When I look at how advanced schedulers are used on real shop floors, I see a familiar set of difficulties:

  • They model today’s chaos. On a floor flooded with work in progress, bottlenecks seem to move every day, so managers believe they have many constraints, and the software is configured to optimise them all. Optimising many constraints at once keeps work in progress high.
  • Rescheduling becomes a habit. If the schedule is recalculated every time something goes wrong, how can delivery promises be kept? Each reschedule quietly moves some customer’s date.
  • Buffers are added by guesswork. To avoid constant rescheduling, buffers are inserted in the schedule. But then the schedule is no longer optimal, and without feedback nobody knows whether the buffers are too big or too small.
  • Time-based priorities mislead in make-to-stock. The schedule may chase an order that is “late” while the warehouse holds plenty of that item, and ignore an order that is “on time” while the item is about to run out.
  • Nobody trusts a black box. Supervisors can’t see why the algorithm chose a sequence. When it doesn’t make sense to them, they follow their own judgement. This is then labelled a “discipline problem”.
  • The data burden grows. Companies that struggled to keep routings and stock accurate for MRP are now asked for even more precise data.

The real problem: a conflict, not a calculation

Ask any experienced production manager and he will describe the real dilemma in one breath. If I focus on keeping machines busy, orders get delayed. If I rush orders through, efficiency drops. So the plant swings between the two: efficiency for most of the month, then a frantic push for dispatches at month end.

Neither MRP nor an optimiser resolves this conflict. MRP actually embedded it, because standard lead times were set from the old, batch-heavy way of working, and product costing rewarded keeping every machine busy. The optimiser lets managers choose which side to sacrifice on any given day.

A conflict cannot be solved by better arithmetic. It is solved by challenging the assumption behind it.

The assumption here is that focusing on flow costs capacity. In most plants, it doesn’t. Touch time is typically less than a tenth of total lead time; the rest is waiting. Excess work in progress doesn’t use capacity well. It hides it.

Real clothes: a simpler way

Theory of Constraints offers a different approach, which needs a change in thinking more than a change in software:

  1. Choke the release. Release material to the floor only a fixed time before the due date, for example about half the current lead time. Work in progress falls, and the real constraint becomes visible.

    Picture a plant that quotes 30 days and has about 30 days of work sitting on the floor, though each order needs only 2 or 3 days of actual processing. Release orders 15 days before their due date instead of as soon as they arrive, and the floor holds about half the work. Orders now wait in shorter queues, so most finish well within 15 days, and the machines that looked overloaded turn out, mostly, to have been busy working on the wrong jobs too early.

  2. Load the real constraint only. Accurate data is now needed in one place, not everywhere. Due dates are quoted from the load on that one resource.

  3. Don’t reschedule; prioritise. Each order gets a time buffer. Work centres follow a simple colour signal based on how much of the buffer has been used. No black box.

  4. Learn from the buffer. Record why orders turn red, and fix the most frequent causes.

This is simple enough for every supervisor to understand and follow, which is exactly why it works.

Perhaps the hardest part is admitting that simplicity can beat sophistication. In planning, it sometimes takes someone willing to say the obvious out loud: the schedule on the screen is not the schedule on the floor.

Related reading

Mohit Saini

Strategy, supply chain & sales transformation consultant. About · Book a discussion