Your production schedule is a spreadsheet rebuilt every Monday

It is wrong by Wednesday, nobody but its author can read it, and your customer lead times are quoted on guesswork.

Thirty minutes, no commitment

Describe your flow. I'll tell you what can be automated and where I would start.

The production manager spends Monday morning — sometimes Sunday evening — rebuilding the week's schedule. They know their machines, their operators and their urgent jobs, and they produce something sound. Then a rush order lands on Tuesday, a machine stops on Wednesday, and the file becomes a historical document nobody opens.

The market's usual answer is to buy an MES or an APS. Between 20 and 250 employees that is very often the wrong decision: the project runs a year, costs six figures, demands a data quality the shop floor does not yet have, and ends in partial use. There is a middle road, and it is the one I propose.

What it looks like in your shop

The schedule and the ERP disagree

The ERP holds the works orders, with theoretical dates calculated from routing times nobody believes. The real schedule lives in a spreadsheet. The two diverge permanently, and the shop follows the second — so the ERP produces false delivery dates that sales admin quotes to customers.

Load per machine is a hunch

Ask « how many weeks of load does the 5-axis mill have? ». The answer comes after a pause and an estimate. That question ought to take three seconds: it decides whether you accept a job, whether you hire, and whether you invest in a machine.

Routing times have never been revised

They date from the ERP implementation, were entered in bulk from an estimate, and sometimes include setups that no longer exist. Because they are wrong, nobody schedules from them, and because nobody uses them, nobody corrects them. That loop has to be broken at one specific point.

Priority is decided verbally

The sales rep comes to see the shop supervisor, the job jumps the queue. That is not illegitimate, but nothing records the trade-off or its cost to the other jobs. At month end, nobody can explain why three customers were delivered late.

One person knows how to run the place

When the production manager is on leave, the shop runs in degraded mode on the team leaders' memory. It is the most serious risk on this list, and the least discussed, because it concerns a person.

What it actually costs

The cost of an unreliable schedule does not appear on any accounting line. It appears in four indicators you can pull this week.

Rescheduling time: 4 to 8 hours per weekMonday morning plus the daily reworks. That is production manager time — the time that should be going into improving processes.
On-time delivery rate: pull it for the last 6 monthsCompare the date promised at order with the actual dispatch date. Many companies have never calculated this and find it 15 to 25 points below what they assumed.
Routing time versus actual: 20 to 50% on older routingsTake ten completed works orders and compare routing time with booked time. It is the most useful measurement in this whole project and it costs half a day.
Investment decisions without load data: 1 machine = €150,000 to €400,000Buying a machine because saturation is « felt », or not buying one when the load justifies it: that is where the real money is decided.

There is a direct commercial cost too: without reliable load, you quote lead times with a safety margin. That margin loses you jobs against a competitor who knows what they can promise.

How it gets fixed

The approach is incremental and deliberately modest. We are not aiming for minute-level optimised sequencing: we are aiming for load that is correct to the day, which is enough to decide.

1. We make the times reliable before scheduling anything
You cannot schedule on false routings. We start by measuring the gap between routing and actual on the part families that carry the volume — not on 3,000 items, on the twenty that make up most of the load. We correct those routings. It is unglamorous and it is the only solid foundation.
2. Progress reports back from the floor without paperwork
A screen per cell or a scan on the works order: start, finish, good quantity, scrapped quantity. Two actions per operation. Without that feedback no schedule stays true for more than twenty-four hours — and that, not the fact that it lives in a spreadsheet, is the real reason spreadsheet schedules fail.
3. Load per work centre becomes a screen, not a file
Load by machine and by week, fed by real works orders and corrected times. Readable by the production manager, sales admin and management, at the same moment, with the same figures. « Can we take this job for the 15th? » gets an answer in three seconds.
4. Trade-offs are recorded
When a job jumps the queue, you record who decided and what it pushes back. This is not bureaucracy: it is what lets you explain late deliveries at quarter end with facts instead of impressions, and have a calm conversation between sales and production.

If a real MES is justified after that, you will buy it with clean data and a team used to reporting progress — that is, in the conditions where those projects succeed.

What this does not fix

What this does not fix — and here the list is deliberately long:

This is not an MES and it is not an APS. There is no automatic finite-capacity sequencing with multi-criteria optimisation. If your situation genuinely requires that, I will say so and I will not take the project.
It does not schedule for you. The production manager still decides; they simply decide on current figures.
It does not work if the shop floor does not report progress. That is the essential condition, and it is a matter of organisation and of dialogue with the workforce as much as of technology. If the team leaders do not believe in it, the project fails, and no tool compensates for that.
It does not increase your capacity. It tells you precisely where capacity is saturated, which is often a surprise and not always good news.
It does not make up for a shortage of skilled staff, which is probably your real constraint if you are in Wallonia looking for machinists.

The questions I get asked

What does it cost?

The two-day audit is €1,500. A load-and-progress project lands between €12,000 and €15,000 fixed price. It is the heaviest project on this list, because making the times reliable takes time and cannot be skipped.

How long before it runs?

Four weeks for progress reporting and the load screen on one area of the shop. Fully cleaning up the routings is continuing work that then belongs to you; I leave you the method and the measurement tool.

Why not simply buy an MES?

Because between 20 and 250 employees an MES most often fails for three reasons: the master data is not reliable, the shop floor is not used to reporting, and there is nobody internally to keep it alive. None of those three is solved by budget. Deal with them first; if an MES still makes sense afterwards, it will succeed.

Do we need to change ERP?

No. Works orders stay in your ERP. What gets added is actual feedback and the load view — two things most SME ERPs do badly or not at all.

Will operators accept booking time?

That is the central question, and the answer depends on how it is presented. Time booking experienced as individual output monitoring gets sabotaged, everywhere, always. Time booking that replaces a paper sheet and makes it possible to show that the load really is excessive gets accepted. I always design these screens with the team leaders, and I recommend talking to employee representatives beforehand, not afterwards.

Can we keep a spreadsheet somewhere?

Yes, and it is often kept for long-term simulation, where it is very good. What it loses is the role of single source of truth about the state of the shop, which it cannot hold across several people.

What if you're no longer available?

The code, documentation and credentials are with you, and everything runs on your infrastructure.

In your trade

This job is not equally urgent everywhere. These pages say where it sits in the running order, trade by trade.

Let's talk for thirty minutes

You describe your flow, I tell you what can be automated, what cannot, and where I would start. If I'm not the right person, I'll say so during the call.