Almost every factory doing SPC is doing it in a spreadsheet, and the spreadsheet usually works. It was built by someone competent, it draws a real chart, and it has been in service for years.
This page is not about Excel being bad at maths. It is about the four things that quietly go wrong, all of which look fine from the outside.
1. The limits move
This is the big one.
In most spreadsheet control charts the centre line and limits are formulas over
the same column the chart plots: AVERAGE(B2:B200), STDEV(B2:B200). Add a row
and every limit shifts a little.
That is not a control chart. A control chart compares what the process is doing now against what it did when it was behaving. If the limits recompute from the current data, a slow drift drags them along with it and the chart never signals — it stays reassuringly in control while the process walks away from where it started. The failure is silent, and the chart looks perfect throughout.
Correct behaviour is to compute limits once from a baseline window when the process was stable, freeze them, and judge everything afterwards against those numbers. That is a discipline a spreadsheet cannot enforce, only permit.
2. Nobody is watching
A spreadsheet chart shows a rule violation to whoever has the file open. If a run of nine points on one side of the centre line happens at 02:00 on a Sunday, it is discovered on Monday, along with everything made since.
Run rules are the reason SPC is worth doing at all: they catch a process moving before it makes scrap. Most of that value needs something watching the data between the times a person is.
3. Only some of the rules are in there
Rule 1 (a point outside the limits) is easy to write and nearly always present. Rules 2 to 8 need run-length logic — nine in a row on one side, six rising, two of three in the outer third — and in a spreadsheet each one is a column of helper formulas that has to be maintained and dragged down whenever data is added.
In practice most spreadsheets implement one or two rules and inherit a chart that misses the patterns rule 1 was never going to catch.
4. It has one author
The person who built it knows why row 4 is excluded, which sheet feeds the chart, and what the hidden column does. When they change job, the file becomes a thing nobody dares modify. This is the most common reason a plant's SPC quietly stops.
There is also no audit trail. When somebody asks why a point passed in March, the answer has to be reconstructed from a file that has been edited since.
What "without Excel" actually needs
- Frozen, versioned limits with the baseline window recorded on them.
- All the rules, evaluated continuously, with an alert when one trips.
- Data arriving on its own — CSV import, API, or straight from the line.
- A read-only view everyone can see without being able to break it.
- A record of every limit change, tolerance edit and re-baseline.
Where Excel still wins
Genuinely:
- Ad-hoc questions. "What did this look like if we exclude the night shift?" is thirty seconds in a spreadsheet.
- It is already there, and everybody can use it.
- Nothing to buy.
Keeping Excel for exploration and moving the standing charts somewhere that enforces the discipline is a perfectly good answer. That split is what most plants land on.
Start from the spreadsheet you have
Export it as CSV and upload it. Columns get mapped in the browser, the characteristic is created from the file, and you get a chart with limits you choose and freeze — nothing here asks you to model your factory first. There are also free calculators if you only want to check what the spreadsheet is telling you.