Seven workbooks in a folder on the server. Three copies of most of them. Formulas written by someone who left in 2019 and one person who knows which version is current. It has worked for years and it is still working, which is exactly why nobody has dealt with it.
This page is about how to tell whether yours has outgrown itself, which part to replace first and what that actually costs.
Most firms I speak to have somewhere between five and nine of these. See how many you recognise.
If your best inspector left on Friday, how long would it take somebody else to answer "is gauge 114 in calibration and where is the evidence" without ringing them? If the honest answer is more than a minute, the system is running on a person rather than on a process.
I want to be fair to the spreadsheet here, because a lot of software marketing is not. Excel is one of the best tools ever made for working something out. It is fast, everybody has it, everybody can use it and for a firm doing a handful of these records a year it is genuinely the right answer. Replacing it would be a waste of money.
The trouble starts when the working-out becomes the evidence. A record has to do things a calculation does not. It has to be findable by someone who did not create it. It has to be provably unchanged since it was signed. It has to be readable in eight years by a person who has never met you. And it has to stay true while several people work on it at once.
A spreadsheet was not designed for any of that and no amount of care makes up the difference once the volume rises. That is not a criticism of whoever built yours. It is a mismatch between what the tool is for and what the job has become.
There is the one on the server, the one on a desktop and the one someone emailed to a customer in March. All three look identical and open the same way. Version control by filename is not version control and quality_log_FINAL_v3(2).xlsx is the symptom rather than the cause.
One person has it open for editing, so the second works on a copy and merges it later. Usually. The merge is manual and the mistakes it makes are silent, which is the worst kind, because nothing tells you a row went missing.
A value changes and the spreadsheet keeps no record of who changed it, when, or what it was before. Nobody is being dishonest. But a record you cannot demonstrate is unchanged is a weaker record than one you can and an auditor is entitled to notice.
A calibration falls due and the cell turns red, but only for a person who opens the file. An action sits waiting three weeks because nothing makes waiting visible. Every deadline in the system depends on somebody remembering to look.
A row gets inserted outside a range. A sort is applied to one column. A lookup points at a sheet that was renamed. The workbook keeps producing numbers throughout, which is why these are usually found long after the decisions based on them.
Whoever can open the file can edit anything in it, including last year's closed records. There is no practical way to let a machinist add a line without giving them the ability to overwrite the sheet.
You will occasionally be told that a spreadsheet quality system will fail an audit. That is not what the standard says and it is worth being precise about what it does say so you can judge the claim for yourself.
ISO 9001:2015 names no software at all. Clause 7.5 deals with documented information and clause 7.5.3 is the one that matters here. It requires that documented information is available and suitable for use where it is needed, that it is adequately protected against loss of integrity and improper use and that it is controlled for distribution and access, storage, version, retention and disposition.
Read that list against the six failures above and you can see where the pressure comes from. A spreadsheet system can satisfy every one of those requirements. Plenty do. It simply takes more discipline each year, applied by more people, as the number of records grows and it is the consistency rather than the capability that usually gives way.
A revised ISO 9001 is scheduled for publication on 16 September 2026, with a three-year transition period anticipated. On the drafts published so far it is an evolution rather than a rewrite. Until it is published nobody can tell you exactly what the final text says. The longer note on the 2026 revision is on the NCR page.
This is the part most people get wrong and it is usually the reason a previous attempt stalled.
Replacing an entire quality system in one go means a long build, a large bill, months before anything is usable and a switchover during a week when everybody is busy, which is every week. It also means that if one part turns out to be wrong, the whole thing is held up while it is fixed.
Doing it one workbook at a time means the worst problem is solved in a fortnight, the thing is in real use while the second part is still being discussed and you find out early and cheaply whether this way of working suits your firm at all.
Most people touching it, shortest deadline attached, worst consequences when it slips. That is usually the non-conformance register or the calibration schedule.
Same fields, same numbering, same rules, with the history imported. From two weeks. The other six workbooks carry on untouched throughout.
Long enough to find out what it is missing and whether people actually use it. Small changes are included in support. This is where a plan gets corrected cheaply.
With real evidence rather than a projection. Sometimes the answer is that the other six were fine and the money is better spent elsewhere.
Not features for the sake of a list. These are the specific things a spreadsheet cannot do that make each record hold up.
I build replacements for a living, so treat this section as the one worth reading twice.
A dozen non-conformances a year and thirty gauges, handled by one person who is not overloaded, is a spreadsheet job. Software would be a cost with no matching return and the blueprint will say so.
Most of the failures above come from several people touching the same file. If only one person ever opens it, the version and concurrency problems largely do not arise and what is left is a backup question rather than a software one.
Capability studies, gauge R&R, cost modelling, anything genuinely analytical. Excel is better at that than anything I would build you and I would keep it. What moves is the record, not the maths.
If you are still arguing about who signs off a concession, build the agreement first and the software second. Encoding an unresolved process into a system makes the argument permanent and expensive.
A calibration register with eighty gauges and one user is not a document control system with revision approval, controlled distribution and three sign-off levels. Quoting a single price for both would mean either promising a number I cannot build to, or charging the simple jobs to cover the hard ones.
So the blueprint comes first and it is free. You get the real figure in writing before you commit anything and it is fixed once you accept it. It does not move because the work turned out harder than I expected. Quotes hold for 30 days.
Your original files are never touched. I work from copies. The workbooks stay in their folder doing what they do now, right up until you decide to switch. If you read the blueprint and decide against it, you have lost nothing.
One caveat, stated plainly. The data comes across as it is. If six years of typing has left three spellings of the same supplier name, those three spellings arrive with it. I will tell you what I spot, but I cannot warrant historic data I did not enter.
What you own at the end. The application, the code and the data, with no licence and no per-user fee. A full export is free on request whenever you ask. Optional support is £750 a year or £75 a month, with the first 30 days included.
Nothing, until it becomes the record rather than the working out. Excel is excellent at calculating and poor at proving.
It cannot reliably tell you which copy is current, who changed a value and when, whether a closed record was edited afterwards, or that the person reading a procedure is looking at the approved revision. Those are record-keeping jobs and it was not built for them.
No and I would advise against it. One workbook, the one causing the most trouble, rebuilt properly and running for a couple of months before anything else is touched.
Replacing seven at once means a long build, a big bill and a switchover nobody has time for. It also means that if one part is wrong, everything waits while it is fixed.
Usually whichever has the most people touching it and the shortest deadline attached. In practice that is normally the non-conformance register or the calibration schedule, because both have consequences when they slip and both involve several people rather than one.
If you are unsure, the blueprint will tell you which one I would start with and why.
No. ISO 9001:2015 names no tool at all. Clause 7.5.3 requires that documented information is available where needed, adequately protected against loss of integrity and improper use and controlled for distribution, access, storage, version and retention.
A spreadsheet system that genuinely meets those requirements is compliant. The difficulty is meeting them consistently, year after year, as the number of records and the number of people grows.
It comes across. The sheets are exported, checked and imported, so on day one your history is already there rather than you starting from an empty screen.
Your original files are never touched. I work from copies and the originals stay in their folder until you decide you no longer need them.
Yes. Every screen exports to a spreadsheet, so your auditor, your customer or the person who simply prefers Excel can have the data in a file they can open without asking me for anything.
That matters more than it sounds. It means your data never stops being yours in a format you control, which is also what makes leaving straightforward if you ever want to.
No to both. It is hosted and runs in a browser on a PC, a tablet or a phone. If your staff can open a website, they can use it and there is nothing to install on anybody's machine.
From £1,950 per part depending on the size of the job, with no VAT to add and no per-user fees. Optional ongoing support is £750 a year or £75 a month, with the first 30 days included.
Before you commit anything you get a written blueprint, free. It sometimes says the job is not worth doing and I would rather tell you that than sell you something you did not need.
From two weeks per part once the quote is accepted. You get a link to try part-way through rather than a surprise at the end.
Your existing workbooks carry on exactly as they are throughout, so there is no period where you are between systems and no weekend when everything has to move at once.
I do. My trade is quality inspection, twenty years of it, most recently on aerospace engine components. There is no team, no account manager and nobody in a call centre.
You email me, I answer, I build it. Which is also why I take on a few of these at a time and will tell you early if yours is not a good fit.
One file, or a screenshot if it is too sensitive to email, plus a line or two on who uses it and what goes wrong. Within two working days you get a written breakdown: what it would become, how long it would take and exactly what it would cost.
Free whether or not you hire me. No sales call unless you ask for one.