The NC tracking spreadsheet lasts for years, and for good reason. It never breaks all at once. It simply stops being useful the day you need it most.

In most quality departments, tracking non-conformities begins with an Excel file. One row per NC, one column per piece of useful information, a few filters. This is not an admission of being behind the times; it is a rational choice: the tool is readily available, everyone knows how to use it, and it adapts to the reality of the shop floor better than many software packages. The problem does not arise when filling in the spreadsheet. It appears years later, when you ask it a question it was never designed to answer.

Why the spreadsheet lasts so long

Excel has three qualities that few quality tools can match. It requires no formal project: you create a column in the morning, and it is in use by the afternoon. It requires no training: the technician, the quality manager, and the buyer all read it the same way. And it requires no one's permission, which matters when an NC arises on a Friday evening and needs to be logged immediately.

This is also why attempts to replace it often fail. A more rigid tool arrives, complete with mandatory fields and workflows. Three months later, a parallel spreadsheet reappears because the team on the ground needed a column the software did not provide. "I don't want to change tools" is not mere resistance to change. It is often the simple observation that the current tool gets the job done.

A file that works—for the person who built it

An NC tracking file reflects the logic of the person who created it. The abbreviations, the order of the columns, the way a cause is noted as "to be confirmed", the colour coding of the rows: all of this is obvious to its author. To everyone else, it is a dialect.

As long as that person is there, the file works, because they know where to look and how to interpret it. The day they change roles, retire, or take responsibility for another site, the file remains. The user manual, however, leaves with them. The rows are still there, but no one really knows anymore what "OK after rework" meant on a 2019 NC.

Search only finds what was written the exact same way

A new NC comes in. The instinct is right: check whether the case has already been addressed. Ctrl+F, "porosity". Zero results. Yet, the very same defect was logged two years earlier under the description "weld seam blowholes" by another inspector.

This is the structural limitation of a spreadsheet: it searches for characters, not meaning. It does not know that "clearance out of tolerance" and "non-compliant dimension" on the same part number might describe the exact same issue. Nor does it read the attached photo, or the PDF report archived in a network folder. The answer existed. It was simply unfindable, and the analysis starts again from scratch.

Multi-site: six files, six versions of the truth

On a single site, a well-maintained spreadsheet may suffice. As soon as there are multiple sites, multiple production lines, or multiple programmes, everyone develops their own. Columns diverge, as do defect categories. One site categorises by part family, another by supplier, and a third by inspection station.

The result: a problem resolved on one site remains invisible to the others. Nobody compares six differently structured files to check whether a root cause has already been identified elsewhere. And when management requests a consolidated view, someone spends several days copying, harmonising, and recounting, all for a report that will be out of date by the following month.

During audits: who modified what, and when

An auditor does not settle for the final row alone. They want to know who qualified the NC, who signed off on the decision, on what basis, and whether the corrective action was verified. A shared spreadsheet is ill-equipped to answer these questions. An overwritten cell leaves no usable audit trail; a deleted row leaves even less.

Teams compensate with dated versions, archive tabs, and cell comments. This holds up until the day two people work on two different copies. The handling of the NC may well have been flawless in practice. The records, however, fail to demonstrate it—and in an audit, what cannot be proven is deemed not to exist.

The difference made by AI that reads your files without asking you to abandon them

The usual response to these limitations is to replace Excel with more structured software, then ask teams to re-enter everything. That is often where the project grinds to a halt. Another approach starts from what already exists: Excel files, PDF reports, photos, and emails remain where they are. They simply become searchable.

An AI grounded in the organisation's quality history reads these sources as they are, with all their abbreviations and inconsistent phrasing. When a new NC arrives, it links cases by meaning rather than exact words, including across different sites. It flags that a comparable defect has already been addressed, shows the basis on which it was handled, and indicates when information is missing rather than fabricating it. The engineer retains the decision. They simply no longer make it without institutional memory.

The day Excel breaks, it is never the file that fails. It is the person who knew where to look.