Most data quality advice is about cleaning: validation rules, deduplication, reconciliation. All useful, all work.
There is a cheaper move that almost nobody makes first, and it is subtraction.
The mechanism
Somewhere in your business is a form, a system, or a spreadsheet with a field that nobody uses. It was added because someone once thought it might be useful. Nothing downstream reads it. No report includes it. No decision has ever turned on it.
The person filling it in knows this. Not consciously, perhaps, but they have noticed that nothing ever comes back about that field, and so it gets whatever gets it past the validation — a dash, a repeated value, the first option in the dropdown.
That is the obvious cost, and it is not the important one.
The important cost is that carelessness does not stay local. A form where four fields matter and six do not is a form that gets filled in carelessly, in general. The habit is formed by the majority of fields, and the habit applies to all of them. You do not get careful data entry in the fields that matter and sloppy entry in the ones that do not. You get one behaviour.
This is why a shorter form usually produces better data in the fields it keeps — not just less data.
The audit
Take any form or record type your business uses. For each field, ask one question:
When was the last time a decision, a report, or an invoice used this field?
Not "could it be useful". Not "we might want it later". When was it last used.
Three answers are possible:
- It is used regularly. Keep it, and make it easy to fill in correctly.
- It is used rarely but genuinely. Keep it, but consider whether it belongs on the main form or somewhere the few people who need it can add it later.
- It has never been used, or nobody can remember. Delete it.
The third category is normally larger than anyone expects — often a third of the fields on a mature form.
The objection, answered
"But we might need it later."
Two responses. First: you can add a field later, and adding it later means you add it when you know what it is for, which means you can define it properly. Adding a field with no defined purpose is how you got here.
Second, and more important: the data you are collecting now, badly, will not be usable later anyway. Three years of a field filled in inconsistently by people who knew it did not matter is not an asset. Analysing it will produce a confident, wrong answer, which is worse than having nothing and knowing it.
Collecting a field you do not use does not bank it for the future. It banks noise.
What this is really about
Every field you collect is a small ongoing tax: on the person filling it in, on the person who has to clean it later, and on the credibility of everything around it.
Deleting an unused field is the only data quality improvement that costs nothing, takes ten minutes, requires no tool, and improves the fields you kept.
It is worth doing before anything more sophisticated, and it is worth doing again every couple of years — because fields accumulate, and nobody is ever assigned to remove them.