Good enough to build on: how to decide which data gaps are worth fixing

Once you start building, the data gaps show up quickly. The skill is knowing which ones change the answer, fixing those at source, and being honest about the rest.

The data gaps worth fixing are the ones that change a decision. If a gap alters a number someone acts on, fix it, and fix it at source. If it does not, record it and leave it. That single test does more to keep a data programme moving than any quality framework.

When you build with data that is usable rather than perfect, the gaps appear quickly. Missing values, inconsistent codes, duplicate customers, a field that means one thing in one system and something else in another. The question is what to do about them.

Why you cannot fix everything

Every business has more data problems than it has time to fix. Treating them all as equally urgent spreads the effort so thinly that nothing is resolved. It also means the problems that genuinely distort the numbers compete for attention with problems nobody would ever notice.

Prioritising is not cutting corners. It is putting effort where it pays back.

Three questions for every gap

When a gap shows up, ask three things.

• Does it change a decision? A gap that moves a number someone acts on, such as revenue, margin, stock cover or headcount, matters. A gap in a field nobody reports on does not, yet.
• How big is it? Ten records out of a million is different from ten per cent of the book. Measure it before debating it.
• Is it getting worse? A stable historical gap can often be documented and lived with. A gap that grows every day, because a process keeps creating it, needs fixing sooner.

The gaps that score high on all three are your priorities. The rest go on a list with a short note of what they affect.

Fix at source, not in the report

The tempting fix is a workaround in the report or the data model. A mapping table here, a filter there, a calculation that quietly corrects a known problem. Sometimes that is the right short-term answer. It should not be the permanent one.

Workarounds hide the problem from the people creating it. They multiply, because every new report needs the same fix. And they break when the underlying data changes. Where you can, fix the gap where it is created: the form, the process, the system configuration or the integration. That is the fix that stays fixed.

Stay honest while you improve

Building on usable data only works if people trust what they see. That means being open about the gaps that remain.

• Say what the report does not yet cover, in the report itself.
• Show data quality where it matters, such as the share of records with a missing value.
• Tell users when a known gap has been fixed, so they know the number has improved.

This is where governance earns its place. Not as a policy document, but as a small set of habits that make sure improvements stick and ownership is clear. Our lightweight governance framework for the mid-market sets out what that looks like.

How this fits the bigger picture

This is the third step of the loop: build, find the gaps, fix them, build better. It follows taking stock of the data you have, and it is what turns a first version into a platform the business can rely on.

For AI, the same test applies with higher stakes. An AI tool will use whatever data it is given, gaps included, and present the answer with confidence. Fixing the gaps that change decisions is exactly what AI-ready data foundations means in practice. Where the fixes belong in the platform itself, a well-built data platform puts the checks in the pipeline so problems are caught before anyone sees them.

If you want help deciding which of your gaps matter, start with a free analytics audit. For ongoing delivery, our Analytics Acceleration Programme puts governed, phased delivery around the work.

Fix what changes the answer. Be honest about the rest.