Usable, not perfect: why waiting for clean data means you never start

Waiting for perfect data feels responsible. In practice it is the most reliable way to still be in the same place six months from now.

Your data does not need to be perfect before you start on analytics or AI. It needs to be usable for the first thing you want to do with it. Those are very different standards, and confusing them is one of the most common reasons data projects never get going.

Almost every business describes its data as a mess. That is normal. Data is created by people and systems doing their day jobs, not by people preparing it for analysis. The question is not whether it is messy. The question is whether that stops you starting.

Why waiting for perfect data does not work

Waiting for clean data feels like the responsible choice. It sounds like discipline. In practice it tends to go the same way.

A data clean-up with no use case behind it has no way to prioritise. Every field, every table and every system looks equally important, so everything gets looked at and nothing gets finished. Budgets are spent on data that nobody ends up using, while the gaps that actually matter go unnoticed because nothing has been built that would expose them.

Six months later the data is a little tidier and the business has nothing new to show for it. The case for the next phase is harder to make, not easier.

Build, find the gaps, fix them, build better

The approach that works is a loop, not a sequence:

• Build something useful with the data you already have.
• Find the gaps. Building is what shows you where the data falls short.
• Fix the gaps that made a difference to the result.
• Build better, on data that is now improved in the places that matter.

Two things happen as soon as you start. You begin creating value with data you already hold. And you discover exactly where the problems are, rather than guessing. Both are impossible while you are still waiting.

What usable actually means

Usable is not a lower bar. It is a more specific one. Data is usable when it is good enough to support a particular decision, with its limitations understood.

A sales dataset with a few missing regions might be perfectly usable for a trend report and not usable at all for commission calculations. The same data, two different answers. That is why the standard has to be set by the use, not by the data in isolation.

It also means being honest about the gaps. A report built on usable data should say what it does not yet cover. Trust in the number is what makes data get used, and trust depends on people knowing where the edges are.

This is not an argument against foundations

Solid data foundations matter, especially for AI. An AI assistant pointed at a messy estate will answer with complete confidence and a number nobody agrees on. We wrote about that in getting your data ready for AI.

The point is that foundations are built best when they are pulled by real work. Build the first use case, and it tells you which foundations to lay first. It is the principle behind our AI-ready data foundations work.

Where to start

Pick one decision the business makes regularly that better data would improve. Work out what data would help, what you can get hold of, and whether it is good enough to start. Then build, and let what you learn set the priorities for what to fix.

If you want a quick read on where your data stands today, our AI data readiness checker takes a few minutes. For a fuller picture, a free analytics audit will show you what you can build with now and what needs work first.

Sometimes you have to start in order to know what to fix.