Start with the data you have: a practical way to take stock before you build

Before you build, list the data that might help, then ask three questions of each item: can we get it, is it usable, and is it worth it. The answers tell you where to start.

The quickest way to assess the data you already have is to start from a use case, list every source that might help, and ask three questions of each one. Can we get it? Is it usable? How valuable would it be? The answers tell you what to build with now and what to fix later.

It is a simple exercise. It usually takes days, not months. And it replaces a vague sense that "our data is a mess" with a specific list of what you have, what it is good for, and where the gaps are.

Step one: write the long list

Start with the decision or process you want to improve. Then write down every piece of data that might help, without filtering. Sales transactions, customer records, stock levels, web analytics, supplier data, timesheets, spreadsheets that live on someone's desktop.

The long list matters because the useful data is often not where people expect. A spreadsheet one team maintains by hand can turn out to be the only place a critical piece of information exists.

Step two: ask three questions of each item

Go through the list and score each item against three questions.

• Can we get it? Is it in a system we can connect to, an export someone produces, or locked in a third party's platform? How often does it change, and can we get it that often?
• Is it usable? Is it complete enough, consistent enough and accurate enough for this purpose? Are the definitions clear, so that "customer" or "order" means the same thing everywhere it appears?
• How valuable would it be? If we had it, how much would it improve the decision? Some data is nice to have. Some changes the answer.

A simple high, medium or low against each question is enough. The point is not precision. The point is to see the pattern.

Step three: sort it into three groups

The scores usually sort the list into three groups.

• Build with it now. Accessible, usable enough and valuable. This is your starting point.
• Fix it next. Valuable, but hard to get or not yet usable. These are the gaps worth a plan.
• Leave it for now. Low value, whatever its condition. Cleaning it would be effort with no return.

The third group is as important as the first. Knowing what not to fix is what stops a data project turning into an endless clean-up.

Step four: build with what is good enough

Build the first version on the data in the first group. Be clear about what it does not cover yet. Then use what you learn to refine the second group, because building will show you which of those gaps actually hold the result back.

This is the loop we believe in: build, find the gaps, fix them, build better. We set out why in usable, not perfect.

What this looks like in practice

For a sales performance use case, the first group might be invoiced sales from the finance system and the product catalogue. The second might be margin data that needs cost prices tidied, and customer segments that are held inconsistently in the CRM. The third might be web session data that would be interesting but would not change a single sales decision.

You can build a useful sales report on the first group this month. The second group becomes a short, specific improvement plan. The third waits until there is a reason for it.

If you would like a structured version of this exercise, our AI data readiness checker is a good first pass. If you would rather have help, our data strategy and leadership team can run it with you, and a free analytics audit is the easiest way to see what it would find in your business.

Take stock, start with what is good enough, and let the work tell you what to fix.