Data quality is not a project, it is a habit

You cannot fix data quality once and be done. The businesses that trust their numbers treat quality as something that runs continuously, not a clean-up you finish.

There is a particular disappointment that follows a data-quality project. A business notices its numbers cannot be trusted, commissions a clean-up, spends real money getting the data into good shape, and celebrates. Six months later the data is a mess again and everyone is quietly baffled. The clean-up worked. The problem is that data quality was treated as a project with an end, when it is a habit with none.

Data does not stay clean on its own. Every day, new records arrive, people enter things in slightly the wrong way, a source system changes, a category gets used loosely. Quality degrades continuously, which means it has to be maintained continuously. The one-off clean-up is not wrong, it is just not sufficient. It is mopping the floor while the tap is still running.

What the habit looks like

• Catch problems where they enter, not where they surface. The cheapest place to fix a data problem is at the door, as it arrives. The most expensive place is in the board pack, after a decision has been half-made on it. Good practice checks data as it comes in, flags what looks wrong immediately, and stops the bad record before it flows downstream into ten reports. Fixing at the source once beats fixing ten symptoms forever.
• Measure quality, so you can see it slipping. You cannot manage what you do not watch. The businesses that keep their data trustworthy have a few simple, ongoing checks running all the time: are the totals reconciling, are there gaps that should not be there, are values inside sensible ranges. When quality starts to slip, they see it as a trend, early, not as a crisis when a number is finally, obviously wrong in front of the board.
• Make someone responsible for it continuing. A clean-up has a project manager and an end date. A habit needs an owner with no end date. Without someone whose ongoing job includes the data stays trustworthy, quality maintenance is everybody’s responsibility, which means it is nobody’s, and it quietly lapses the first busy quarter.
• Build the checks into the platform, not the calendar. Quality maintenance that depends on someone remembering to run a monthly review will lose to the month everyone is too busy. The version that lasts is automated: the checks run themselves, every day, and only ask for a human when something needs a human. Effort spent once building the checks buys quality that maintains itself.

The return on this is not glamorous and it is enormous. It is the difference between a business that acts on its numbers without hesitation and one that quietly double-checks everything by hand because it has been burned before. Trust, restored once and then maintained, is what makes the whole analytics estate worth having.

If your data gets cleaned up periodically and then drifts back, the fix is not a bigger clean-up, it is turning quality into something that runs continuously. Talk to us at hello@hoptonanalytics.com.