A beautiful dashboard built on bad data is not a help: it is a trap. It gives you a sense of control while leading you to the wrong decisions with complete confidence. That is why, before we talk about charts, technology or dashboards, it is worth talking about what holds everything else up: data quality.
“Garbage in, garbage out”: why quality wins
The oldest rule in computing is still the most ignored: garbage in, garbage out. If the data going into your Business Intelligence system is incorrect, incomplete or inconsistent, no analysis, however sophisticated, is going to fix it. All a good visualisation does on top of bad data is make an error more believable.
The cost rarely shows up on an invoice, but it is there: entire meetings spent arguing over which figure is the right one, budgets built on misattributed sales, campaigns aimed at duplicate customers. Data quality is not a technical luxury, it is the foundation on which decisions are made.
The dimensions of quality
Talking about “good data” in the abstract is useless. Quality is measured across concrete dimensions, and it pays to know them so you can tell what you are getting wrong:
- Completeness. Is all the data that should be there actually there? A country field left blank for 30% of your customers breaks any geographic analysis.
- Consistency. Does the same fact say the same thing everywhere? “Spain”, “ES” and “España” for the same country create three groups where there should be one.
- Accuracy. Does the data reflect reality? A mistyped amount or a reversed date is accurate as a format, but false as a fact.
- Uniqueness. Are there duplicates? The same customer registered three times inflates your counts and distorts your metrics.
- Timeliness. Is the data up to date? A correct report built on month-old data can be as useless as a wrong one.
A piece of data can be accurate and yet incomplete, or unique but out of date. Measuring each dimension separately is what lets you prioritise where to act.
Symptoms of bad data
Nobody ever announces that the data is wrong. You notice it through the signs:
- Reports that do not add up. The sales total in the dashboard does not match the one in accounting.
- Different figures for the same thing. Marketing and sales quote different numbers for “active customers” because each defines and filters it in its own way.
- Widespread distrust. The most serious sign: when people stop looking at the dashboard and go back to their spreadsheets “because that one is actually reliable”.
That last symptom is the one that kills a BI project. A dashboard nobody trusts is money wasted, however well it is built.
How to improve it
The good news is that data quality can be worked on systematically. It is not solved in one go, but it can be brought under control:
- Profiling. Before fixing anything, measure. Analyse your tables to find null values, inconsistent formats, duplicates and impossible ranges. You cannot improve what you have not quantified.
- Validation rules. Define what a valid value is and enforce it as close to the source as possible: date formats, allowed ranges, mandatory fields, closed lists of values. The sooner an error is caught, the cheaper it is to correct.
- Ownership. Every data domain needs an owner: someone accountable for making sure customer data, or product data, or sales data, is reliable. Without owners, quality is everyone’s job and therefore nobody’s.
- Monitoring. Quality is not a project with an end date, it is a state that degrades on its own. Track quality indicators continuously and raise a flag when something gets worse, rather than discovering it on the day of the board presentation.
The relationship with data governance
Data quality is the most visible part of something broader: data governance. Governance sets the rules —who defines each metric, who can change what, how the origin of a piece of data is documented— and quality is the result you get when those rules work.
That is why the two go hand in hand. You can run one-off clean-up campaigns, but if you do not change the rules and responsibilities that produced the dirty data, you will simply dirty it again. Sustained quality is the consequence of good governance, not of a heroic clean-up every six months.
At Digital Fox Data we help companies get their data in order —profiling, validation rules, ownership and monitoring— with our Business Intelligence consultancy on the market-leading tools, so that your BI becomes a source of reliable decisions rather than of arguments. If you want your data to stop misleading you, let’s talk.