www.simplify-data.de
A new dashboard is presented, and in the very first meeting someone says: "That number is wrong." Soon afterwards, people go back to their familiar Excel lists. The cause rarely lies in the BI tool; it almost always lies in the data behind it. This article explains which quality issues are typical and how to tackle them systematically.
Why data quality decides adoption
A dashboard makes visible data that used to be scattered across systems and spreadsheets. It also exposes errors that nobody noticed before because they were quietly corrected in manual reporting.
Trust is lost quickly. Once users spot one wrong figure, they start doubting all the others. That is why data quality should not be an afterthought but part of every dashboard project.
The typical dimensions of data quality
Data quality problems are usually grouped into a few dimensions. For mid-sized companies, these are the most relevant in practice:
| Dimension | Question | Typical example |
|---|---|---|
| Completeness | Are all required values present? | orders without a sales rep or cost centre |
| Uniqueness | Does each object exist only once? | the same customer with three customer numbers |
| Accuracy | Do values reflect reality? | wrong postcode, outdated contact person |
| Consistency | Do systems agree? | product price in the ERP differs from the web shop |
| Timeliness | Is the data recent enough? | stock levels are only transferred once a week |
| Uniform format | Are formats the same? | "Ltd", "Ltd." and "Limited" in customer names |
Common causes
Master data without an owner
Master data is a company's long-lived core data: customers, suppliers, products, employees. It is often created by many people and maintained by no one. Without clear ownership, duplicates and outdated records accumulate.
Free-text fields instead of picklists
Where staff type industry, region or complaint reason freely, countless spelling variants appear. Such fields are hardly usable for analysis by industry or region until they are cleaned up.
Several systems without shared keys
Web shop, CRM, ERP and accounting often use their own numbering. Without a shared key, records can only be matched by name or email address, which is error-prone.
Unclear definitions
Sometimes the data is correct but the terms are ambiguous. Does a cancelled order count towards order intake? Is an "active customer" someone who ordered in the last twelve months or the last 24? Different answers produce different numbers, even though everything is technically correct.
Manual corrections in reporting
If errors are fixed in Excel every month after the fact, the problem remains in the source. When you switch to an automated dashboard, those corrections suddenly disappear.
Making data quality measurable
What is not measured rarely improves. You can make data quality visible with simple validation rules and track it in a small data quality dashboard of its own.
- Share of empty mandatory fields: for example customers without an industry or products without a product group.
- Number of potential duplicates: for example customers with the same VAT ID or address.
- Differences between systems: total invoices in the ERP compared with accounting.
- Data age: time of the last successful refresh per data source.
These metrics show not only the current state but also whether improvement measures are working.
Where to fix errors
As a rule, fix errors where they originate. Cleaning data in the BI tool can make sense as a temporary measure, but it hides the real problem.
- In the source: merge duplicates, fill in missing values, introduce mandatory fields and picklists in the ERP or CRM.
- In the process: clarify who may create master data and what information is required.
- In data preparation: consistent formats, mapping tables (for example old and new product numbers) and documented rules for special cases.
- In the dashboard: notes on known limitations, such as "Web shop data is refreshed daily at 6 a.m."
Assigning responsibility
Data quality is not purely an IT task. IT can automate checks and adjust systems, but only the business departments know which of two customer records is the right one.
A proven approach is to name a business owner for each important data domain, for example inside sales for customers and purchasing for products. This person decides on definitions, reviews the quality metrics regularly and triggers corrections.
What BI tools can contribute
Modern BI tools such as Power BI or Metabase do not solve data quality problems automatically, but they help make them visible. In Power BI, Power Query lets you standardise formats, flag empty values and apply mapping tables while loading data. Data profiling in the Power Query editor also shows at a glance how many empty or erroneous values a column contains.
In Metabase, you can write validation queries in SQL and save their results as questions of their own, such as "customers without an industry". Alerts then send the owners a regular overview.
What matters is that such rules are documented. A clean-up step that only one person knows about is just as risky as an undocumented Excel macro.
Checklist for your next dashboard project
- Before the project starts, list the key data sources and their master data.
- Take samples: count empty fields, duplicates and spelling variants.
- Define metrics unambiguously and document them.
- Agree on shared keys between systems or build a mapping table.
- Name a business owner for each data domain.
- Automate validation rules and set up a small data quality dashboard.
- Flag known limitations transparently in the dashboard.
- Review data quality together at regular intervals, for example quarterly.
Conclusion
A dashboard is only as good as the data it is built on. Checking data quality early, clarifying ownership and fixing errors at the source lays the groundwork for figures to be accepted and actually used for decisions. The effort is usually smaller than the time otherwise spent arguing about diverging numbers.
How Simplify Data can help
Simplify Data helps companies in Schleswig-Holstein, Hamburg and across Germany analyse data sources, build validation rules and implement reliable dashboards with Power BI or Metabase. Find out more on our Business Analytics page.