Data Quality in BI Strategy – How to Build Trust in Your Numbers 2026

“The numbers don’t add up.” It is the kiss of death for any BI initiative. When nobody trusts the data, everyone goes back to Excel and gut feeling. Data quality is not a technical problem – it is a trust problem. Here is how you solve it.

At Excel Department we have implemented BI for 60+ Swedish companies. The most common reason for failed BI is not the wrong tool or poor design. It is poor data quality that undermines trust from day one.

This guide does not present academic theories about data governance. It presents practical steps that Swedish mid-sized companies can actually carry out.

Why data quality determines BI success

You can have the best-looking Power BI dashboard in the world. But if the CEO opens it and says “That’s not right”, the whole investment is worthless.

The cost of poor data quality

A trading company we worked with had order data in three systems: CRM, accounting system and warehouse management. The same order could have different order numbers, different values and different statuses in the three systems.

Result: – Sales reports showed different revenue depending on the source – The management team spent 45 minutes every Monday “reconciling the numbers” – Decisions were delayed by weeks while waiting for “the right number” – The BI dashboard was not used because nobody trusted it

Cost in money: – 3 people’s time, 2 hours/week comparing numbers = SEK 30,000/month – Delayed decisions that cost missed opportunities = unmeasurable but significant

Trust takes time to build and seconds to destroy

When a BI system is launched you have 2-3 months to prove that it works. If numbers are wrong during that period, it is extremely difficult to rebuild trust.

A manufacturing company launched a production dashboard. In week two the production manager found an error – a machine was shown as “idle” even though it was running. The error was fixed in 10 minutes, but the production manager stopped trusting the system. Six months later he was still using his Excel files.

Lesson: Perfect data from day one is impossible. But transparent handling of errors and quick corrections build trust.

The five dimensions of data quality

Data quality is not a binary question (good/bad). It consists of several dimensions that must all be managed.

1. Accuracy

Definition: Does the data reflect reality correctly?

Examples of accuracy problems: – Customer address: “Stokholm” instead of “Stockholm” – Product price: SEK 1,000 instead of SEK 10,000 (a zero was missed) – Date: 2025-13-01 (month 13 does not exist)

How to ensure accuracy: – Validation at entry (dropdown instead of free text where possible) – Automatic checks (price must be between X and Y) – Regular audits against source systems

Tools in Excel/Power BI: – Power Query: IF statements that flag improbable values – Excel: Data validation on cells – Power BI: DAX measures that count the number of “suspicious” records

2. Completeness

Definition: Is all the necessary data present?

Examples of completeness problems: – Customer record without an email address (needed for campaigns) – Order without an assigned salesperson (impossible to calculate commission) – Products without a category (cannot analyse by category)

How to ensure completeness: – Mandatory fields in input systems – Identify critical vs nice-to-have fields – Measure completeness: % of records with complete information

Example from a real customer:

We analysed their customer data (3,200 customers): – 100% had a name and customer number (good!) – 87% had an email (acceptable) – 34% had an industry registered (poor – could not segment)

Action: A campaign to fill in the industry. After 3 months: 78%.

3. Consistency

Definition: Is the same thing represented in the same way everywhere?

Examples of consistency problems: – Company name: “Volvo AB”, “Volvo”, “VOLVO AB”, “AB Volvo” – Product name: “Widget-123”, “Widget 123”, “Widget123” – Date: “2026-01-30”, “30/1/26”, “30 January 2026”

This kills BI analyses: Try to calculate sales per customer when the same customer has four variants of the name. You get four customers instead of one.

How to ensure consistency: – Master data register (one official source for customer names, product names etc.) – Standardised formats (dates always YYYY-MM-DD, amounts always X XXX SEK) – Automatic conversion on import (Power Query can standardise formats)

Power Query example:

= Table.TransformColumns(Källa,{{"Datum", each Date.From(_), type date}})

This forces all dates into a standard format regardless of how they look in the source.

4. Timeliness

Definition: Is the data current enough for the decision to be made?

Examples of timeliness problems: – The dashboard shows last month’s figures (too old for daily management) – Stock levels are updated once a week (too infrequent for e-commerce) – Sales data is updated in real time (unnecessary for a monthly board report)

The right timeliness depends on the use case:

Decision	Timeliness needed
E-commerce stock status	Real time/every 15 min
Daily sales follow-up	Overnight (yesterday’s figures)
Monthly financial report	Once a month
Annual report	Once a year

Common mistake: Trying to do everything in real time. It costs performance and is often unnecessary.

Solution: Define the refresh frequency per dashboard based on actual need.

Read more: Power BI for business – Refresh strategies

5. Validity

Definition: Does the data follow defined rules and formats?

Examples of validity problems: – Email: “contact#company.se” (@ is missing) – Phone: “08123456” (too few digits) – Organisation number: “12345” (wrong length/format) – Price: -SEK 500 (negative price not allowed)

How to ensure validity: – Data validation in source systems (Excel, CRM, ERP) – Regex checks on import – Power Query transformations that flag invalid values

A practical process for improving data quality

Data quality improves through systematic work, not through magic automation.

Phase 1: Data quality audit (Weeks 1-2)

Step 1: Identify critical data sources – Which systems contain data for your BI? – Which tables/fields are most critical?

Step 2: Measure the current state

For each critical data source, measure: – Accuracy: Spot checks against reality – Completeness: % empty fields – Consistency: Number of variants of the same thing – Timeliness: How old is the data? – Validity: % of records that break rules

Example report:

Dimension	Measure	Result	Target
Completeness	% customers with industry	34 %	80 %
Consistency	Number of name variants for the top 100 customers	2.3 per customer	<1.1
Timeliness	Age of order data	18 hours	<12 hours
Validity	% invalid email addresses	8 %	<2%

Step 3: Prioritise

Not all problems are fixed at once. Prioritise based on: – Business impact (does this block important decisions?) – Difficulty (how easy is it to fix?) – Visibility (does this affect trust in BI?)

Phase 2: Quick wins (Weeks 3-4)

Fix a few obvious problems quickly to build momentum.

Examples of quick wins: – Standardise name variants for the top 50 customers (manually in master data) – Implement data validation for email in the CRM – Set up automatic nightly refresh instead of manual

Result: Visible improvements that build trust in the process.

Phase 3: Systematic improvement (Months 2-6)

Implement processes for ongoing quality:

1. Data governance structure – Data owner: Responsible for the quality of specific data (e.g. the sales manager for customer data) – Data steward: Carries out daily quality work – Data governance board: A management team that prioritises data quality work

2. Master data management

Define one official source for: – Customers (customer number, name, contact info) – Products (article number, name, category) – Employees (employee number, name, department) – Organisation structure (departments, cost centres)

3. Data validation at the source

The best place to secure quality is at entry: – The CRM system does not allow invalid email – The accounting system requires a cost centre to be specified – The product register has a dropdown for category instead of free text

4. Automatic quality checks

Build dashboards that flag quality problems: – New duplicates detected – Fields that have become empty – Values outside normal ranges

Example Power BI measure:

Andel kunder utan e-post =
DIVIDE(
    COUNTROWS(FILTER(Kunder, ISBLANK(Kunder[Email]))),
    COUNTROWS(Kunder)
)

Visualise this over time. When the number increases, act.

Phase 4: Continuous monitoring (Ongoing)

Monthly quality report: – Trends for the five dimensions – New problems identified – Completed improvement actions – Next month’s priorities

Quarterly data quality audit: – Spot checks of critical data – Validate that processes are followed – Adjust based on new needs

Tools for data quality

Different tools for different needs and budgets.

Excel and Power Query (Free, included in Office 365)

Areas of use: – Clean and standardise data on import – Find duplicates – Validate formats – Convert between formats

Example: Find duplicates in a customer list

Power Query steps: 1. Import customer data 2. Group by “Customer name”, count rows 3. Filter where count > 1 4. Result: All duplicates listed

Advantage: No extra cost, most people already know Excel.

Disadvantage: Manual process, not scalable for very large data volumes.

Power BI Data Quality Dashboard

Build a dashboard that measures quality: – % completeness per field over time – Number of duplicates detected – Number of invalid values – Age of data

Advantage: Makes quality visible, drives improvement.

SQL Server Data Quality Services (For larger organisations)

If you have SQL Server, DQS can be used for: – Automatic duplicate detection – Data standardisation – Matching between source systems

Advantage: Powerful, automated.

Disadvantage: Requires SQL Server expertise and infrastructure.

Third-party tools

Products such as Talend, Informatica, Ataccama – powerful but expensive and comprehensive.

Recommendation for Swedish SMEs: Start with Excel/Power Query and Power BI. Upgrade when volume requires it.

Next steps – Improve your data quality

How do you get from “we know our data is messy” to actual improvement?

Week 1: Map the current state

Simple checklist:

For your 3 most important data sources, answer: – [ ] Do you trust the numbers from this source? – [ ] Are there duplicates? – [ ] Are fields empty that should be filled in? – [ ] Are the formats consistent? – [ ] How old is the data?

If you answer “no” or “don’t know” to the first question or “yes” to 2-3, you have data quality problems.

Weeks 2-3: Quantify the problem

Actually measure: – Completeness: Open Excel, count empty cells – Consistency: Use Power Query, group and count variations – Timeliness: Check the latest update date

Output: A simple report that quantifies the problem.

Month 2: Implement quick wins

Choose 2-3 improvements you can make straight away: – Clean up the top 50 duplicates – Add data validation in source systems – Automate a manual refresh routine

Output: Measurable improvements that build trust.

Months 3-6: Systematise

• Assign data owners
• Build a quality dashboard
• Establish monthly follow-up

Need help?

Excel Department helps Swedish companies improve data quality as part of their BI strategy.

Our approach: 1. Data quality audit – quantify the problem 2. Master data strategy – define sources and structure 3. Implementation of quality processes 4. Automation with Power Query/Power BI 5. Training for data owners and stewards

Book a free data quality review (1 hour)

Summary

Data quality is the foundation of successful BI. Nobody uses dashboards they do not trust.

The five dimensions: – Accuracy: Is it correct? – Completeness: Is everything there? – Consistency: Is it uniform? – Timeliness: Is it current? – Validity: Does it follow the rules?

A practical way forward: 1. Measure the current state – quantify the problem 2. Quick wins – build momentum 3. Systematise – processes and responsibility 4. Automate – dashboards and checks 5. Monitor – continuous improvement

Perfect data quality is impossible. But good enough to make better decisions than gut feeling is absolutely achievable.

Start by measuring. What you can measure, you can improve.

---

Related guides: – Business intelligence for mid-sized companies – Overall BI strategy – Excel as a BI tool – Power Query for data cleansing – Power BI for business – Data modelling and quality