How to Build a Data Quality Culture That Actually Sticks
By Amara Osei, VP of Customer Success at Valido · October 24, 2023 · 8 min read
Why most data quality initiatives fail
You’ve bought the tool. You’ve set up the rules. You’ve even convinced the engineering team to run a nightly validation script. Yet, three months later, the same data quality issues are slipping through the cracks, and the tool is gathering digital dust.
The problem isn't technical. It’s organizational. Most data quality initiatives fail because they treat quality as a feature to be bolted on, rather than a shared responsibility woven into the fabric of the team. When quality is seen as a "burden" for the data team rather than a "win" for the business, it never sticks.
Building a culture of quality requires shifting the mindset from reactive firefighting to proactive ownership. It’s about making data quality visible, rewarding the right behaviors, and embedding checks into the very definition of a successful delivery.
Step 1: Assign data ownership — not everyone owns quality, which means no one does
The most common pitfall is the "everyone" trap. When every stakeholder claims data quality is their responsibility, it effectively becomes no one's. You need to define clear data owners for every critical dataset.
These owners aren't necessarily the people who built the pipeline; they are the business leaders who rely on that data to make decisions. They are the ones who can answer, "If this data is wrong, what is the impact?" When the business owner is accountable, the technical team gets the support they need to fix it.
Step 2: Make quality visible — dashboards, scorecards, and weekly data health reviews
You can't manage what you can't see. If data quality metrics are buried in a Slack channel or an obscure spreadsheet, they will be ignored. You need a single source of truth that is visible to the entire organization.
Implement a weekly "Data Health Review" meeting where the data owners and the technical team review the scorecards together. Discuss the trends, not just the errors. When the health of the data is treated as a standing agenda item, it becomes a priority for everyone.
Step 3: Celebrate catches, not just launches — reward the engineer who finds a problem early
In many organizations, the only time an engineer is praised is when a pipeline goes live. This creates a perverse incentive to ship fast and ignore warnings. You need to flip the script.
Create a recognition program that rewards engineers who catch data quality issues before they reach production. Frame these discoveries as "wins" that saved the business from a potential crisis. When the team sees that finding a bug is celebrated as much as shipping a feature, they will be proactive rather than defensive.
Step 4: Embed quality into the delivery process — definition of done for pipelines
Quality checks should not be optional add-ons; they should be gatekeepers. Update your Definition of Done (DoD) for data pipelines to include passing a set of automated quality checks.
If a pipeline fails a validation rule, it does not go to production. This forces the team to build quality into the code from the start, rather than patching it later. It shifts the cost of quality from the operations team back to the development team, where it belongs.
Step 5: Tooling that reinforces culture, not replaces it
The right tool can be a powerful enabler of culture, but it cannot replace it. Avoid tools that are purely alert-based and create noise. Instead, choose platforms that provide context, explain why something is wrong, and offer a clear path to resolution.
Valido, for example, focuses on providing a calm, honest view of data health. When a tool is easy to use and provides immediate value, it becomes a trusted partner in the workflow rather than a nagging alarm system.
Real-world transformation: NovaStream
NovaStream, a mid-sized FinTech startup, was struggling with inconsistent customer onboarding data. Their CDO, Sarah Jenkins, realized that their data quality tool was ignored because it was too complex and created too many false alarms.
Sarah implemented a new culture framework: she assigned data owners for every customer table, instituted weekly health reviews, and updated their DoD to require passing quality checks. She also switched to Valido for its AI-driven root-cause analysis.
“Within 30 days, our data quality score went from 72 to 94. But more importantly, our engineers stopped dreading the nightly checks and started using them to improve their models,” says Jenkins. “We went from reactive to proactive in a quarter.”
Start building your culture today
Don't let data quality be an afterthought. Download our starter kit to get your team aligned on the fundamentals.
Get the checklist, templates, and meeting agenda.
Ready to trust your data again?
Join hundreds of data teams who've made data quality a feature, not a fire drill. Set up takes under 10 minutes — no engineering sprint required.