How to make data-driven decisions (without being a perfectionist)

In the fourth episode of The View on Data, hosts Faith McKenna, Erica “Ric” Louie, Paige Berry, and new co-host Jerrie Kenney talk about one of the most persistent myths in data work: that data-driven decisions always require perfect data.

Spoiler alert: they don’t.

From civic engagement indexes built on grocery store data to dashboard decisions driven by intuition, the group explores what it really means to work with “enough” data and how to build confidence in your analysis, even when the dataset is messy, incomplete, or inconsistent.

This episode is for anyone who’s ever agonized over missing values, hesitated to share a chart, or wondered if their work was “good enough” to make the call.

Please reach out at podcast@dbtlabs.com for questions, comments, and guest suggestions.

🎧 Listen & subscribe: Spotify | Apple Podcasts | Amazon Music | YouTube

What does it really mean to be data-driven?

For many folks entering the field, “data-driven” can feel like code for statistically significant, p-value certified, decision-changing insight. But in practice? It’s often more about building directional confidence.

“I think people conflate data-driven with 'I’m going to change the company strategy,’” Ric shared. “But sometimes the data-driven decision is: don’t change anything.”

The hosts discussed how business decisions often rely on imperfect or incomplete data, and how it’s more important to ask good questions upfront than to get bogged down in chasing precision.

What’s “good enough” when it comes to data quality?

Jerrie shared a story about helping a client rethink their entire approach to data quality testing, starting with the “why.”

“They were so focused on thresholds and configs, they forgot who they were doing it for,” she said. “We ended up having to zoom out and ask: what decisions are we trying to support here? And how perfect does the data really need to be?”

Paige echoed that point, describing the diminishing returns of chasing exactness when the goal is clarity, not control.

“Sometimes good enough is better than perfect,” she said. “Especially when what your stakeholders actually need is just confidence to move forward.”

Using incomplete data to drive impact

Throughout the episode, the group shared examples of when “directionally correct” was more than enough:

• Jerrie talked about using social media signals and grocery store data to identify civic engagement hotspots. This helped a health equity team decide where to host in-person town halls.
• Faith described making curriculum decisions based on a small number of user requests in course feedback: “What’s the worst that could happen? People don’t like it, and I try something else.”
• Paige highlighted how institutional knowledge and intuition helped her decide whether a metric spike was internal testing or external usage without spending hours chasing a rabbit hole.
• Ric emphasized the value of pairing data with strategic context: “Not every weird number needs fixing. Ask what decisions are riding on this, and how much we’re willing to invest in fixing it.”

The theme? Use what you have. Pair it with experience. And don’t let perfection block progress.

How to build confidence when your data isn’t perfect

For data folks, especially those who are self-taught or newer in their careers, it can feel risky to present work that isn’t 100% airtight. The team talked about the role of impostor syndrome in driving overwork and perfectionism.

“I used to feel like I had to make an ironclad case for every recommendation,” Faith shared. “But I’ve had to learn to trust my experience and say: this is enough.”

Ric and Jerrie both emphasized the power of up-front conversations with stakeholders to set expectations:

• What is the data for?
• What will the decision be?
• What are the consequences if it’s wrong?

“If it’s for a board report or compliance, sure, go for precision,” Jerrie said. “But if the cost of being wrong is low, then directionally accurate is often all you need.”

Communicating clearly when data is messy

The episode wrapped with a conversation about how to present data clearly, especially when your audience ranges from execs to engineers.

Paige shared her approach: “I try to understand the experience of the person reading the chart. How much time do they have? What are they trying to do with this?”

Ric emphasized using context-first messaging in Slack and reports: leading with the TL;DR, followed by charts and only the most relevant details.

And everyone agreed on the power of simple visuals:

Favorite chart types?

• Ric: Bar chart with line overlay
• Paige: Sankey (for the drama)
• Jerrie: Mini line charts (“spark lines”)
• Faith: Big number with small number (and triangle indicators)

Episode takeaways

This conversation is a practical reminder that in real-world data work, perfection isn’t the goal, progress is.

Here are a few takeaways from the episode:

• Data-driven doesn’t mean statistically significant. It means directionally useful.
• Ask what decision the data supports before obsessing over accuracy.
• Stakeholder context matters more than fancy charts.
• Show your work, but only share the summary.
• Confidence in your experience is part of the job.