Karen Hsieh

Data 📊 Empower 💜 Product

[Karen Hsieh @ijac_wei]

‘Let me check row counts on all 47 downstream models' every data practitioner validating dbt changes.
That's $$ in warehouse costs just to find the 3 models that actually changed.

Metadata shows you those 3 models instantly, before you run a single query.

[Karen Hsieh @ijac_wei]

When you have limited time to make changes, the realistic way to balance quality is running ad-hoc queries 🤞. But when you work on new models or join a new company, you start over building your validation from scratch.

Impact Radius solves this problem with metadata alone.

[Karen Hsieh @ijac_wei]

You can run `dbt run` and then `dbt docs generate` to view the lineage as shown in the first screenshot. With the artifacts generated by these two commands, you can actually view column-level lineage and visualize changes and their impact at the column level.

[Karen Hsieh @ijac_wei]

See exactly which dbt models your changes will break in 2 minutes, without any setup.

Last week, we launched 👋 https://t.co/xlgX5Unjio for users to experience Recce more easily by uploading dbt artifacts or using our sample artifacts.

[Karen Hsieh @ijac_wei]

🤔 Data teams keep asking: "What do I actually need to validate to ensure my data is right?"
See real example, real workflow, real time and cost savings. 💰
https://t.co/KGKUUtsdpf

[Karen Hsieh @ijac_wei]

Every data transformation begins with metadata modifications: column additions, type changes, new dependencies, or altered join logic.

Can we use this metadata to find potential issues before running costly queries 💸?

👉Explore metadata diffing https://t.co/0eVWceZlYP

[Karen Hsieh @ijac_wei]

"The PRs created by John are always high quality. I can review them easily."
Our most successful users catch issues before submitting PRs. But getting there is painful 🙈

Most people give up. 😭
More on our fix coming soon 👇

[Karen Hsieh @ijac_wei]

"I love the concept, but I can't get it working with my data."
Reading about "dbt artifacts" and "environment setup" doesn't magically give you infrastructure knowledge.
The most motivated users were getting the most frustrated. 😤
More on what we learned coming soon 👇

[Karen Hsieh @ijac_wei]

Teams like Prefeitura do Rio de Janeiro crushed our setup process.
https://t.co/jJhuwkgPp0
The other 80% of sign-ups? They just want to validate changes before merging PRs and may not have engineering experience.
More on how we're fixing this 👇

[Karen Hsieh @ijac_wei]

After our OSDE East 2025 talk, our booth was swarmed. Everyone wanted to see the lineage diff we'd demoed. People were genuinely 🤩 excited.

Then setup conversations: "I need to ask other team." 🙄

Join our journey to fix this.

[Karen Hsieh @ijac_wei]

A partial breaking change can have no impact on downstream models.

Though breaking change analysis works at the column level, identifying a partial breaking change still isn't enough to get the precise impact radius.

Why 👇
https://t.co/WlDKeb1ZU6

[Karen Hsieh @ijac_wei]

🚨 Breaking Change Analysis Gap Confusing Data Teams
Software Dev: Breaking changes = planned improvements
Analytics Engineering: Breaking changes = unplanned problems

Solution: Column-level precision showing what actually needs validation.
https://t.co/WlDKeb1ZU6

[Karen Hsieh @ijac_wei]

A partial breaking change can have full impact on downstream models.

Though we do breaking change analysis to the column level, so we know it’s a partial breaking change. It’s still not enough to get the precise impact radius.

Why👇
https://t.co/WlDKeb1ZU6

[Karen Hsieh @ijac_wei]

'If we can do breaking change analysis at the column level, could we narrow the impact radius to the column level too?’

Simple. Just combine the two tools, right? Wrong. 😵

🧵Form validate everything downstream → validate only what uses it.
https://t.co/WlDKeb1ZU6

[Karen Hsieh @ijac_wei]

'What if we could see impact at the column level?'

While building breaking change analysis and column-level lineage, we couldn't help but think this question.

This simple thought changed everything and lead us to Impact Radius 👇
https://t.co/WlDKeb1ZU6

[Karen Hsieh @ijac_wei]

We built an amazing data validation tool. Users loved the demo. But 80% didn't finish setting it up. 😭

The setup requires technical knowledge that data engineers handle, not the day-to-day analysts who need validation most.

We're fixing this! Join our journey 👇

[Karen Hsieh @ijac_wei]

Breaking change analysis tells you WHAT changed, but not what to DO.
You know a model has a partial breaking change. Great. But which downstream models need validation? Which columns?

A partial breaking change can have full impact on downstream models.

https://t.co/WlDKeb1ZU6

[Karen Hsieh @ijac_wei]

'Validate everything downstream' is expensive and wasteful.

Impact Radius changes that to 'validate exactly what matters' with column-level precision.

This journey took us months of research, dead ends, and breakthroughs.

🙌 Join us: https://t.co/WlDKeb1ZU6

[Karen Hsieh @ijac_wei]

"What do I actually need to validate to ensure my data is right?"Every data engineer asks before merging a PR.

See how we broke it down into two parts:
1️⃣ WHERE to validate? (Which actually matter?)
2️⃣ HOW to validate? (What checks are enough?)
https://t.co/ntotTvFoGW

[Karen Hsieh @ijac_wei]

From “validate everything” to “validate exactly what matters” took us months of research, dead ends, and breakthroughs.

We finally launched Impact Radius to help you analyze changes and s downstream impacts at the column level.

🎒 See our journey https://t.co/WlDKeb1ZU6