Today’s drop is built for data teams who move fast and don’t have time for mystery failures.

We’re launching Soda Data Observability—an AI-powered system for detecting anomalies across all your datasets.

It sets up in under 5 minutes, delivers results immediately (no model long training time required), and works at the scale you’re already operating at.

It’s 70% more accurate than Prophet-based systems.

It runs on billions of rows.

It doesn’t require a lot of config.

And it’s built to help you start right and shift left → Access Now

Why Metrics Observability?

Data testing is great—if you already know what to look for.

But most issues in production aren’t what you expected.

Observability helps you catch:

• Unexpected schema changes

• Freshness gaps

• Partition dropouts

• Sudden volume shifts

No manual rules. No complex setup. Just results—fast.

Start Right. Shift Left.

With Soda Metrics Observability, you can:

✅ Set up in under 5 minutes

Just connect Soda, scan your datasets, and turn it on. Done.

✅ Get results immediately

No need to wait for model training. No tuning. You’ll see anomalies on your very first scan.

✅ Backfill up to 1 year

If you have historical data, we’ll use it—automatically. You’ll get instant visibility into past anomalies.

✅ Monitor everything, then harden

Turn on observability for all your datasets, spot weak points, then codify what matters into Soda checks and data contracts.

This is how you go from unknown unknowns → tested expectations.

From brittle pipelines → resilient platforms.

Under the Hood

⚡️ Proprietary anomaly detection

70% more accurate than Prophet-based models. Designed and benchmarked specifically for data quality.

📈 Explainable anomaly visualizations

Each anomaly is shown with expected ranges, impact, and trend history, so you know what changed and why it matters.

🔁 Feedback-aware detection

Mark false positives, confirm true issues—our engine learns from your input.

Already Proven at Scale

We’ve already scanned over 1.1 billion rows in 64 seconds on Databricks.

This system isn’t theoretical. It works, today, in production.

Bonus: Win a $1,000+ Keyboard

Sign up this week and you’ll be entered to win a custom mechanical keyboard built for data engineers.

Company emails count double → Request Access

What’s Next

Tomorrow we will release collaborative data contracts. They bridge the gap between business and engineering.

But for today?

Turn it on. Start right. Shift left.

Read our docs on Data Observability to learn more about the new features.