How Adaptavist Slashed Data Downtime and Brought Observability Culture to Life

A few days ago, we went live with a new webinar — a casual, honest, and refreshingly real conversation between our Head of Content Lindsay and Callum O’Connor, Senior Analytics Engineer at Adaptavist (and proud owner of a very healthy dose of British wit).

No buzzwords. No slides full of acronyms. Just two people digging into what it really takes to scale observability across a modern data team — and how Callum’s crew went from firefighting to forward-thinking.

This post is your debrief: the highlights, the takeaways, and the stories that made us laugh (and also made us rethink how we manage data quality).

If you’re short on time, this is your shortcut. If you’ve got a coffee and 45 minutes, the full webinar’s right here.

‍

☕ Friday Energy and a Bit of Chaos

From the first few minutes, it was clear this wasn’t going to be your average webinar.

"If you try and talk to me or get me to do anything important before noon… good luck." – Callum

The tone was casual, honest, and unmistakably human. We were in good hands.

👀 Observability = Seeing and Understanding

Callum broke down observability in two parts:

1. Seeing your data: where it lives, how it flows, and what it connects to.
2. Understanding what that data means and why it matters.

It’s not just about monitoring what you know might go wrong — it’s about surfacing what you didn’t even think to look for.

🚀 From Skeptic to Superuser

Before Sifflet, Adaptavist was juggling a chaotic patchwork of tools — DBT logs, Confluence docs, QuickSight dashboards. Sound familiar?

Callum was skeptical that Sifflet’s cross-repo lineage would work.
Two days later, he had a working POC.
One week later, the whole setup was live.

“We’re now debugging in minutes, not days.”

🛠️ Refactoring Without Downtime

Massive refactor project. Zero downtime allowed.

Callum’s team needed to retire a whole repo of DBT assets in favor of Terraform-based pipelines — without breaking downstream models or reports.

With Sifflet:

• They mapped the blast radius of each change
• Alerted impacted users before things broke
• Validated results pre-production

The old process? “Merge and pray.”
The new one? Observability-first, risk-aware, rollback-ready.

"The Secret to Mornings Without Alerts" “With Sifflet, you know before 9 a.m. if things are fine — because you’ve already tested everything.”

Callum O'Connor

❤️ Observability Culture is the Endgame

Tooling alone doesn’t cut it. That was one of Callum’s biggest points.

“Observability won’t fix everything if you don’t have the culture.”

So Adaptavist is investing in:

• Clear ownership of alerts
• Health scoring across assets
• Native dashboard feedback for stakeholders
• And even “observability on observability” (yes, really)

One of our favorite ideas? Baking observability into sprints, not treating it as a side project. When building or refactoring, teams now surface lineage, write monitors, and document assumptions by default.

🧪 Platform Health, Reimagined

Ever wonder why something’s failing… again?

Callum walked us through their next evolution: a health score for every asset, based on:

• Freshness (data and pipeline timing)
• Quality (validation tests and row anomalies)
• Reliability (how often something breaks)

It’s observability that tells a story — not just noise, but actionable diagnostics. And it’s made possible by Sifflet’s ability to import/export data across the stack and power custom views.

📈 The Business Impact

What does this all translate to?

• Less downtime
• Faster debugging
• Fewer surprises for stakeholders
• More confidence in the data — across the org

“No news is good news. If business users aren’t complaining, we’ve done our job well.”

🛤️ What’s Next for Adaptavist?

• Better analytics on incidents
• Smarter routing of alerts
• Prioritization based on DAG importance
• Observability sprints as a team norm

Callum put it best:

“We’re trying to build a culture where if you’re alerted — you own it.”

And frankly, we love that.

🎥 Missed the webinar? Watch the full replay here

‍