The pipeline failed Thursday. Thanks to Sentinel, the board never knew

A short story about a pipeline, a CFO and a Monday morning nobody wants.

It's 8.52am on a Monday. You arrive at the board meeting prepared. The board pack went out last Friday, and the meeting starts in eight minutes.

You've barely swallowed your coffee when, halfway down page three, your CFO stops. "Why doesn't this revenue figure match the one finance sent last week?" A little coffee dribbles down your chin and onto your crisp, freshly dry-cleaned white shirt.

An awkward pause hangs in the air until somebody says, "Let me check that and come back to you." Everyone knows what that means.

By lunchtime, the data team has the answer. A pipeline failed late on Thursday night. Nothing crashed and no alarm went off. The data simply stopped arriving, and the report carried on as if nothing had happened.

Nobody made a mistake. Nobody was watching the pipeline, that's all. And you're far from alone.

Sentinel changes that. It watches your Snowflake platform continuously and tells the right person when something needs attention, before it reaches a report, a decision or a board meeting. It's built natively inside Snowflake, so there's no data movement and nothing new to run.

What that means for you, in 30 seconds

1. You hear about problems first. Your team finds out in minutes, not from a stakeholder days later.
2. Fewer surprises reach leadership. Failed loads and bad data are caught upstream, before they reach a decision.
3. Root cause in minutes. Full lineage shows exactly where a number came from and everything it affects.
4. Mornings back for your data team. Targeted alerts replace hours of manual checking.
5. A trust score on every data product. Leaders can see at a glance which dashboards, models and analyses are safe to rely on today.
6. Nothing leaves Snowflake. No extracts, no extra infrastructure and no gaps in your governance.

How Sentinel works, in plain English

Sentinel watches two layers of your platform, so nothing slips through the gap between them.

When something needs attention, Sentinel tells the right person in Teams, Slack or email, with the lineage to trace it straight back to its source.

Under the bonnet: for the practitioners

If you're the one who'll be asked "but how does it actually work?", here's what matters:

• It's a 100% Snowflake Native App. It installs from Snowflake Marketplace and runs inside your own account, under your existing role-based access control. It has also completed a Snowflake security review.
• It's built on your dbt project. Sentinel reads your dbt artifacts (manifest.json and run_results.json) directly from a Snowflake stage. There are no agents to deploy and no separate observability stack to maintain.
• It monitors five pillars. Freshness, Schema, Volume, Quality and Product Health, covering both the Floor and the Ceiling in one view.
• Lineage is built for you. Interactive lineage graphs are generated automatically from your artifacts. You can trace test failures and skipped models from raw sources through to downstream dashboards, with critical paths highlighted.
• Alerts go to the people who own the problem. Granular routing rules send specific dbt test and build failures to the responsible domain team via Slack, Microsoft Teams, PagerDuty or email, which cuts alert fatigue.
• Every data product gets a Trust Score. The Data Product Catalog keeps a governed inventory of your dashboards, ML models and analyses, each with an automated Trust Score derived from the health of everything upstream.
• Governance takes a few clicks. The Tagging Wizard classifies models by domain, medallion level (bronze, silver, gold) and test severity tier, so ownership is clear and alerts reach the right team.
• Leaders get an executive cockpit. A single operational inbox and a trust trend chart show current data reliability at a glance.
• You stay in control of telemetry. Sentinel shares some app events with Lumilinks, and Snowflake lets you review and audit those events and logs before they're shared.

What it looks like in practice

For data and technology leaders

The pain: You've invested heavily in a modern data platform, and it only takes one wrong number in the wrong room for people to doubt all of it.

With Sentinel: Issues are caught and flagged early, and you get one clear view of platform health instead of a report after the fact. When something does go wrong, you hear it from your own team, with a fix already under way.

The numbers: Gartner puts the average cost of poor data quality at $12.9 million a year per organisation.

Ask yourself: When did a stakeholder last find a data problem before your team did?

For data engineers and platform teams

The pain: Every morning starts with checking pipelines and dashboards by hand, just in case. It's tiring, it's invisible and it never really ends.

With Sentinel: Continuous monitoring of freshness, schema, volume and quality across your dbt project, with alerts only when something actually needs you, and lineage that takes you straight to the failing model.

The numbers: Monte Carlo's 2022 survey found data professionals spend 40% of their time checking data quality. The average organisation deals with around 61 data incidents a month, each taking about 13 hours to identify and resolve.

Ask yourself: What would your team build with two days a week back?

For analytics and BI teams

The pain: A dashboard looks fine on the surface, but it's built on data that stopped refreshing two days ago, and you're the one who gets the awkward email.

With Sentinel: Freshness and quality checks sit underneath every report, and lineage shows exactly which dashboards are affected when something upstream breaks.

The numbers: In a 2023 survey, more than half of data professionals said 25% or more of their revenue was exposed to data quality issues, with the average at 31%.

Ask yourself: How confident are you that every dashboard opened this morning is showing today's data?

For data product owners and governance teams

The pain: You have dozens of dashboards, models and analyses, and no easy way to say which ones the business can trust today. Ownership is fuzzy, so alerts land in a shared inbox and nobody picks them up.

With Sentinel: Every data product carries an automated Trust Score based on the health of everything upstream. The Tagging Wizard assigns domains, medallion levels and severity tiers in a few clicks, so every model has an owner and every alert has a home.

The research: Gartner warns that organisations with basic or manual metadata management will struggle to make their data AI-ready.

Ask yourself: Which of your dashboards could you put a trust score on today?

For AI and innovation leads

The pain: A model will act on bad data with complete confidence. It won't stop to ask whether a number looks odd.

With Sentinel: A live view of whether the data feeding your models is complete, fresh and fit for purpose, so you can scale AI on foundations you trust.

The numbers: Gartner predicts that through 2026, organisations will abandon 60% of AI projects that aren't supported by AI-ready data. It names data observability as one of the practices that gets data ready.

Ask yourself: Would you know if the data behind your AI changed overnight?

Why Sentinel

• It's Snowflake-native. It runs entirely inside your environment, so nothing is extracted or copied out.
• It covers both layers. Pipeline health and data quality, the Floor and the Ceiling, in one view.
• It alerts you where you already work. In Teams, Slack, PagerDuty or email, routed to the team that owns the problem.
• Lineage is built in. You can trace any number from report back to source.
• Trust is visible. Every data product has a Trust Score, so leaders know what they can rely on.
• There's nothing new to run. No separate observability infrastructure for anyone to look after.

It's built by Lumilinks, a Snowflake Elite Partner and winner of the 2025 Snowflake Startup Challenge, with a proven track record of getting clients to value up to 5x faster.

How to get started

1. See it. Book a 20-minute walkthrough and we'll show you what Sentinel would pick up on your own dbt project.
2. Install it. Get Sentinel from Snowflake Marketplace and it runs inside your own account.
3. Switch it on. Point it at your dbt artifacts, set your alert routes, and let it watch.

Rewind to Monday morning

It's Thursday at 11.14pm and the pipeline fails. Within minutes, an alert lands in the data team's Slack channel, showing which pipeline failed and which reports depend on it. On Friday morning, an engineer fixes it over their first coffee, and the board pack goes out on time with the right numbers.

On Monday, you arrive at the meeting with two cortados in hand: one for you, and one for the CFO, who reads page three, nods and moves on to page four.

Nobody ever knows how close it came. That's exactly how it should feel.

Sentinel, from Lumilinks. Because practical makes powerful.