Semantic Layer: the silent saboteur

We’ve been spending more time with our tech partners lately. We’re at their conferences, co-hosting events (come to our data meets magic event), and working closely with the product development teams shaping modern data technologies.

This helps us stay close to the tech to draw out what matters to you and your business.

I don’t need to tell you that there’s no shortage of product announcements and launches right now. But very little of it is grounded in how your data and business teams operate.

In this issue, we focus on technology and how it translates into real-world impact. We cover key tech updates, a practical dbt health check, and a broader take on semantic layers.

Below you'll find:

📖 Our Recommended Read: The Essential Guide to a dbt Health Check

🔥 Hot Topic: The semantic layer stopped being optional

🛠️ Emerging Tech Insights: Databricks One Mobile, Power BI writeback, and Sigma dataset deprecation

🎩 Chicago Event: Where Data Meets Magic: join us post Databricks AI Days

😂 LOL Moment: Single source of lies

Let's get to it!

Tracey Doyle

Chief Marketing Officer, Analytics8

Was this email forwarded to you? Subscribe here >

Our Recommended Read

📖 What's Changed in Your dbt Project Since You Last Looked?

The number of our clients adopting dbt in their data stack continues to grow because it reduces manual effort for data engineers. But that doesn’t mean it runs itself.

When was the last time you stepped back and looked at how your dbt project is really running?

What tends to happen over time:

You fall behind on how dbt has evolved. New features and improvements roll out, but your project keeps running the way it was originally set up
Standards start to slip. Naming, modeling patterns, and development practices become less consistent as more contributors get involved
Work gets repeated without realizing it. Logic shows up in multiple places instead of being reused, making changes harder than they should be
It gets harder to trust and maintain. Documentation lags, onboarding slows down, and it takes more effort to understand what’s happening across the project

The Essential Guide to dbt Health Checks →

Hot Topic

Your Semantic Layer Is No Longer Optional

Gartner made it official at their Data and Analytics Summit last week:

By 2030, universal semantic layers will be treated as critical infrastructure alongside data platforms and cybersecurity. Their guidance to every CDO and CTO walking out of that conference was direct: budget for semantic capabilities is non-negotiable.

The semantic layer conversation has been building for years. What's different now is that the use cases finally caught up to the concept.

The most important role of the semantic layer is NOT "everyone uses the same metric definition" (though that matters).

It's broader: a semantic layer should be your central source of context.

You must confidently be able to answer: What is this data? What does it mean to the business? How should it behave when a human queries it versus when an AI agent acts on it?

The AI part is where the urgency comes from, because AI agents don't ask clarifying questions. They take the context they're given… and run with it 😬

The takeaway here: Modern tools make it easier than ever to access and act on data- but they don’t automatically enforce meaning. If your semantic layer is incomplete, inconsistent, or nonexistent- your agents will confidently produce wrong answers. And they'll do it at scale. Platforms like Coginiti are emerging to address exactly this problem.

Our Data Strategy Practice Lead, Christina Salmi put it this way:

"Organizations don't just need a data and AI strategy anymore. They need a context strategy. How is meaning defined, stored, and shared across every too, model, and agent that touches your data? That question used to be a nice-to-have. it isn't anymore."

Emerging Tech Insights

A few of the tech updates on our radar:

Databricks Brings the Lakehouse to Your Phone

Databricks just released a mobile app that gives business users access to data from their phone. You can ask questions about your data in plain language, get answers as charts or tables, and pull up dashboards — all from iOS or Android.

Most analytics tools still require you to be at a computer to do anything useful. This shifts that. A sales leader can pull revenue trends on the way to a meeting. An ops manager can check pipeline status from the floor. The data doesn't wait for you to be at your desk anymore.

The practical implication: This raises the bar on data quality and governance. If business users can now query data from anywhere, at any time, the cracks in your semantic layer and metric definitions are going to surface faster — and in front of more people.

Power BI Lets Users Update Data Without Leaving the Report

Power BI's new writeback functionality is now generally available (for the Microsoft Fabric ecosystem), meaning business users can update, add, or delete data records directly inside a report.

Analytics and operational systems have always been kept separate for a reason; one is where you look at data, the other is where you change it. This collapses that wall.

The practical implication: Before you get excited about this, make sure your governance house is in order. Write access through a reporting layer without clear ownership, audit trails, and change controls is how you end up with a data quality problem that's very hard to unwind. And for now, this only works within the Microsoft Fabric ecosystem, so if you're not fully there yet, this one can wait.

Sigma is Deprecating Datasets - Here's Your Timeline

Sigma is retiring its dataset feature in favor of data models. Starting June 2, you won't be able to create or edit datasets. By Sept. 15, existing datasets can no longer be used as a data source at all.

You shouldn't wait too long. This isn't a distant roadmap item; the first cutoff is about ten weeks away. If your team has workbooks, dashboards, or reports built on Sigma datasets, those will stop working unless you migrate them to data models before the deadline.

The practical implication: Audit your Sigma environment now. Sigma has a built-in migration tool and an admin view that shows every dataset, its migration status, and how many documents still reference it. Start with the most-referenced datasets first - those are the ones that will cause the most disruption if they break.

LOL Moment

Have a great week!

Tracey

Transform your business with data.