Laying the Foundation for Business Intelligence on Microsoft Fabric: Stiga RM Project Story

Started in 1994 as a full-cycle forest management company, Stiga RM has grown to become one of the largest privately held forestry companies in Latvia. Over time, the company has expanded its product range to offer roundwood transportation and logging services to other industry incumbents, as well as established itself in manufacturing to produce and export world-class birch plywood.

With the recent implementation of Dynamics Business Central as the new ERP system across the group's companies, Stiga RM initiated the cooperation with Scandic Fusion to establish a data analytics platform and business intelligence initiative to streamline management reporting—moving away from the effort-intensive reporting previously done in Excel.

Although Microsoft Fabric was still a relatively new platform when the project began in the summer of 2024, Scandic Fusion had already successfully implemented it for another client. Based on this experience, we recommended Microsoft Fabric to Stiga RM as a modern, future-proof solution. The platform aligns with Microsoft’s strategic product direction while also enabling a flexible, small-scale start—one that can seamlessly scale as analytics needs grow.

Thus began the pilot project—to establish the Microsoft Fabric platform, create a best-practice data model centered around debtor analytics, and introduce management to modern, best-practice dashboarding in line with one of our core values: “Data is art.”

The Team is the Key!

When assembling project teams, Scandic Fusion always strives to select the most suitable people for the job—something that proved especially valuable in this case. The PM who took up Stiga RM implementation already had delivered a Microsoft Fabric implementation for a large pharmaceutical client. Similarly, since Scandic Fusion had just implemented Dynamics Business Central as our own ERP system, we assigned the same senior analyst—who had led our internal implementation just months prior—to this project.

Projects involving both a new data platform setup and data mart development with dashboarding are most efficiently executed in two parallel tracks: technical and analytical. While one part of the team focused on setting up the best-practice medallion architecture, load monitoring, and CI/CD, the other part worked closely with business stakeholders—gathering requirements, preparing specifications for facts and dimensions, and sketching dashboard designs. Architect and PM ensured architecture consistency and overall synchronization of these two tracks.

As is often the case with new and rapidly evolving technologies, we encountered a few challenges along the way. For example, we noticed that Notebooks in Fabric sometimes took excessively long to start, and that SQL Endpoint synchronization with data in Lakehouse tables was not always immediate after a load appeared to be finished—making data loads occasionally unreliable. To address these issues, we implemented necessary workarounds (e.g., timed retries) and successfully delivered the project to production within 12 weeks.

“Data is art”

When creating dashboards for clients, we pay equal attention to the data-driven user story and visual design, integrating elements from the client’s brand book (color palette, icons, images, logos). This is by no means an easy task—striking the perfect balance where nothing essential is missing and no unnecessary details clutter the experience for key users.

We planned three iterations, and it took all three to arrive at the final result—a dashboard with a cover and four data pages, each providing different levels of detail.

For a more in-depth look at dashboard development, check out our blog: What is the Recipe for a Perfect Dashboard?

Is Microsoft Fabric a Feasible Choice for Small DWH/BI Initiatives?

At the start of the project, Microsoft Fabric was still a very new technology, and its pricing model—based on Capacity Units (CUs)—did not immediately make clear whether the smallest capacities (F2/F4) were viable for small projects in production and could reliably handle regular data loads and transformations, combined with Power BI Semantic Model refreshes and usage.

The best way to monitor platform usage in relation to purchased capacity is the Fabric Capacity Metrics App (learn more). However, note that admin rights for the capacity are required to use it.

Unlike traditional on-premises DWH platforms, where server resources are provisioned to handle peak loads, Microsoft Fabric charges for what is better understood as average capacity. To manage short-term required capacity spikes, Fabric employs a process called “bursting”, allowing temporary exceeding of purchased capacity before throttling mechanisms activate—leading to delays or process rejections.

A key distinction should be made between interactive and background operations (learn more):

• Background operations (such as data loads and transformations) have a 24-hour smoothing window.
• Interactive operations (such as Power BI dashboard interactions) have a much shorter smoothing window of only 5 minutes.

Even though data refreshes are relatively CU-intensive, their scheduled nature—occurring only a few times per day in our case—allows the smoothing process to distribute the load, making F2 capacity sufficient for the project’s needs in terms of background operation requirements.

To assess the feasibility of F2 capacity for interactive operation requirements, we conducted a test where four users simultaneously interacted with the dashboard (built on an Import Mode Semantic Model) for several minutes. This pushed interactive CU usage close to 50% and total CU usage close to 65%. Allowing us to conclude that size and complexity of the Semantic Model, along with the number of concurrent users, are critical factors in determining whether a given capacity is sufficient for demand of interactive operations.

After completing our initial tests, we monitored real user behavior with the new dashboard for a week. Since we did not observe spikes similar to those seen during our test, we concluded that F2 capacity is sufficient (at least for now) also for interactive processes.

^"We are pleased with our collaboration with Scandic Fusions as the pilot project on Microsoft Fabric has significantly improved Debtor control processes and ensures that all involved colleagues operate with the latest and consistent data. We will likely continue cooperation by integrating data from other systems and implementing dashboards for other metrics and processes. _{Raivis Cakuls, CFO at Stiga RM}

‍

‍

‍