What We Got Right (and Wrong) about PowerBI

Monday, March 30, 2026

by Dr. Jennifer Weber

Sometimes the most effective solution is not adding another layer of complexity, but knowing when to stop.

Most of us in Institutional Research didn’t set out to become data architects. We just needed better answers. Over time, the tools got more powerful, and so did our expectations. What started in spreadsheets evolved into something far more sophisticated. But somewhere along the way, more powerful morphed into more complicated.

I first dipped my toes into Power BI back in 2017. I don’t even remember how I came across it, but I do remember downloading it from Microsoft and just playing. Up to that point, my work lived almost entirely in Excel’s VLOOKUPs, pivot tables, and Analysis ToolPak. System wide analysis was possible but only through a lot of manual effort, with separate flat files for each institution, appended using some MS-DOS Command code.

What hooked me early was Power Query, first in Power BI, and later realizing it lived in Excel, too. That discovery completely changed how I worked. Data still came from flat files, but now I could drop them into a single folder and let Power Query handle the appending and transformations. The process became repeatable, auditable, and far less subject tp manual error. It felt like a turning point. Excel did not disappear, but it was no longer the tool of choice to build and transform data.

As our work matured, so did our infrastructure. We moved toward shared datasets, data models, and SQL endpoints, connecting directly to curated, validated data sources rather than pulling multiple files from our system of record. That shift alone saved countless hours that had previously been spent running and stitching files together.

Over time, our analyses became more complex, but easier to produce. Power Query and Power BI gave us the ability to connect, shape, and visualize data in ways that simply were not possible before. My role shifted away from manually wrangling spreadsheets and toward something closer to problem solving. We were figuring out how to bring the right data together to answer questions we had not been able to answer before.

At the same time, dashboards became increasingly popular across NDUS. Staff and administrators value having current data at their fingertips and being able to interact with that data visually only increased that appeal.

But over the past year, something interesting has happened. In a few specific cases, it has become clear that Power BI, despite all its strengths, is not always the right tool for the job.

Take our legislative funding formula as an example. It is based on completed credits across eleven institutions, with a fair amount of complexity. Different course levels, subject areas, institutional breakdowns, and weighted values layered throughout. We built the first Power BI dashboard for this in 2019, and it is now on its fifth iteration.

In the beginning, it worked well. Even with the complexity, the model and visuals dramatically reduced the workload for the system office financial team, who had previously been managing the process manually.

But each funding cycle brings changes. New categories, revised weights, and small tweaks in how things are calculated or displayed. Over time, those changes have added up.

What used to be a quick adjustment in a spreadsheet has become something much heavier. A simple change to a weight now requires multiple IR staff digging into the Power BI model, tracing dependencies, and carefully modifying measures, all while hoping nothing downstream breaks. With each added layer, performance takes a small hit as well.

At some point, it became clear that we were not using the tool in the best way.

Power BI is incredibly powerful, and Power Query enables transformations we could not have imagined a decade ago. But flexibility, especially the ability to keep modifying logic inside the model, can become a liability when a process is complex and changing.

Here is what I have come to believe: for projects like this, Power BI and Power Query are at their best when they are used to create clean, validated datasets. This is where the value lies, and in many ways, where the role of Institutional Research should end.

The final layer, the ongoing adjustments to weights, assumptions, and scenarios, is often better handled outside the model, in a tool the end users control. In this case, that means spreadsheets used by the financials team.

By separating those responsibilities, we get the best of both worlds. Institutional Research can focus on delivering trusted, well-structured data without the constant churn of model revisions. The financial team gains the flexibility to adjust quickly without needing to rely on IR for every change.

Power BI did not fail us. We just asked it to do more than it should.

It turns out the goal is not to build the most sophisticated model possible. It is to build a process that works, one that is sustainable, responsive, and clear about where responsibility lives.  We are still working on getting there. 

Dr. Jennifer Weber is the Director of Institutional Research and Chief Data Analyst for the North Dakota University system.  Her primary functions are to oversee the department and provide system level enrollment reporting to the State Board of Higher Education.  Jennifer also manages system-wide IR Shared Services, works closely with the State Longitudinal Data System (SLDS) and serves as the state coordinator for federal reporting.   As the NDUS-IR is also contracted through the North Dakota Department of Public Instruction (NDDPI) for data analysis and reporting, the NDUS-IR department is ultimately responsible for the data of all students attending public institutions in the state of North Dakota, pre-kindergarten through graduate school.