There's a quiet tension at the heart of this Power Query problem, and it's one worth naming: the tool that keeps your data fresh is also the tool that wipes away your team's judgment. The user here isn't asking for a complex fix or a dramatic overhaul. They're asking for something simpler, a way to let the machine refresh while the humans annotate. That's not a niche request. That's the daily reality for anyone who's ever watched a query run and felt that small pang of dread about what just got overwritten.
The core issue isn't really about Power Query or Excel or even Power BI. It's about the unspoken contract between automation and human input. When the query refreshes, it owns the table. New rows appear, order shifts, and suddenly the sprint numbers that people carefully typed into specific rows are floating in the wrong places or gone entirely. The user's instinct to isolate the manual column, to let people add only the sprint number and nothing else, is exactly right. It's a recognition that the tool should serve the workflow, not the other way around. The problem is that the workflow, as it stands, treats manual entry as an afterthought rather than a first-class citizen.
What's striking here is the user's clarity about the constraint: many people, infrequent edits, but those edits matter. This isn't a case of sloppy process or lazy users. It's a genuine design flaw in how we think about refreshed data. The answer isn't to fight the refresh or to demand that people copy values somewhere safe before the next run. It's to separate the concerns entirely, let the query own the source data, and let the manual annotations live in a place the refresh can't touch. That could mean a lookup table, a separate sheet, or a proper relational setup. It could mean Power BI's own data model, where the refresh updates the facts and the sprint column becomes a dimension tied to a stable key.
The practical takeaway is this: stop trying to make a single table do everything. If a column is manually maintained, it shouldn't live inside the refreshed output. Give it a home of its own, keyed to something stable like a title or a date, and let the query merge it back in after the refresh. The user is already halfway there by asking the right question. The next step is to stop accepting data loss as a side effect and start designing for the way people actually work, sparingly, deliberately, and with a need for the system to respect that.