There's a quiet kind of frustration that builds when your data looks clean in the editor but falls apart the moment it hits Excel. That's exactly where this user finds themselves, staring at a "dataformat.error cannot convert to number" with no obvious culprit in sight. Our take is simple: this isn't a dead end, it's a map. The error message is telling you something specific, even if it doesn't feel that way yet.
The key clue here is the word "TFR" appearing in the details. The user assumes it's a data value, but in Power Query, error details often reference column names, steps, or internal identifiers rather than actual content. When you see something like that, your first instinct shouldn't be to search for the string in your data. It should be to look at the transformation steps that run right before the error occurs. The fact that the raw data loads fine on its own tells you the problem isn't in the source. It's in the pipeline you built on top of it.
What this means for you in practical terms is that your debugging strategy needs to shift from hunting for a value to auditing your steps. Start by isolating the exact query step that triggers the conversion. Use the "Remove Errors" or "Replace Errors" options on a duplicate of your query to see if the offending row disappears, then check the row count before and after. That will tell you whether it's a single bad row or a systemic type mismatch. Also, remember that Power Query's error preview often shows "Error" in the cell, but the real detail is hidden in the "Details" section. Clicking into that gives you the column name and the specific value that failed, even if it looks like gibberish at first glance.
The user's instinct to separate raw data from subsequent queries was smart, and it's exactly the right diagnostic move. But the next step is to look at the type conversion step itself. If you're forcing a column to number and one value is text, that's your error. The "TFR" might be a misread of a column header or a step name that got truncated. So, here's the concrete move: duplicate your query, remove the last few steps, and re-add them one by one until the error appears. That's not guesswork; that's how you pin down the exact moment your data stops being trustworthy. Do that, and you won't just fix this error. You'll understand your own workflow better than you did before.