Power Query engineering

What you're learning

Query parameters

A named, editable value (a folder path, a year, a threshold) referenced throughout a query — change it once in one place instead of hunting through every step.

Custom functions

A query written once and invoked with different inputs — the same idea as a LAMBDA, built for M instead of formulas, so five near-identical import queries become one function called five times.

Folder ingestion patterns

Point at a folder, not a file — every file inside combines automatically, including ones added after the query was built, as long as they match the expected shape.

Error rows & exception handling

A row that fails a type conversion doesn't have to vanish — route it to a separate "Rejected Rows" query so someone can see and fix what didn't load, instead of a silently short count.

Staging queries

An intermediate query set to "Connection only" — it does real work (cleaning, typing) but never loads to a sheet itself; other queries reference it instead of repeating its steps.

Load destinations

Load to a Table, to the Data Model, or Connection Only — the same imported data serves three completely different jobs depending on where it lands.

Refresh diagnostics

When a refresh is slow or wrong, check row counts before and after, look for a step that reads more than it needs to, and confirm every source file actually matched the expected pattern.

M code readability

Rename each Applied Step to say what it does ("Remove Cancelled Orders," not "Filtered Rows1") — the step list is documentation whether you write it that way or not.