Kickstarter Analytics Engineering
Campaign snapshots, data grains, and an analytics model with Power BI project files.
A question. A process.
Work you can explore.
Follow the decisions behind the implementation, then inspect the original work.
Chapter 1 of 4 · The question
The question behind the build.
Historical dumps and recurring crawls overlap. A campaign can appear repeatedly with changing states, while regional benchmarks exist at a different grain. Appending everything into a single flat table can inflate results.
Follow the implementation.
- 01
Stage sources independently and normalize encoding, naming, types, and dates.
- 02
Apply documented source hierarchy and recency rules to deduplicate campaign records.
- 03
Separate campaign outcomes, campaign snapshots, and regional benchmarks in a galaxy model.
- 04
Keep the semantic model reviewable in Git using PBIP and TMDL, with organized DAX measures.
See the work. Inspect the details.
The model treats grain as an explicit design decision. Campaign outcomes and time-series snapshots can answer different questions without being mixed into a misleading total.
Choose where to look deeper.
Search the linked source inventory or narrow it to the kind of artifact you want to inspect. Files open at their original source.
8 linked source files
No linked files match this view.