tokpipe

The problem

TikTok does not offer an analytics API for creators without an approval process, and TikTok Studio exports are raw CSVs that change format between app versions with no documentation or warning. Any script that parses them breaks within weeks. The goal was a pipeline that survives those changes without manual intervention.

The data

TikTok Studio exports: total and unique views, per-second retention curve, interactions (likes, comments, shares), traffic sources and daily followers. Columns, decimal separator and date format vary between app versions.

Approach

• Automatic format version detection: the parser identifies the CSV schema by heuristic before attempting to parse
• Metric normalisation: real retention rate (not the dashboard's, which uses a different methodology), reach/follower ratio, engagement per view — the metrics that matter, not the ones the app shows
• Clean DataFrame output with correct types, ready for any downstream analysis or dashboard

Tools

python

pandas

Result

Pipeline in use with my own data. Survives TikTok Studio updates without code changes. The difference from opening the CSV manually is that the output is reproducible, has correct types and computes the derived metrics TikTok's dashboard does not show.