The operating problem
Search results are not a market census. Overlapping queries, repeated jobs, mixed budget types, and inconsistent labels can make a report look precise while misrepresenting the underlying sample.
What we built
We built a research pipeline that deduplicates marketplace records, preserves their search provenance, and organizes demand into a reviewed taxonomy. Python and notebook workflows support preparation, while an interactive report supports exploration.
Fixed-price budgets are kept separate from hourly ranges. Classification corrections and quality checks are part of the workflow so that a visualization does not hide problems in the source sample.
What the work enables
The result is an inspectable analytical artifact for understanding the collected sample. It is not presented as a complete market size estimate or a guarantee of future demand.
Good market intelligence makes the limits of the sample visible alongside the findings.