Reporting, dashboards and analysis for small businesses.

Scott Campbell. I fix monthly reporting for small businesses and agencies, from Excel macros to Power BI dashboards.

Available for freelance data and reporting projects.

The hybrid improves on TF-IDF. The neural model alone does not.

Reported arXiv recommender evaluation. Mean average precision at ten, higher is better.
Model	MAP@10 (higher is better)	Score	95% interval
Hybrid	0.199	0.199	0.186 to 0.213
TF-IDF	0.149	0.149	0.137 to 0.161
MiniLM	0.121	0.121	Not shown

2,000 evaluation seeds. Reported evaluation: 29 June 2026.

Figure 1. arXiv Recommender, mean average precision at ten on held-out seeds from a 28,000-paper OpenAlex snapshot. Hand-set blend. Restricted citation graph. Read the evaluation

Projects

CellwrightFree VBA migration tool. Read all 30 test workbooks on Windows and Ubuntu.
Grid Intelligence5.49% model MAPE vs 8.30% for EIA day-ahead. Rolling snapshot; coverage differs.
Manufacturing Intelligence PlatformRecovered the signals seeded into 35.3 million synthetic records.
National Health Workforce RegistryVolunteer. Facility registry and API for a national health regulator: 8,627 facilities.

See all projects

Have a report that needs fixing?

I use AI coding tools for implementation and check every result myself.

Contact Scott Campbell Consulting