AI-Powered Web System for Small Business Financial Forecasting and Advice
Authors/Creators
Description
Small and medium-sized companies often depend on spreadsheets for cash flow planning, which introduces manual effort and reduces accuracy, speed, and transparency. This research evaluates the usability of a web-based prototype that processes CSV ledgers to produce next-month financial forecasts along with concise, auditable guidance designed for non-specialists. A three-month moving-average spreadsheet baseline was used for comparison through a five-session, task-based pilot. Metrics included forecasting accuracy (MAPE and MAD), end-to-end workflow time, perceived usability (System Usability Scale), and output clarity. All datasets, instruments, and baseline rules were archived for replication.
Results showed significant improvement across all measures. Mean MAPE decreased from 22.7% (spreadsheet) to 9.4%(prototype). Average task time reduced from 416 s to 43 s, an 89.7% improvement. The prototype achieved a System Usability Scale mean of 82.0, exceeding the benchmark of 68, and a clarity score of 4.44 / 5. Within the limits of a mini-pilot using realistic synthetic ledgers, a transparent next-month forecasting pipeline can enhance SME decision readiness while remaining easy to adopt. Future work will focus on real production ledgers, stronger baselines, version-locked dependencies, and long-term impact evaluation.
Files
Financial_Forecasting_Paper.pdf
Additional details
Software
- Programming language
- Python , JavaScript , TypeScript , HTML , CSS
- Development Status
- Active