HT
← All projects

Working prototype — In development

EcomOS AI

E-commerce decision support: scoring, explainable recommendations, PDF reports.

Period2025 — présent
My roleDesigner & developer (SaaS prototype)
TechnologiesPython · Pandas · Streamlit · ReportLab · YAML · Jupyter
01

The starting point

Help a merchant turn sales exports into a structured view of product performance. The prototype combines metrics, decision rules and a readable report.

Bilingual Python/Streamlit prototype: CSV import, data preparation, business metrics, YAML rule-based scoring and PDF reports. Recommendations use business rules; direct e-commerce platform integrations remain a planned improvement.

EcomOS AI
Project visual · open to view full size
02

How it works

Simplified functional workflow
  1. 1CSV import
  2. 2Mapping & cleaning
  3. 3Metric calculation
  4. 4Scoring & YAML rules
  5. 5Dashboard & PDF
Architecture & implementation
  • Pipeline: CSV import → schema validation → cleaning → KPI calculation (ROAS, margin, CTR) → scoring → recommendations → PDF report.
  • Weighted multi-criteria scoring engine (margin, volume, ad profitability, risk).
  • YAML-based decision rule engine: nested AND/OR conditions, priority levels (SURVIVAL, GROWTH).
  • Automatic generation of contextualized recommendations and PDF reports (ReportLab).
  • Streamlit interactive UI for data exploration and decision visualization.
Technical decisions
  • YAML rule engine for understandable and configurable recommendations.
  • Fast prototype to validate business logic before industrialization.
  • ReportLab for professional PDF report generation without external dependency.
Lessons learned
  • Data quality directly impacts recommendation relevance.
  • Explainability is essential for adoption in decision-support tools.
  • A fast prototype helps validate complex business logic.
Possible next steps
  • Connect to Shopify or other real data sources.
  • Improve report quality and personalization.
  • More intelligent and automated layer (ML).
03

Results & practical value

  • Working prototype: import, analysis, recommendations and PDF export in one Streamlit interface.
  • Rules make recommendations explainable: each suggested action follows business conditions.
  • Interest in AI guides future work: compare a learned model with this explainable baseline using validated data and objectives.

Delivered so far

  • Working prototype with CSV import and column mapping.
  • Metrics, scoring, YAML rule-based recommendations and PDF reports.
  • Local analysis history; further user validation planned.
04

Scope & limitations

  • Current recommendations use rules, not a trained machine learning model. Direct Shopify/WooCommerce integrations remain future work.
  • Margin estimates depend on available costs. Missing costs should not be interpreted as zero, nor margins as validated net profit.
  • Merchant validation and measurement of commercial gains remain to be completed.

Source: supplied Python/Streamlit prototype. Demonstration image; displayed figures do not represent commercial gains attributable to the product.