A polished, Python-based local web app to compare two systems — System A: NCH/CVM and System B: Shared Systems (SS) — column by column. It shows, for every column, how many claims matched vs mismatched, the % match, and the action to take; surfaces the top mismatched columns via an AI-style Intelligence Tracker; links every field to its NCH→NCH STTM business rule; and even lets you upload your own two systems to run the same analysis.
- Two-system comparison: matched / mismatched claim counts, % match and the recommended action for every column, with a summary → drill-down (stream → table → column).
- 🤖 AI Intelligence Tracker: automatically reports the % of columns that mismatch, how many columns need to change, the claims tied to those actions, and the Top 5 columns by both mismatch percentage and unmatched claim volume.
- 📖 Use Cases: each comparison scenario from the SSDS presentation (perfect match, SS-has-more-info, zeros/blanks/nulls, reconciliation-needed, not-expected-to-match, concept-does-not-apply, SS-not-sending-data, etc.) explained, with live counts and real example columns.
- 🧪 Bring-your-own-data demo: upload any two system extracts (CSV/XLSX), pick a key and the columns to compare, and get the full matched/mismatched, top-mismatch and action analysis — tunable to any two systems. Also supports a pre-computed comparison file and a built-in synthetic sample.
- Cross-dataset connections: joins the comparison results to the business-rule / STTM
mapping on
NCH Target Table+NCH Target Column, measures rule coverage, and flags where the documentedData Match? (Y/N)verdict conflicts with the actual result. - Correlation analysis: correlation matrix and scatter plots relating claim volume, matched/unmatched counts and match rate, with auto-generated insights.
- Professional UI: branded hero header, KPI cards, AI-agent bubbles, use-case cards, custom theming and styled charts.
- Data cleaning: normalizes inconsistent dispositions (e.g.
Good/Needs action) and sub-classifications, and parses claim counts and match percentages into numeric types.
- 📈 Executive Overview – KPIs, classification by stream, disposition mix, match rate by stream, top gaps.
- 🤖 Intelligence Tracker – AI read-out of % columns mismatch, columns to change, claims tied, Top 5 mismatched columns (by % and by volume) and a watch list.
- ⚖️ System A vs System B – per-column matched/mismatched claims, % match and action, with summary → drill-down and CSV export.
- 🎯 Data Integrity – integrity gauge, match-rate distribution, Stream × Classification heatmap, lowest-matching fields.
- 📖 Use Cases – sub-classification distribution plus a guide to every comparison scenario with live counts and example columns.
- 🔗 Connections – rule coverage, rule-vs-actual verdict agreement, Stream→Classification→Disposition Sankey flow, conflict inspector.
- 📊 Correlations – correlation matrix, volume vs. match-rate scatter, match rate by classification.
- 🔍 Explorer – searchable, column-configurable table with CSV export.
- 📋 Business Rules – searchable STTM/business-rule mapping plus additional reference sheets.
- 🚨 Action Items – records flagged Action needed, broken down by stream/classification, with CSV export.
- 🧪 Custom Data (Demo) – upload any two systems (or use the built-in sample) and run the same comparison analysis.
-
Install Dependencies:
pip install -r requirements.txt
-
Run the Application:
streamlit run main.py
-
Access the Website: The application will be available at
http://localhost:8501.
The application uses the Excel files located in the Data folder:
All Streams June 2026.xlsx: Main NCH vs. Shared Systems comparison results.CMS Version NCH to NCH STTM 12_13_2024.xlsx: Business rules and STTM mappings (sheets: Lynnette's Version, NCH STTM to NCH Data & % Pop, Missing STTM Cols with Data).