# Data Quality Issues Analysis Based on the evidence gathered from the `callyeah_analytics_2026_07_23` table, I've identified several significant data quality issues: ## 1. **Incorrect Data Types (Critical Issue)** All numeric and date columns are stored as **VARCHAR** instead of appropriate data types: - `Total Calls` (VARCHAR) - should be INTEGER - `AI Handled` (VARCHAR) - should be INTEGER - `Human Handled` (VARCHAR) - should be INTEGER - `Voicemail` (VARCHAR) - should be INTEGER - `AI Resolution %` (VARCHAR) - should be DECIMAL/FLOAT - `Human %` (VARCHAR) - should be DECIMAL/FLOAT This prevents proper numeric operations and sorting. ## 2. **Mixed Data Structure (Critical Issue)** The table contains **multiple distinct data structures mixed together** in a single flat table: From step-14, the 216 total rows break down as: - **30 date rows** (daily trend data: 23-Jun through 22-Jul) - **7 day-of-week rows** (Thu, Fri, Sat, Sun, Mon, Tue, Today) - **10 summary/total rows** (aggregates and totals) - **4 header rows** (section headers like "=== DAILY TREND — LAST 30 DAYS ===") - **8 blank rows** (empty separator rows) - **~150+ call log entries** (individual call records with Call IDs like CY-2847) This violates database normalization principles - the table mixes: - Weekly summaries - Daily time-series data - Call outcomes aggregates - Escalation reason breakdowns - Individual call transaction logs ## 3. **Inconsistent Column Mapping** The same column names contain **different types of data** depending on the row: - `Day` column contains: day names ("Thu"), dates ("23-Jun"), section headers ("=== CALL OUTCOMES..."), and Call IDs ("CY-2847") - `Total Calls` contains: call counts (405), dates ("2026-07-22"), and text ("Count") - `AI Handled` contains: counts (319), times ("06:02 p.m."), and percentages ("77.50%") ## 4. **Unnamed Columns** Three columns are unnamed: `_unnamed_7`, `_unnamed_8`, `_unnamed_9` (shown in step-1), though step-8 reveals they likely represent: - Duration (seconds) - Duration (m:ss format) - Sentiment Score/Label ## 5. **Inconsistent Date Formats** Multiple date formats appear: - "23-Jun" (day-month abbreviation) - "2026-07-22" (ISO format in call logs) - "Today" (relative reference) ## 6. **Missing Data Validation** The `Human %` column is frequently empty even when other percentage data exists (visible in steps 6, 13, 15). ## Recommendation This appears to be a **report/dashboard output incorrectly imported as a database table**. The data should be: 1. Split into separate normalized tables (daily_metrics, call_logs, escalation_reasons, etc.) 2. Converted to proper data types 3. Cleaned of header rows and formatting artifacts 4. Standardized with consistent date formats