Mushab Mon K

-- Delay cohort analysis with running averages
WITH delivery_metrics AS (
    SELECT 
        region,
        dispatch_date,
        address_type,
        delivery_time_hrs,
        AVG(delivery_time_hrs) OVER (
            PARTITION BY region 
            ORDER BY dispatch_date 
            ROWS BETWEEN 6 PRECEDING AND CURRENT ROW
        ) AS rolling_7d_avg
    FROM deliveries
    WHERE market IN ('KSA', 'UAE', 'EGY')
),
delay_cohorts AS (
    SELECT *,
        NTILE(4) OVER (
            PARTITION BY region 
            ORDER BY delivery_time_hrs
        ) AS delay_quartile
    FROM delivery_metrics
)
SELECT region, address_type,
    COUNT(*) AS total_deliveries,
    ROUND(AVG(delivery_time_hrs), 2) AS avg_delivery_hrs,
    ROUND(AVG(rolling_7d_avg), 2) AS trend_avg
FROM delay_cohorts
WHERE delay_quartile = 4  -- Top 25% slowest
GROUP BY region, address_type
ORDER BY avg_delivery_hrs DESC;

-- Address quality scoring with error classification
WITH validation_results AS (
    SELECT 
        country,
        city,
        address_format,
        validation_status,
        error_type,
        COUNT(*) AS case_count,
        SUM(CASE WHEN validation_status = 'PASS' 
            THEN 1 ELSE 0 END) AS pass_count
    FROM address_audits
    WHERE audit_date >= DATEADD(month, -3, GETDATE())
    GROUP BY country, city, address_format, 
             validation_status, error_type
),
quality_scores AS (
    SELECT country, city, address_format,
        SUM(case_count) AS total_cases,
        ROUND(SUM(pass_count) * 100.0 / 
              SUM(case_count), 2) AS quality_pct,
        RANK() OVER (
            PARTITION BY country 
            ORDER BY SUM(pass_count) * 100.0 / 
                     SUM(case_count) ASC
        ) AS quality_rank
    FROM validation_results
    GROUP BY country, city, address_format
)
SELECT * FROM quality_scores
WHERE quality_rank <= 10  -- Bottom 10 per country
ORDER BY country, quality_pct ASC;

-- Optimized partitioned query for cost reduction
CREATE TABLE delivery_analytics
WITH (
    format = 'PARQUET',
    partitioned_by = ARRAY['country', 'year_month'],
    external_location = 's3://data-lake/analytics/'
) AS
SELECT 
    delivery_id,
    driver_id,
    CAST(delivery_timestamp AS DATE) AS delivery_date,
    delivery_time_minutes,
    address_quality_score,
    CASE 
        WHEN delivery_time_minutes > threshold THEN 'DELAYED'
        ELSE 'ON_TIME'
    END AS delivery_status,
    country,
    FORMAT_DATETIME(delivery_timestamp, 'yyyy-MM') 
        AS year_month
FROM raw_deliveries
WHERE delivery_timestamp >= DATE '2024-01-01'
  AND country IN ('KSA', 'UAE', 'EGY');