Case Study: Chuỗi bán lẻ 12 cửa hàng tiết kiệm 210 triệu/quý nhờ AI Agent phát hiện thất thoát tồn kho

Case Study: How a 12-Store Retail Chain Saved VND 210 Million per Quarter with an AI Agent for Inventory Loss Detection

1. Chẩn đoán rủi ro cụ thể: thất thoát không nằm ở một món hàng, mà nằm ở điểm mù vận hành Một chuỗi thời trang có 12 cửa hàng...

1. Specific Risk Diagnosis: the loss was not about a few missing items, it was an operational blind spot A 12-store fashion retail chain once told HimiTek something we hear...

from datetime import datetime

RISK_WEIGHT = {
    "stock_mismatch": 35,
    "transfer_shortage": 30,
    "repeated_sku_gap": 20,
    "high_value_small_item": 10,
    "shift_adjustment_spike": 25
}

def score_inventory_risk(event):
    score = 0
    reasons = []

    if abs(event["pos_qty"] - event["stock_qty"]) >= event["threshold"]:
        score += RISK_WEIGHT["stock_mismatch"]
        reasons.append("POS và tồn kho lệch vượt ngưỡng")

    if event.get("transfer_sent") and event.get("transfer_received"):
        if event["transfer_sent"] > event["transfer_received"]:
            score += RISK_WEIGHT["transfer_shortage"]
            reasons.append("Chuyển kho nhận thiếu")

    if event.get("sku_gap_count_30d", 0) >= 3:
        score += RISK_WEIGHT["repeated_sku_gap"]
        reasons.append("SKU lệch lặp lại trong 30 ngày")

    if event.get("is_high_value_small_item"):
        score += RISK_WEIGHT["high_value_small_item"]
        reasons.append("Sản phẩm nhỏ, giá trị cao")

    if event.get("shift_adjustment_rate", 0) > event.get("avg_adjustment_rate", 0) * 2:
        score += RISK_WEIGHT["shift_adjustment_spike"]
        reasons.append("Ca làm có điều chỉnh tồn bất thường")

    if score >= 70:
        level = "nghiêm trọng"
    elif score >= 40:
        level = "trung bình"
    else:
        level = "thấp"

    return {
        "store_id": event["store_id"],
        "sku": event["sku"],
        "risk_score": score,
        "risk_level": level,
        "reasons": reasons,
        "created_at": datetime.utcnow().isoformat()
    }

from datetime import datetime

RISK_WEIGHT = {
    "stock_mismatch": 35,
    "transfer_shortage": 30,
    "repeated_sku_gap": 20,
    "high_value_small_item": 10,
    "shift_adjustment_spike": 25
}

def score_inventory_risk(event):
    score = 0
    reasons = []

    if abs(event["pos_qty"] - event["stock_qty"]) >= event["threshold"]:
        score += RISK_WEIGHT["stock_mismatch"]
        reasons.append("POS and stock quantity exceed mismatch threshold")

    if event.get("transfer_sent") and event.get("transfer_received"):
        if event["transfer_sent"] > event["transfer_received"]:
            score += RISK_WEIGHT["transfer_shortage"]
            reasons.append("Transfer received short")

    if event.get("sku_gap_count_30d", 0) >= 3:
        score += RISK_WEIGHT["repeated_sku_gap"]
        reasons.append("SKU mismatch repeated within 30 days")

    if event.get("is_high_value_small_item"):
        score += RISK_WEIGHT["high_value_small_item"]
        reasons.append("Small high-value item")

    if event.get("shift_adjustment_rate", 0) > event.get("avg_adjustment_rate", 0) * 2:
        score += RISK_WEIGHT["shift_adjustment_spike"]
        reasons.append("Shift has abnormal stock adjustment rate")

    if score >= 70:
        level = "critical"
    elif score >= 40:
        level = "medium"
    else:
        level = "low"

    return {
        "store_id": event["store_id"],
        "sku": event["sku"],
        "risk_score": score,
        "risk_level": level,
        "reasons": reasons,
        "created_at": datetime.utcnow().isoformat()
    }