• APPROACH

    We forecasted the total spend on two types of coupons – inflight and out-of-box coupons. We did this by:

    • Creating a data architecture in Azure to automate data extraction process
    • Multiple forecast techniques using coupon parameters such as face value, duration, days in market of the coupon
    • Transformation of ML prediction to data structures that can be consumed by PowerBI

    KEY BENEFITS

    The solution provided a coupon stimulator which predicts the duration and face value of the coupon to optimize spends

    It also provided an interface to user with report on coupon spend forecast to identify drivers of change and deep dive at a coupon level the reason for change from previous month

    RESULTS

    • Our solution was able to reduce the time required for data extraction process from ~800 hours/year to less than 100 hours/year
    • Our solution was able to achieve ~94% accuracy for In-flight coupons and ~80% accuracy for out-of-box coupons

    快猫 亿万狼友的选择