Skip to main content
Jul 17, 2020 1:05 PM UTC
Question

Forecast Accuracy Intelligence v12.1

With the release of v12.1 and the functionality of Forecast Accuracy Intelligence, is it possible to measure forecast accuracy on Causal Forecasting?
The input on causal forecasting is based on product rollout and failure rate. The product rollout projection is based on a sales forecast of product, but over time this forecast can be inaccurate. Is it possible in to track for example the input of the last 12 months and the differences?
Also the failure rate input can be incorrect, resulting in a too high forecast. Is this part of the functionalities in v12.1? Measuring forecast accuracy on this is very important for us, also enabling the discussion on failure rate inputs internally. 

    3 replies

    Jul 17, 2020 2:09 PM UTC

    Hello Mark,

     

    Yes, we do calculate forecast accuracy for causal forecast method. Some details below:

    • In SPM, we calculate 1 period lag based accuracy (few different metric) on the causal forecast stream on overall demand and archived forecast.
    • In PAI, we calculate 3 month, 6 month, and Lead time lag based accuracy( multiple metrics) on the causal forecast stream on the overall demand and forecast historicized each period.
    • Additionally, we can see historical failure rates and current calculated failure rate in SPM screens and do some sort of comparison.
    • However, we don't have any direct accuracy calculations on the product rollout or failure rate itself. That may be a good enhancement to plan on our road map.

    Please let us know if you have any additional questions.

     

    Thanks,

    Jai

    Jul 17, 2020 2:15 PM UTC

    Thanks Jai!

    So, is it possible in v12.1 (either SPM or PAI) to compare historical product rollout input (without any calculation)?

    Jul 20, 2020 12:29 PM UTC

    Hi Mark,

     

    Yes, on the current SPM Product roll out screen, we can see both historical and future product roll out records. However, we don't have separate table for actual roll out vs future roll out provided.

     

    In current PAI, we can historicize the product roll out table and compare the same period roll out (predicted) vs actual. For e.g, if we projected June's roll out to be 20 in May and when system date changes to July, we get actual roll out for June to be 18. Then we can compare roll out table snapshot from May vs July for the period: June.

     

    We can put it in road map to make it more out of the box.

     

    Thanks,

    Jai