Ideas on the best way to forecast when system has been configured at lowest level

Please give me ideas on the best way to forecast when the system has been configured at the lowest level.  For example, we can only forecast at an item, location, and account chain level.  Also, we can enter the forecasts only by week.  What advice would you give to overcome this hurdle?  Only look at certain items at certain locations at certain account chain levels?  When management is giving you the information at a monthly level, what is the best way to make necessary forecast adjustments all in a week's time frame?

 

Thank you very much!

Tagged:

Answers

  • @Elsie 

     

    What format are you given the monthly data?

  • Hello Michael,

    Data is given in weeks.  I have to export to Excel to make monthly, quarterly, and annual calculations.

    Any advice you can give would certainly be appreciated!  Thank you!

  • Data is given in weeks.  I have to export to Excel and make the monthly, quarterly, or annual calculations.  Any help you can give would be appreciated!

  • Hello @Elsie 

    "Data is given in weeks.  I have to export to Excel and make the monthly, quarterly, or annual calculations.  Any help you can give would be appreciated!"

    Before venturing into suggestions for forecasting at different levels of the hierarchy, may I ask where are you exporting the data from? Anaplan? and why are you using Excel to aggregate the weekly data?

  • Yes, I am exporting the data from Anaplan.  I am using Excel to aggregate the weekly data to view the data at a higher level (monthly, quarterly, annually).  Anaplan, at this company, is not configured to view the data at a higher level, only weekly, by item, by manufacturing location, and finally by account chain.  The weekly data is sometimes too detailed depending on your audience.

  • @Elsie 

    Like @einas.ibrahim I'm a little confused on the exact use case but I think I see where you might be going. The core part of your question seems to be how to disaggregate the monthly forecast to weekly forecast. If so, there are two important things to consider:

    1. A model can only have one "model calendar". If the calendar of your model is monthly, you'll have to either set up a "fake" weekly list or you will have to create another model and import your data. Either is perfectly acceptable but you'll need to decide which is best based on your business requirements.
    2. To disaggregate from month to week, you have to decide on the level of accuracy you want. Ideally, you'll have a basis to refer to like last year's data by week that will give you a forecasting profile (indexed by week). If you have this you'll be in good shape, you'll just need to account for the 53rd week and any holiday or seasonal shifts. More accuracy can be obtained if you draw down the forecast to a daily level so you can ensure the monthly forecast is exactly allocated to the right weeks (weeks don't end up on the first of the month and the last day of the month).

    To go back and forth between a Gregorian Calendar and a Weekly Calendar, I recommend you peruse this best practice article which gives you step by step instructions. 

    If I'm way off, give us some additional details about the dimensions of the source data and the dimensions of the target data. This will help a lot.

  • Hello @Elsie 

     

    Could you please elaborate on what you mean by "Anaplan, at this company, is not configured to view the data at a higher level, only weekly".
    I am not aware of any Anaplan configuration that will bar you from having data aggregated by time at higher levels such as months, quarters,....etc.

    One way I could think of is if the summary method for the line item(s) is intentionally set to NONE for time.

    We can easily have a module view with just weeks and another view for months,..etc

     

    Moving the data from Anaplan to Excel and then Anaplan again is one of the fundamental things Anaplan is helping us avoid.

     

    I am still not clear on what exactly is the issue you are facing, I wanted to clear the export into Excel issue first before we dive into details of forecasting.

     

    • If your issue is converting from weeks model calendar to monthly forecast then @JaredDolich provided good suggestions.
    • If the issue is that you have the data on a weekly and SKU level but higher management for example plan/forecast on a monthly and Product levels then that's a different issue.
  • Hello,  

     

    I really appreciate your advice.  I need the exact opposite.  I want to aggregate the weekly forecast to a monthly level, quarterly, and then annually.  I was told Anaplan, at this organization, is only able to forecast at the lowest level.  There is no button to aggregate at a higher level.  

     

    But, I appreciate your advice about faking.  Possibly, we can fake the model calendar into higher buckets, such as monthly.  I am also assuming we can fake the item, manufacturing, and chain account into a higher level such as item only and have the forecasts disaggregate according to their historicals and the best fit forecasts.  

     

    The problem is that it will be very time consuming to enter forecasts by the item, manufacturing location, and account chain so I am exploring other solutions I have used in the past.  

     

    Your advice was very helpful, and I will discuss with the consultants further.  I cannot thank you enough!

  • Thank you, Einas!

     

    How do I check the summary method for the line item is intentionally set to NONE for time?  I am assuming that I once I flip the switch the update will happen in mass and I do not have to complete this step for every item, manufacturing location, and account chain combination?  I believe the earlier team wanted to have Anaplan configured at the lowest level but they are no longer with the organization so I recently joined to help make the forecasting process work.  

     

    My issue is that I do not want to have to make forecast adjustments at every item, manufacturing location, and account chain combination.  There could be at least 20 different combinations for one item.  This is too time consuming, and we would never finish the forecasting cycle in a month's time.  Yes!  You summarized my issue in your second bullet point about having my data on a weekly and SKU level, and adding account chain.  I do not want to get into all of the details with executive management.  I want to discuss at a higher level and only for exceptions talk in details.  

     

    Loved your suggestion about the summary method for the line item and also Jared's about "faking" the calendar!  THANK YOU!  I will wait for your advice about how to check if the summary method for the line item is intentionally set to none.  

  • Hi @Elsie 

    In order to check/edit the summary method for a line item, you need to;

    • Go to the blueprint of the module
    • Scroll to the right until you find the column titled Summary
    • Click on the 3 dots on the right side.
    • A dialogue will open up to allow you to determine the aggregation method for both list hierarchy and time.
      image.png

    That's the easy part in terms of utilizing Anaplan to aggregate values on the time scale.

     

    Now, regarding your forecast issue; there are a couple of things you need to think about.

    1. You have to know the level where you forecast. For example, is it per city and product or per store and product? 
    2. You need to determine if the forecasting will be top-down, bottom-up, or both.
    3. You need to determine how you will disaggregate the data from a higher to a lower level.
      For example, if a product manager planned # units sold at the level of Products and Quarterly, Assume for Product A1 and Q1 the value was 1000 units, you need to distribute this value by month or week, do you distribute the 1000 units evenly across Jan, Feb and March or you use some sort of seasonality. 

    Most probably you will use a combination of Top-Down and Bottom-up approach and the good news is Anaplan allows you to do so. You just need to plan for it and answer questions like the above.

     

  • Einas!

    I cannot thank you enough!  Your detailed explanation and forecast suggestions were exactly what I needed!  THANK YOU!