I’ve lost count of the times that I’ve been shown a forecast spreadsheet which shows the sales forecast in monthly buckets – and of course we’re going to want to import that into Dynamics AX.
So we face two problems – the first of course is that the Demand forecast table holds the forecast in separate records – so we’ll need to convert our matrix of monthly forecasts into a columnar list – or we’ll need to use a customised import function. In the past I’ve used both options, but that’s not the subject of this post. This post’s about the second problem, converting a monthly forecast to a weekly demand pattern.
The standard Dynamics AX solution is the ‘Period key’. So let’s create a Demand forecast for an item. Inventory management > Inquires > Forecast > Demand forecast:
I’ve entered a Date (1st December 2015), and an Item number and a sales quantity. The system’s defaulted the sales unit, the sales price and a warehouse. I suppose that we should note in passing the warehouse might not be needed – we could have setup our Storage dimension group so that ‘Coverage plan by dimension’ isn’t active for the warehouse – but I don’t see that setup being used very often.
So, we already know that a monthly forecast like that isn’t very helpful to production planners who are working on weekly (or 2-weekly) production plan cycles – or buyers who want to create purchase orders every week or two. To solve this problem we’re going to use an Allocation method of ‘Key’ – but first we’ll need to setup a ‘Period key’. (Incidentally the other Allocation option of ‘Period’ is useful for creating recurring test data – where it’ll quickly duplicate your entered forecast on a weekly or monthly basis up to an entered ‘End date’). But back to the Period key. Inventory management > Setup > Forecast > Period allocation categories:
I’ve created a new key; clicked on the Lines button; and added four lines (25% per line), with an offset of 7 days per line. Your allocation doesn’t have to add up to 100% but mostly it will. Back on your demand forecast, select the Allocation method Key and enter your Period key (and click Create lines or Save) and the system will split your entered forecast as specified:
So typically, what I’ve seen done is that the forecast import allows the user to specify a Period key – and often that’s only used for the first few months of the forecast. Once you get out past your short-term planning horizon, having a forecast in monthly buckets is probably acceptable – and not using a Period key on all forecasts will speed up the import of the forecast, and master planning’s run time. But a major problem with that type of import is that it ignores the 4,4,5 weeks to months conversion pattern that gives you a more ‘real-world’ conversion. So really we should build a bit of intelligence into our import (by which I mean another parameter on the calendar we’re using) so that we can use two (or more) period keys to convert the monthly forecast to planning weeks.
Another problem we need to address when we’re importing forecasts is what happens when the forecast date (typically the first of the month) falls on a non-working day. This year (2015) 1st November is a Sunday. Enter a forecast for 1st November 2015 and run master planning and you’ll get:
That first line shows a Requirement date of 30th October, not the 1st November. The system’s reacting to demand on a non-working day by prompting you to supply against the demand on the first working day before the forecast. Unfortunately most planners won’t want that behaviour and your forecast import will need to use a calendar to determine the first working day after the forecast date. Actually, the other forecast I entered on 1st December shows another problem. The 1st December 2015 is a Tuesday. Do your planners want the forecast to be entered on a Monday (or perhaps a Friday?) – more customisation of your import is the only way forward here.
Finally – I’ve shown here a Period key with an even distribution (25%; 25%; 25%; 25%), but you don’t have to use period keys like that. It may be that in your company most orders are received at the end of the month, and period key like (15%; 15%; 20%; 50%) might give you a better approximation of an item’s actual demand pattern. I once saw monthly forecast’s being split into 60% at the beginning of the month and 40% two weeks later. That worked well for that company because they had a 2-week production cycle (most products were made once every 2 weeks) and they could use the 60% forecast to drive a ‘spike orders’ report, which highlighted those items that had sold more than 60% of the monthly forecast in the first 2 weeks.
Of course, you might be able to avoid all of these issues by converting your marketing department into giving the planners a weekly forecast. Good luck with that one.