Andrew Lencsak - Demand Forecasting in Azure, D365, and Power BI

[@wirksworthsrailway]

Hello Drew! Lovely to see you here!

[@lylehenderson9222]

Andrew, just starting to evaluate Machine Learning for demand forecasting. However, when following your tutorial, I run into an issue when trying to capture the Web Service Endpoint Address. When I run the Test (at around 32:20 into the video), the system is asking that I enter Web Service Parameters. Not sure what it is looking for here.

[@luisevivar]

Click on REQUEST/RESPOND not in test

[@craigmcdermott2964]

Hello Andrew. If you don't mind a question: If you are starting form scratch with D365 forecasting, can you / should you run forecast for previous periods individually and work your way up to running it for the current period? Can you fore cast in weekly buckets instead of monthly buckets? Thank you - great presentation!

[@andrewlencsak4568]

hi Craig, it may be useful to run previous periods, however, it is not necessary to generate a forecast for the current period. It may be more accurate with the more forecasts you have, but the real important thing is the historical data set you choose, the more data you have, typically, you'll be able to analyze more trends and get more accurate results, however, you also can pick up more outliers. So not necessarily critical you do that. Yes you can definitely forecast in weekly buckets. You have the choice to select Days/weeks/months and the quantity when you run the "Generate Statistical baseline forecast" step to generate the forecast. Additionally, if you are manually creating forecasts in the Demand forecasts line form, you set the dates for the forecast and the dates of those forecasts can fall in weekly yearly, monthly, weekly, or daily buckets. You could even forecast at one level but reduce or consume the forecast at a more regular interval, i.e. forecast monthly but reduce or consume the monthly forecast weekly by a percentage or amount.

[@craigmcdermott2964]

@@andrewlencsak4568 Perfect. Thank you. I see the forecast buckets in the Stat Baseline. And for the obligatory follow up question... :-) How might forecasting in weekly buckets instead of monthly buckets alter the planned PO's that are presented to purchasing? Would Master planning simply look at the sum of the weekly forecasts across the time fences that are specified by Coverages? Do you think it would tend to increase the number of Planned PO's generated? Thank again for sharing your knowledge - greatly appreciated. No more follow up after this one - I promise.

[@andrewlencsak4568]

@@craigmcdermott2964 It could definitely lead to an increase in planned orders, but that can be mitigated by the setup of your coverage groups as Period based coverage type to sum the requirements into a single planned order, you would see multiple forecast lines pegged to the planned order representing the total demand.

[@georgygursky4303]

Good explanation, thanks!

[@fsaetre]

What a great video! I think this is a missing piece of many puzzles out there. Great overview and demo. Hope that this vide help many get going on this. ...and Batman is DC, not Marvel. 🤩

[@andrewlencsak4568]

Cheers Fredrik, appreciate the support. I'll blame the comic misnaming on how tired I was that day. Luckily for you all, you can see that in HD with the bags under my eyes :) :)

[@empowerd3657]

@@andrewlencsak4568 @Fredrik I don't care how tired someone is... No excuse for mixing up DC and Marvel... unforgivable. This help fill a hole for me brotha and I appreciate the content as always. Hope you are doing well and "staying curious and not judgmental" like my favorite super hero!

[@andrewlencsak4568]

@@empowerd3657 where was DynCon helping me out?! The sessions were edited and nobody caught that!! 🤣 now the world judges me by my knowledge of comics