I was recently asked a question regarding the advantages and disadvantages of DP. I thought I would take part of the email exchange and post it here.
______________________________________________
Questioner:
I would be interested in knowing how DP is different from demand planning in Manugistics.
Especially:
1. Weaknesses
2. Strengths
3. Level of integration with supply planning (eg ability to effectively measure forecast consumption)
4. APO suitability for products with short life cycle (eg movie/music DVD)
Shaun:
Well, ok here is a more technical reply. I will think on it, I have articles you can read already to get a general background. Here are the DP related posts on sapplanning.org
http://sapplanning.org/2009/05/22/dp/
However, I am not sure I get the short life cycle issue. As you are well aware, there is nothing new under the sun in forecasting, DP uses the same forecasting approaches that you have seen a million times, and has an auto select feature like everybody else. The main thing about DP is that it is difficult to configure because it uses the Data Warehouse Workbench and requires the learning of an entirely new vocabulary (InfoObject, InfoProvider), and a strong tolerance for pain to administer. It can forecast short or long term, like any other forecasting system. So what do they mean by applicability to short term forecasting, is there some trick or method companies who do short term forecasting use? SAP F&R, which is replenishment focused handles short term forecasting by allowing store managers to input their order amounts, and decentralizes the ordering and short term forecasting.
I can also say, I have never seen DP used for short term forecasting. DP typically has the longest time horizon, SNP shorter, PP/DS shorter still and so on.
As for DP to SNP, this is the most commonly implemented integration in APO-land. In terms of measuring forecast consumption, here is how it works.. and I quote from Supply Chain Management with APO by Jorg Thomas Dickersback.
“The horizons for the forecast consumption (number of days forwards or backwards from the confirmed date of the sales order) and the consumptive mode (backwards only, forwards only or both) are maintained in the product master.”“The forecast is calculated each time a sales order is created, changed or deleted and each time the forecast is released from DP.”“the consumption of the forecast is performed on the location product level, although it is possible to restrict the forecast consumption to a more detailed level using descriptive characteristics” (furthermore, forecast consumption can be set as backwards, (sales orders consume forecast quantities that lie before the requirements date), backward/foreward (sales orders consumer forecasted quantities that lie before the requirements date. If the actual demand is no satisfied, the sales orders then consume forecast quantities that lie after the requirements date in order to satisfy the remaining demand, forward consumption (sales orders consume forecasted quantities that lie after the requirements date). – The recommendation is that the for fast moving products, the consumption can be set backwards/forwards for a shorter period of time (say a week),and for slow moving products, the consumption can be set for a month or longer.Descriptive characteristics are a different conversation, but essentially, consumption is inherent each time the forecast is released to SNP. The question of how inherent it is if it is released to SAP ERP and not SNP (an alternative implementation), is another question. I think it probably is, but that would become a question for the integration mechanism, called the CIF, and is another line of inquiry.
Strengths:
All the methods every other forecasting system has, including autoselect or best fit:
- Integration with SCM
- Integration with SAP ERP
Weaknesses:
- Data Warehouse Workbench
- The necessity to learn a new vocabulary to use
Questioner:
What type of exception reporting (eg actual greater than forecast) are available in APO? How difficult is to create your own report (does this require coding or is it similar to combining fields like in access DB)?
Short term forecasting (the way you described it) is managed a lot better via retail inventory management solutions or VMI.
Does APO support:
- Easy elimination of certain data points (to exclude from forecast)?
- Good graphical depiction of interpolation (past data) and extrapolation (future forecast)?
- Causal analysis (or factor-based regression)
- Fair share logic (when forecasting is done at a sub-assembly level and then allocated to final SKUs using %%)
- Profile-based forecasting (with some factors impacting profile characteristics – eg slope)
- Calculation of statistical parameters (which helps to determine the quality of forecast)
- If a client outsourced all manufacturing and only drops orders to be fulfilled using APO DP for demand planning and its own tool for VMI (on the customer side) does it make sense to use CTM (with VMI generating replenishment orders)?
- Since production is outsourced, client is not concerned with manufacturing capacity while it is critical to replenish on-time since you are dealing with time-sensitive product and WallMart.
Shaun:
Exceptions are handled by the alert monitor. See this paper for the details.
http://snappdocs.files.wordpress.com/2007/12/how_to_use_the_apo_alert_monitor_for_reporting.pdf
It is very flexible and access can be given to whoever the project sees fit. We are trying to figure this out now on my current project.
- Not sure about easy elimination of outliers, I don’t think so. Sounds like a report that would have to be run off of the DP data. SAP would answer differently, that all of that is handled by BW, but it still has to be coded, and its a separate system, not integrated — so the process would be disjointed.
- Graphical depiction is weak. The charting in DP is nonexistent. I am a big fan of the charts in the forecasting in the SPP module, but that is not DP so does not apply. (DP and SPP development teams don’t seem to talk)
- DP can do some causal based forecasting, is basic but functional. SPP has a stronger functionality called “Leading Indicator” forecasting, which is causal, but since it’s not implemented more than a few places, it’s not very well tested. My take is it’s not that important because very very few clients do causal forecasting. But I would like to hear your take on it.
- Yes, the forecast can be disaggregated this way, and a number of other ways as well. SAP makes a big thing about this functionality.
- I am not sure what profile forecasting is. I think this is just forecast parameters (i.e. a field for the trend component and so on), if that is what you mean then yes.
- Well DP has best fit. But I don’t see how calculation of the statistical parameters helps determine the quality of the forecast. Isn’t the quality of the forecast determined simply by the comparison of the old forecast vs. the old actuals? Maybe you can explain some more here.
