Software Category Analysis – Demand Planning

Introduction

When forecasting applications were first developed external to the ERP system, a narrow spectrum of applications was developed for demand planning and these were primarily designed for the statistical forecasting process. However, statistical forecasting is only one of several forecasting processes, and over the years the variety of demand planning applications has increased significantly. Therefore, attempting to force all of a company’s forecasting needs through a single application is no longer necessary or desirable. The diagram below represents the various forecasting processes:

Major Forecasting Processes

A common mistake companies seem to make is trying to use statistical forecasting packages to manage the other forecasting processes. Part of the reason that companies do this is because both the major vendors (who tend to only offer statistical packages) and the major consulting companies incorrectly advise companies that statistical demand planning software can be used to manage non statistical process forecasting. This analysis package will primarily be focused on statistical forecasting applications, but where the application can branch out into other areas, we will point this out.

It is important to understand where demand planning fits among the different supply chain applications, as shown in the graphic below.

Common Supply Chain Application Categories Demand Planning

Demand planning one of the major categories of supply chain software. When companies implement an external supply chain planning module to be connected to their ERP system, they most often start with demand planning.

If you read most forecasting books they tend to focus very heavily on the mechanics of forecasting. They talk about forecasting methodologies (simple exponential smoothing, regression, etc.), however not enough get into the business process or how forecasts are used in real life. The first wave or generation of applications leveraged the ability to place not so sophisticated and quite sophisticated forecasting methods in software. These applications had high failure rates because they were generally not designed to be easily implemented or easily maintained. Applications in this category are what we refer to as first generation advanced planning products. A good marker of a first generation advanced planning forecasting application was if it expended most of the development effort in the application of complex methods and would often have a proprietary forecasting algorithm. During the period when first generation the prevailing wisdom was that the application mostly came down to how complex forecasting methods could be applied in an automated fashion. This period in software, which has been the dominant software development approach lead to some very bad habits. In fact, there little evidence that sophisticated mathematics can improve the forecast of difficult-to-forecast products, and this is a problem. Some studies do not show improvement from more advanced methods, but firstly the improvement is never very large, and secondly other studies come by later to contradict the original studies. In addition, complex methods should have to exceed a higher bar. Academics can apply complex methods in a laboratory environment over a few products far more easily than can be done by industry. This fact, along with the point that sophisticated methods are much more expensive for industry to implement than simple methods, is rarely mentioned. This point is made very well by J. Scott Armstrong:

Use simple methods unless a strong case can be made for complexity. One of the most enduring and useful conclusions from research on forecasting is that simple methods are generally as accurate as complex methods. Evidence relevant to the issue of simplicity comes from studies of judgment (Armstrong 1985), extrapolation (Armstrong 1984, Makridakis et al. 1982, and Schnaars 1984), and econometric methods (Allen and Fildes 2001). Simplicity also aids decision makers’ understanding and implementation, reduces the likelihood of mistakes, and is less expensive.

This idea, that forecasting software just came down to most sophisticated forecasting algorithm has been incredibly durable, especially since the research into this area clearly demonstrates that when actually implemented (that is not in a controlled academic environment where only a small number of items are being forecasted), more complex methods have a hard time defeating simple methods. The major limiting factor that companies face in leveraging forecasting functionality? That would be getting these types of applications to be understandable and to work consistently. That is making them easier to use. Unfortunately, statistical forecasting software vendors do not tend to compete on the basis of how easy their applications are to use, instead the competition tends to center around the sophistication of the mathematics that is within the forecasting methods. Having covered this subject extensively and worked in the field for quite a few years, we can say with confidence that simply having sophisticated mathematical forecasting algorithms/methods within the application is nowhere near enough to obtain consistent forecast accuracy. This leads many companies to have a false sense of security with regards to their statistical forecasting application – and it leads to companies asking themselves “We have statistical forecasting software, why can’t we get a decent forecast?” A certain cynicism has crept into statistical forecasting – however, forecasting is a disciplined endeavor and many implementing companies are not used to performing the type of data accuracy and discipline and testing that is required for forecasting. Forecasting vendors have also been responsible for overselling the ease by which a forecast can be improved by a system without the required work. Finally, many forecasting software vendors have been let down by major consulting companies who seem to primarily be able to staff IT resources for forecasting – that can configure the system, but lack very much experience or an understanding of forecasting beyond textbook knowledge.

Attribute-based Forecasting

Attribute-based forecasting is one of the most important developments since enterprise demand planning software began being used. I know this is a big statement to make, but I make it based upon research into the history of demand planning and in light of my research I am quite confident that my statement is true. After one uses an application capable of attribute-based forecasting, it’s difficult not to come to the same conclusion. And perhaps most interesting is the fact that attribute-based forecasting is still only used in a minority of companies (even though so many vendors say their applications are good at dealing with attributes). Attribute-based forecasting can allow different groups and departments to perform forecast aggregation as they are interested in seeing the data, and does not require that one single static hierarchy be used for all users. With attribute-based forecasting systems, the implementation approach for forecasting and analytic projects, which currently focuses on the technical details of complex database setup, can be changed. Instead of explaining the concept of realignment and spending seemingly endless hours debating what the “one” static hierarchy should be, that time can now be refocused onto determining and explaining how the business can get the most out of the forecasting application. This is an enormous benefit to forecasting system implementation projects, which have tended—along with many other supply chain planning software implementations—to become overly technical affairs with more emphasis on meeting deadlines and IT objectives than on adding value to the business.

Is the Word Out on Attribute-based Forecasting?

I cannot find a good explanation for why the term is searched for in search engines so infrequently. According to SEOMoz.com, the number of searches typed into Google per month for either the term “attribute forecasting,” or “attribute-based forecasting” or other derivations of these terms is negligible. There few Internet articles on this topic as well. Even a search through Google Books does not bring many results (this is usually a very comprehensive way to search for a topic). Attribute-based forecasting may not be used that commonly now; but I predict it will be in the future.

Forecast Disaggregation

Every statistical forecasting vendor that I am aware of states that they can perform forecast aggregation and disaggregation; however, there is a large gap between statistical forecasting vendors in terms of capability. This functionality is too important for clients to simply accept the statement from a vendor that “our product can do that.” In fact, aggregation and disaggregation should both be extensively demonstrated and tested by the company’s planners prior to selecting an application for purchase, in order to discern how easy the aggregation functionality is to use in competing systems. Aggregation and disaggregation capability cannot be an afterthought. Instead these capabilities must be designed into the application from the database layer up. The following quote on this topic is instructive.

Demand Sensing

We point out all trends in each software category that we cover, the valid and invalid. One of the invalid trends that we have been tracking is called demand sensing. The information on the definition of demand sensing is currently and primarily controlled by software vendors. Demand sensing did not come out of the academic community, so there have been few unbiased descriptions of what demand sensing is. Demand sensing is the adjustment of forecasting inside of the lead-time of the product, and therefore when the supply plan cannot respond. If our lead-time is 2 weeks, then demand sensing means changing the forecast less than 14 days out. Demand sensing is the adjustment of forecasting inside of the lead-time of the particular product, and therefore when the supply plan cannot respond. Because demand sensing changes the forecast within the lead-time, demand sensing cannot be considered a forecasting approach. To understand this, its important to understand that while broadly speaking a forecast is a prediction of a future event; in practical terms a forecast is a prediction of a future event that is given with sufficient advanced notification to be worthwhile. For instance, a forecast could be improved for a football game by waiting until 1/2 the game is over. However, when half the game is over, it’s too late to place a bet on the game. Therefore the forecast is not particularly useful because it occurs within the lead-time of when a some benefit from be received from it. The forecast of the game could be further improved by waiting until the minute before the game ends, but again its hard to see how anyone would accept this as a forecast. One could not want to compare the forecast accuracy of a person who forecasts games while in progress versus those that forecast the game before the game begins. This of course brings up the topic of demand sensing and forecast accuracy “improvement.”

So Where Does Demand Sensing Belong?

Instead, it is a method of creating the illusion of improving the forecast accuracy by change the forecast in way that can never translate into an improvement in supply chain performance. This is extremely appealing to any demand planning group that cannot meet its forecasting goals (which is not to say they are realistic or unrealistic).  Interestingly, the vast majority of articles on demand sensing describe it as a method of improving the forecast, and categorize it as a forecasting approach. While its true that forecast accuracy can be improved by waiting until the last minute, this is blurring the line of what forecasting actually is. Vendors and IT analysts like Gartner have completely confused the issue by combining demand sensing with demand shaping.

The list of basic things that most companies cannot do in their forecasting systems is often amazing. A company that should start with doing proven things like those above to improve the forecast, will instead choose to go with the latest “trend,” and buy demand sensing software. They prefer to use an approach that is untested and has no academic research to support the contentions that the vendors in this area (notably Terra Technologies and SAP) make about it.

Software Category Summary

All companies would like to improve their forecast accuracy. Forecasting continues to be one of the great areas of opportunity within companies. Many companies are still using some combination of ERP (which is not a good platform for forecasting) combined with Excel. This is a difficult way to improve forecast accuracy. There have been many mistakes made on demand planning projects, and gaining value from a demand planning application means analyzing those mistakes and adjusting the implementation methodology. Application of the same approaches will lead to the same outcomes.

Overall this software category has several very compelling applications, and also a growing application where we are not sure how it will develop in the future.

The links to the specific research you have paid for is included at the beginning and end of this Software Category Analysis. You will only be able to access the pages that apply to your subscription.

MUFI Rating & Risk

See the MUFI Ratings & Risk below for all of the applications we cover.

Vendor NameApplication
Big ERP
SAPMUFI Rating & Risk – SAP ECC
OracleMUFI Rating & Risk – JD Edwards EnterpriseOne
EpicorMUFI Rating & Risk – Epicor ERP
SageMUFI Rating & Risk – Sage X3
InforMUFI Rating & Risk – Infor Lawson
Small and Medium ERP
SAPMUFI Rating & Risk – SAP Business One
OracleMUFI Rating & Risk – JD Edwards World
ProcessProMUFI Rating & Risk – ProcessPro
RootstockMUFI Rating & Risk – Rootstock
ERPNextMUFI Rating & Risk – ERPNext
OpenERPMUFI Rating & Risk – OpenERP
MicrosoftMUFI Rating & Risk – Microsoft Dynamics AX
Financial Applications
IntacctMUFI Rating & Risk – Intacct
IntuitMUFI Rating & Risk – Intuit Quickbooks Enterprise Solutions
FinancialForceMUFI Rating & Risk – FinancialForce
NetSuiteMUFI Rating & Risk – NetSuite OneWorld
PLM
SAPMUFI Rating & Risk – SAP PLM
Arena SolutionsMUFI Rating & Risk – Arena Solutions Arena PLM
Hamilton GrantMUFI Rating & Risk – Hamilton Grant Recipe Management
Demand Planning
SAPMUFI Rating & Risk – SAP APO DP
TableauMUFI Rating & Risk – Tableau (Forecasting)
Business Forecast SystemsMUFI Rating & Risk – Forecast Pro TRAK
Demand WorksMUFI Rating & Risk – Demand Works Smoothie
JDAMUFI Rating & Risk – JDA Demand Management
ToolsGroupMUFI Rating & Risk – ToolsGroup SO99 (Forecasting)
Supply Planning
SAPMUFI Rating & Risk – SAP SNP
SAPMUFI Rating & Risk – SAP SmartOps
ToolsGroupMUFI Rating & Risk – ToolsGroup SO99 (Supply Planning)
Demand WorksMUFI Rating & Risk – Demand Works Smoothie SP
PlanetTogetherMUFI Rating & Risk – PlanetTogether Galaxy APS Superplant
Production Planning
SAPMUFI Rating & Risk – SAP APO PP/DS
DelfoiMUFI Rating & Risk – Delfoi Planner
PreactorMUFI Rating & Risk – Preactor
AspenTechMUFI Rating & Risk – AspenTech AspenOne
PlanetTogetherMUFI Rating & Risk – PlanetTogether Galaxy APS
BI Heavy
SAPMUFI Rating & Risk – SAP BI/BW
SAPMUFI Rating & Risk – SAP Business Objects
OracleMUFI Rating & Risk – Oracle BI
SASMUFI Rating & Risk – SAS BI
MicroStrategyMUFI Rating & Risk – MicroStrategy
IBMMUFI Rating & Risk – IBM Cognos
TeradataMUFI Rating & Risk – Teradata
ActuateMUFI Rating & Risk – Actuate ActuateOne
BI Light
SAPMUFI Rating & Risk – SAP Crystal Reports
QlikTechMUFI Rating & Risk – QlikTech QlikView
TableauMUFI Rating & Risk – Tableau (BI)
CRM
SAPMUFI Rating & Risk – SAP CRM
OracleMUFI Rating & Risk – Oracle RightNow
OracleMUFI Rating & Risk – Oracle CRM On Demand
InforMUFI Rating & Risk – Infor Epiphany
Base CRMMUFI Rating & Risk – Base CRM
SalesforceMUFI Rating & Risk – Salesforce Enterprise
SugarCRMMUFI Rating & Risk – SugarCRM
MicrosoftMUFI Rating & Risk – Microsoft Dynamics CRM
NetSuiteMUFI Rating & Risk – NetSuite CRM