Previous Webcasts--
Guidelines for S&OP in the Multi-site Enterprise
Wednesday,January 11, 2006, 2pm Eastern, 11am Pacific
Principal Speakers
Share Green, Director, Program Management, Teledyne Technologies EMS , Glen Margolis, SVP Services, Steelwedge Software
Click here to view recording

Three Keys to Successful S&OP
Wednesday, October 26, 1pm Eastern, 10am Pacific
Principal Speakers
Tim Vaio, Managing VP,
Hitachi Consulting
Seema Phull, Director
Process & Technology,
Enterasys Network
Click here to view recording


State-of-the-Art Sales Forecasting Management
September 13, 1pm Eastern, 10am Pacific
Principal Speakers -
Professor J. Tom Mentzer, University of Tennessee
Todd Jones, Sr. Director
Sales Operations, QLogic
Click here to view recording

Sandhill.com Nov 28, 2005
A Hybrid Strategy for
On-Demand Success
By Timothy Campbell

Manufacturing Business Technology (MBT) November 2005
Cover Story: Balancing Supply & Demand
By Sidney Hill, Jr.

Baseline Magazine (11/08/05)
Air Products in the Pipeline
By Mel Duval, 11/08/05

START Magazine July-August 2005
Looking Forward: Forecasting and Consensus Planning
By Christine Pfefferle, Director of Global Demand and
Order Management

Turbo Charging Your Sales Forecasting Process
By Anil Gupta, The Applications Marketing Group

The Sales Funnel - a Critical Part of the Demand Planning Process
By EJ Tavella, Senior Director
of Business Development, Steelwedge Software

The Importance of Exception-Based Planning

By Anders Gjerde, Senior Manager, Steelwedge Software, Inc.

In today’s world of tera-byte databases many planning organizations are struggling with the issue of how to transform all their data into useful information to improve planning and – ultimately – their business performance.

Many planning and business intelligence systems have helped automate the planning process by running reports on a monthly, weekly or even daily basis. However, few planning systems have succeeded in proactively extracting important issues – it is up the reviewer to sift through the information and find and interpret important issues. Moreover, business intelligence tools typically do not support taking corrective actions based on the information provided.

This is where exception-based planning comes in – the ability to extract anomalies or sudden changes that can impact the business, to bring these exceptions to stake-holders in a timely manner, and where stake-holders – in turn – easily can take corrective actions.

Steelwedge Software has developed a methodology to enable exception-based planning following these steps:

• Define
• Communicate
• Act
• Measure

This article describes how Steelwedge has helped customers implement exception-based planning.

The first step is to define the exceptions – these must be measurable quantities, and they need to be associated with threshold values for triggering the exception. In a complex, dynamic system where values can be viewed at different levels of aggregation and/or in different business hierarchies, alert measurements are often best stated in percent terms – e.g., % change in forecast value from one day to the next. The pitfall with percent changes is, of course, that small (insignificant) numbers can have huge percent changes even if the absolute change is insignificant to the business overall. Hence, percent changes with additional criteria for minimum absolute changes often works well.

In defining exception metrics, a number of considerations must be made:

• What anomalies or changes can affect the business? In particular, what changes in one end of a business process are important for downstream processes?
• Are we only interested in BIG changes, or are many small changes over time (that add up to a BIG change) also important?
• What are reasonable threshold values? Setting the threshold value too low can overload planners with too much information, and setting them too low can lead to overlooking important signals.

Here are some examples of metrics:

• % change in forecast from yesterday
• % change in forecast since last week
• Orders equal to or above forecast
• Inventory value below X% of forecast

Another consideration to make is whether exceptions should be generated
as soon as they occur, or if they should be generated as part of a systemic process that evolves around a calendar. The trade-off here is often one of running the risk of being bombarded with too much information (in case of single alerts each time a threshold value is exceeded), or of getting the information slightly delayed (when a scorecard is distributed).
A third option – not exclusive of the other options - is the ability to run
exception reports on command.

With a set of well-defined exception metrics, the next step is to determine how to communicate them to stake-holders. This communication starts with a business process that incorporates exception management. Based on this process, exceptions can be communicated via email as single alerts or as scheduled dashboards or scorecards, or they can be queried on-demand. Either way, the ability to drill-down on the information to understand what the situation is and be able to effectively delegate or transfer a task is important.

Example Dashboard Email

An exception remains FYI only until it can be acted on. In some cases it requires no other action than communicating to stake-holders that the game has changed. In other cases is requires corrective actions – e.g., when a planning error has been entered into the system and puts down-stream processes at risk of being based on bad information. In this case it is imperative that the analyst can go in and make the corrective action expediently. This can be facilitated by having the ability to change plans in the same system as the alert system – thus enabling detection and direct navigation to the offending record(s).

Like any planning process, measurement is the last and arguably the most important step in the process. In particular, effective measurement needs to track two dimensions: (1) Does the process work?, and (2) Does exception planning improve results?

Measuring how well a process works is not easy. One step towards monitoring a process is to get a clear view of who is participating, and who isn’t:

Similarly, a more detailed view towards monitoring a process would be to generate a reports that shows who made what changes, and when.

Another, more involved way to measure process efficiency is to “tag” corrections resulting from exception alerts as “exception corrections” when submitting a change. With this, management can compare forecasts or plans with or without the exception-based process, as follows:

In the above simple example we see that forecast error without exception alerts would have been 20%, whereas the error with exception alerts was 10%. As always, when measuring “what could have been” we need to interpret the results with a grain of salt – as indications rather than hard facts.

A different category of measurements has to do with the results from a planning process over time; is the forecast or plan accuracy improving over time? Here are some common forecasting or planning accuracy measurements (described in more detail in the following article:
http://www.steelwedge.com/newsletter/may05/newsletter_Gjerde_May05.html)

1. Percent Error
2. MAPE
3. BIAS

The above steps describe how exception management can be implemented and how it can work. In practice, it is not a one-shot deal – continuous improvement during and after implementation is critical to get the most out of your investment in exception-based planning.

About the author:

Anders Gjerde is a Senior Manager and Business Analyst at Steelwedge Software. Since joining Steelwedge in 2002, he has worked with customers
to implement innovative solutions to help them solve a wide range of planning and performance management problems. Prior to joining Steelwedge, Anders was Director of Global Client Solutions at Decision Focus/Talus Solutions (acquired by Manugistics, Inc in 2001). Anders holds an MBA from the Norwegian School of Economics and Business Administration