Posts Tagged ‘Forecasting’

Too Big to Fail? Too small to save? Too inefficient to work with? Too old school to educate?

Tuesday, November 17th, 2009

Policy makers hang on to the idea, or more like fear, that it is too risky to let big banks fail. There is a lot of talk and surmise that Lehman Brothers failure may have caused some of the precipitation last year. Not for sure.

This fear on the part of the policy makers has created a put option or a gamble of sorts for Wall Street. This TBTF policy has created loads of easy money. The printing press is running non-stop and the Fed is flying helicopters all over the country and sprinkling greenbacks (or may be just on the backyards of the big banks).

V-Shaped Recovery

V-Shaped Recovery

With essentially zero-cost money, the likes of Goldman Sachs, are now taking bigger risks in the markets and other asset investments. Savers faced with zero yield on their deposits are also pushing to take on more speculative risks and risky investments.

If this TBTF policy is going to cripple us, then why let anything be too big? The Big ones think of it as a right to come hat-in-hand, at the slightest problem that they suspect is systemic. If the Government and the Fed have the obligation to save the Big ones when in trouble, do they also earn the right to NOT let them become too big? Sounds fair enough.

W-Shaped Recovery

W-Shaped Recovery

Perhaps policy makers of the 21st century need to think about a TBTF premium and collect monies from financial companies that are beyond an acceptable size. Perhaps policy makers can collect this insurance premium or levy on a tiered basis that increases with the size. Maybe they also have a say on how big they can become.

The reluctance of the Fed to mop up the excess liquidity and the lack of any developing financial regulation are puzzling facts.

* Will this lead to Financial crisis 2?
* Is America disadvantaged by the sliding dollar?
* How will it affect your plans and forecasts?

Is the recovery V-shaped or W-Shaped?

If you are c-Level executive, you need to wonder about these issues in your intermediate and long-term planning process.

Excellent participation in the Forecasting Tutorial

Tuesday, November 3rd, 2009

The forecasting workshop on Oct 22 and 23 held at Whippany, NJ was well attended by companies from different industries from Consumer goods, to medical devices to Technology companies. We had participants from

  • SMSC
  • J&J
  • Tools Group
  • Bush Brothers
  • Merck
  • Crabtree and Evelyn
  • Avon Products
  • Niles Audio

Thanks to the great audience, the workshop was very interactive with people talking about their real world planning experiences and comparing notes on how they dealt with specific situations.  A number of questions were addressed during this interactive forum.  I truly enjoyed the participation and nothing better than conducting a workshop with an engaging audience.  Thank you.

The major focus on the first day was forecasting in the current economy.  There was a lot of discussion on the V-shaped recovery and how to forecast for it.  Although exponential smoothing is an adaptive technique that normally catches up with a lag, it is difficult when the demand suddenly drops and then sharply recovers a few months later.  We all discussed the importance of scenario planning and other techniques that are important like leading indicators.

The guest speakers Mark Temkin and Jay Nearnberg also provided valuable insights to the group.  There was discussion on inventory optimization, S&OP and demand metrics.  There were number of questions on modeling and metrics, particularly the usage of MAPE and the methodology to compute the MAPE.  The weighted Mean Absolute Percent Error seemed to be the most common performance metric used by most organizations, although some had used a variation of it.

Please feel free to post any follow-up questions on this workshop either here or in our Linked-In Group for DemandPlanning Net Training.

Mark Chockalingam

Woburn, MA

November 3, 2009.

Can we use Multiple Linear Regressions to model sales promotions?

Wednesday, July 29th, 2009

The simplest approach to model promotions is through interventions. This is also called event modeling in the literature. We collect a list of events and classify into types of events by the intended effect or by the sensitivity on the volume. For example, retailer tabs may have a 10% lift on volume while an FSI may have a 20% lift.

Then we create different indicator variables and use them as independent variables to do the volume forecast. In some of the tools, you can specifically calculate the volume lift by event type.

In our training workshops we have used Forecastpro and Excel to illustrate event modeling. Excel takes a little bit more effort to set up these models. Ideally you want to use event models with either ARIMA models or using Exponential smoothing models with events.

If you want to learn more on event modeling in demand planning, you may want to join us for the DP Web workshop to be held on Friday July 31 at 10AM EST.

For more information, see

http://demandplanning.net/workshops

The mystery of Moving Averages

Friday, July 10th, 2009

A user in an APO forum lamented:

“I am trying to use the Moving Average option in APO DP but I get a static forecast that remains constant for all the future months based on the history.  I would like the statistical forecast to be a moving target not a constant.  Not sure how to achieve this?!”
“All I get is the same statistical forecast volume for all 24 months. This is not want we are looking for.  The system is giving me a forecast like this:  month1 – 915, month2 – 915,… month 24 – 915″
“But I want the system to give me forecast that is different each month: 
Future: month1 – 915, month2 – 902,… month 24 – 908.”

Before we go too far with the micro details of these numbers, let us first realize some qualities of a good statistical forecast. 

One of the important qualities of a forecast is robustness. Robustness means the forecast does not change like a yo-yo just based on new historical data point.

So the objective is not to mimic the history but to produce a forecast that minimizes the error. Theoretically, Moving averages should produce a constant forecast into the future at least after the first two periods.  The idea of the moving average is that it will change by at least one third of the impact of the new observation that is different from the
forecast.
Other than this, let us understand the difference in error in what is being proposed here.  A forecast that is 915 each month is off by 1% from another forecast that varies between 902 and 915 over the entire forecast horizon. 

A contrived model that looks fancier with oscillations in results between 902 and 915 perhaps can be 1% better on average than another model that proposes to use a constant 915 every month. I don’t think the trade-off to improve forecast quality by 1% is worth the model search to come up with a more complex model that mimics
the history better.

The fact that you are choosing moving average means that the data series is relatively more stable. We as planners should let moving average do its job and move on to more complex items that need your attention – items that
have a persistent trend or seasonality or both.
Trying to fit a holt-winters model when the series begs you for a constant model is NOT a good use of time. Note that SAP APO classifies moving average under constant models.  In fact, specifying that you want Holt-Winters models in such a scenario will give you a First Order smoothing model which again gives you constant forecasts into the horizon. 
If you want to learn more about error metrics, you may be interested in http://www.demandplanning.net/apoMetrics_webworkshops.htm.