Comments on Brightwork Articles on Demand Sensing in Forecasting

Table of Contents: Select a Link to be Taken to That Section

Executive Summary

This article contains comments from articles on demand sensing in forecasting.

Introduction

These comments are in response to the articles on demand sensing in forecasting.

Comment #1: Tibor Vister

Hi Shaun,

As already started discussion in another topic I would like to give you my comments here as well.I see demand sensing useful technique when used in combination with postponement/risk pooling techniques (e.g. ship through one CDC, keep safety stocks in CDC, etc.), especially when applied in companies with multi-echelon network (large and complex organizations). In such environment you deal with “multiple” lead times from supplier to CDC (e.g. 20 weeks out), from CDC to local hubs (e.g. 2 weeks out), etc., and having improved and most updated/accurate forecast on 2 weeks out time fence on which you need to execute local hub replenishment is of great importance if you want to deploy supply from CDC to local hubs most optimally (based on 2 weeks out forecast version and not on 22 weeks out version).

I agree that such changes of forecast within lead time won’t help you to balance supply and demand on supplier lead time (and will add some nervousness to the forecast), but in case of risk pooling you can balance positive and negative forecast errors. You know that forecasts are always more accurate on short term and when aggregated up the hierarchy, considering that demand sensing makes much sense. Regarding question which forecast version is the original one… Original one is the one which on the supplier lead time “time fence”, namely the forecast version on which first commitments were made and $$$ invested in supply (out of this time fence forecast can be changed without any impact on supply chain if there are no other agreement with suppliers). But as forecasting/demand planning is a re-planning process (this is my personal standpoint) companies should save different versions of forecasts (lag versions) and apply accuracy measures to each of them to see how accuracy is improving with shorter lags.

Most important is to save forecasts as they were on decision points (if we take previous example this would be: on 22 weeks out time fence submit PP to supplier, and on 2 weeks out time fence replenish local hubs) and apply accuracy measures to these versions. Final KPI could be mix of both, with more weight on 22 weeks out version.I believe these are good practices that add value to overall supply chain planning.Please let me know what you think about it. Tibor

Tibor,

Well, you covered quite a lot of ground.

I will try to address all of your points. I don’t see how whether a company has a multi-echelon network that makes demand sensing valuable.

The question of demand sensing can be restricted to its benefit at a single product location. If it can’t show a benefit at a single product location, it won’t be useful on a larger network no matter how complex. On the second point, the fact that forecasts are more accurate in the short term is not an argument for demand sensing. Let us say that we have a lead time of 2 days. Then to me, changing the forecast less than 2 days out is demand sensing. 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 product, and therefore when the supply plan cannot respond. So no, just because you can improve a forecast accuracy is immaterial to whether you should improve a forecast that cannot be met. The last part of your response lost me. Demand planning can be changed up down and sideways…..up until it impinges on the supply planning lead times. After that point, it may not be changed because at that point the horse has left the stable. So, unfortunately, Tibor, I still don’t see any logic under any circumstance where demand sensing makes any sense and should be performed. (although I am still open to listening). However, the logic for demand sensing gets even less obvious once when considers the state of modern forecasting today. Companies can’t do the most elementary forecasting properly. I can do things on my laptop with a $3500 application that the largest companies with the largest IT spends cannot do.

One reason is IT has taken over the software selection process at many companies, and they don’t understand forecasting, and shut demand planners completely out of the process. If you choose a bad forecasting application, obviously you will forecast at a low level. I think the question needs to be raised if demand sensing, which does not have any logical support is really the best investment of forecasting resources when most companies can’t perform attribute-based forecasting, do not control for bias, and don’t know their pre-manually adjusted forecast accuracy versus the system generated forecast accuracy. I say, let’s conquer the baby steps first, and do things that decades of academic research support as being how to manage the forecast. If we can’t do that, we don’t have a very good platform for proposing new and unproven methods.

Secondly, demand sensing is inconsistent with the broad research on manual adjustments to forecasts. This research which has been compiled by J Scott Armstrong is that most manual changes to the forecast to not improve it and that the only positive correlation between manual adjustment is when high forecasts are brought down significantly. When a new method contradicts a large body of research, we have to sit back and take notice. Demand sensing requires enormous evidence to be taken seriously, and as of yet, I have not seen any evidence presented. Rather I am hearing a lot of claims by software vendors. I have a better reason for why demand sensing is becoming popular. Companies across the country generally don’t know how to forecast, yet have accuracy targets they must meet.

One of the biggest investments in time that companies make is faking their forecast accuracy. Demand sensing is a very convenient tool for changing the forecast at the last minute. The demand planning department will use a term like “demand sensing” to in effect fake out other departments that rely upon the forecast into telling them that they are using a legitimate technique to improve “forecast accuracy.” But it won’t work. Demand planning departments that lie to the other departments will eventually lose their credibility with these departments.

Tibor, your message promoted me to write another article on this topic. So even though we may disagree on this topic, I appreciate your comment. I am sure readers will as well.

Comment #2: Alessandro Paggiaro

I was looking for an objective opinion on Demand Sensing, and I found your article on scmfocus.com. My cultural background makes myself completely agree with you, but I know that several giant companies (like P&G, Unilever, Kimberly…) are using the demand sensing to improve short-term forecast. Thus my question is simply: why?
Improve short-term fcst (7-14 days) can help to save stock and money (in term of safety days) at Distribution Center level, but in theory this gap could be covered with an excellent S&OP process and with the hard work of the DP team (e.g working with APO alerts, improve estimation and correction for promotional activities, etc.).
What’s your comment about that?

Alessandro, I think there is a misunderstanding as to what forecasting is. Forecasting is producing value for the future. Short term changes to the forecast are unfortunately within the lead time of the product.

So in this sense, they are non-sensical. P&G, Unilever, etc.. many companies do many counterproductive things in supply chain planning. The fact a large company does something is not evidenced that it is correct to do. Again, I question whether there is a “short term” forecast.

Secondly, once the forecast is created, does the “short term” forecast become adjusted? So what is the original forecast error? There are so many areas where just “normal” forecasting can be improved, trying a concept which is not solidly based in anything, and seems to be a way of “gaming” the forecast is not the best use of time.

Would You Like to Comment and Have it Added to This Thread?

Just provide your comment in the chatbox in the lower left of this screen.