The Infosys global supply chain management blog enables leaner supply chains through process and IT related interventions. Discuss the latest trends and solutions across the supply chain management landscape.

« Designing optimal customer order allocation by taking into account of inventory fluctuations can be quite complex.... | Main | Reduce your TCO with migration to SRM MDM Catalog »

What’s the right forecasting approach in the current business environment?

I am talking about short term operational forecasting here, since that’s the one that drives business on an ongoing basis and has a direct impact on the financial performance of an organization. Forecasting has been a very statistically driven exercise with too much weightage given to quantitative techniques. Quantitative techniques take historical data as the reference or basis for forecasting, and therefore, the results are acceptable as long as the industry and other players work in similar economic environmental conditions. Now, when the consumer demand is extremely volatile and sentiments are down across almost all the industry segments, the historical consumer sales data cannot be considered as the right reference for generating forecasts. And therefore, dependence on quantitative techniques may not be the right approach and might lead to serious consequences such as rise in excess and obsolete inventory, increase in blocked working capital and finally, shortfall in achieving sales targets.

It is a known fact that the recession and slowdown in economy has impacted organizations on both their topline and bottomline, leaving no room for error in any process. Since the tolerance levels for absorbing inefficiencies are at its minimum, organizations are doing a critical health-check by identifying non-value adding processes & cost elements and eliminating them as far as possible. In such an environment, it is highly imperative to re-look at one’s forecasting approach and adopt techniques that are more appropriate, implementable, and let one stay conservative.

In my opinion, I feel that quantitative techniques should be used to generate just a baseline forecast that should get further fine tuned and refined to make it a final forecast, and this final forecast should be used to drive the entire planning process. Doing this at a SKU level may not be feasible, but demand planners should definitely attempt to do it at a product category level and then break the aggregated numbers to SKU level with right business rules. Now the most important part is the fine-tuning of this base forecast, where I feel that demand planners should adopt a two-pronged collaborative forecasting approach and work very closely with:

a)      “Internal stakeholders such as marketing and sales teams”. The marketing and sales teams play a very crucial role in boosting sales in a downturn economy, by creating innovative strategies to drive channel push and consumer pull. Usually termed as ‘demand shaping’, its end objective is to drive consumer demand and preferably, organizations focus their efforts to just key brands and geographies. Demand planners should work with these teams to gather inputs and insights about such programs and get an estimate of expected impact on consumer demand. The final forecast can then be prepared by fine tuning base forecast based on these critical inputs which otherwise would have been lost, if there was no such collaboration process in place.

b)      “External supply chain team such as distributors and channel partners”. The other critical piece is working with channel partners and distributors, atleast, the A-class partners and geographies and receiving consumer preferences and signals as early as possible. Termed as ‘demand sensing’, its end objective is to be as close to the consumer as possible, so that any demand signal can be quickly picked up and made visible to all members back in the supply chain. Demand planners should work with these partners through periodical consensus planning meetings and get involved in their forecasting process, help them in improving it and use it as the basis for fine tuning its own generated base forecast. Again, let me emphasize here that, it is nearly impossible to capture this precious information through the historical data an organization has. Having very close and continuous interactions with channel partners is a very significant step towards getting the forecast numbers right.


Now, having said that, it doesn’t mean that it will result in a perfect forecast but yes, definitely, it will have all the supply chain members signed in for the final forecast number that drives all the other planning processes. The risks are shared and organizations will take less time to react to any untoward instances especially, when we all are in the middle of this crisis situation. Please do share your comments and views on this; I am willing to hear your experiences and feedback.


Shaping and Sensing will definitely help towards improving forecasts. But what I have observed is that the number of demand planners is so limited in many of the organizations and the forecasts make and review meetings are so many that they barely have time to actually put their minds to it or make the effort to reach out to everyone who can add value to the forecasts.

While improving forecasts should always be an aim, the organization should remember that -
a. Forecasts are always wrong (you can only try to reduce their error)
b. Longer the horizon of the forecasts, the worse would be its accuracy

The only real solution to improving the forecasts is to try and eliminate them through crunching lead times.

Nice write-up.
In my opinion, I feel historical data is still relevant under this situation but probably dependence on the immediate past (say past 7 days) is more reliable than the usual data set consideration. Probably the forecast should react more to the immediate past than the older data.
Also one more option could be what you had mentioned. Use the history to derive the baseline sales while using external factors (negative) to adjust for the fall in sales. This negative factor could be derived as an index of some generic economic indicators (sensex, inflation, etc.).
Another consideration could be on change of the order calculation methods, if the earlier calculation method was to meet the sales and safety stock, under slow moving conditions, it could be just the presentation levels at the shelves that should drive the ordering.
The two options you’ve mentioned are very much valid but the only consideration would be how much quantifiable reliable data you can obtain from these sources.

Got your point. Could you please explain the last point that you have made. What do you mean by "improving the forecasts is to try and eliminate them through crunching lead times."

I do agree that forecasts have to depend on historical data which shall provide us the trend for reference. Besides, a good model shall best depict the forecasts accurately. Nowadays, the business situation is changing rapidly, the ability to run a forecast quickly is of importance.

Sophisticated statistical forecasting techniques (such as MLR) can incorporate macro-economic factors (such as GDP, price indices) into the forecast.

Such concepts are used in certain industries (e.g. Metals, Energy) but for most products statistical forecasts are nothing more than a starting point and need to go through several cycles of adjustment before going into a planning process or system.

I have a few comments about the original post as well some of the following on entries.

I do agree overall with the approaches recommended by Aatish. I think to sum it up, demand planners need to ensure that they are not relying on simply one input (the quantitative forecast driven from demand history) but rather to consider (and challenge) inputs from multiple providers including sales, marketing, product management and key customers in order to make more informed decisions. Sales forecasts tend to be more accurate in the short term while marketing forecasts tend to be more accurate further out along the curve. Working with key customers on collaborative forecasting programs can help improve the quality of the forecast because you are getting closer to the ultimate source of the demand. From experience, it’s also a great way to improve overall customer relationships because it’s often the case that the customer sees this as a value-added service that can distinguish you from other vendors.

The demand planner is the individual in the organization that is responsible for providing sober, second-thought to all of the inputs gather during the forecasting process. They typically are the only role that brings a historical perspective to the process because everyone else either has a short memory or would prefer to forget past predictions. How often has this customer backed out at the last minute or overstated the business that they thought they were going to do? What happened the last three times we ran this kind of promotion in the spring? How quickly did demand materialize the last time we introduced a new model of this product? These are all the classic types of questions that, if given the opportunity, the demand planner needs to ask when driving to a consensus of future demand and that’s just to get to ‘square-one’ i.e. what is demand going to be given what sales, marketing and customers are planning to do. The next step is to then use that information to try and shape demand in order to align it with what the business wants to do (represented by high-level management forecasts propagated down to the SKU or product level). This might mean shifting marketing focus to higher-margin products or regions, implementing specific add-on or upgrade incentives with key customers, or adding key features to a new product.

Just two final points. Aatish started off his post by saying ‘Now, when the consumer demand is extremely volatile and sentiments are down across almost all the industry segments…'. I would argue that the comment about extreme volatility has been the norm for quite a while now at least in high-tech and consumer electronics segments. The approaches being suggested are really becoming the minimum mandatory for companies that want to survive the operational and margin pressures faced by a lot of today manufacturers. The other key point is related to Vikram’s comment about demand planners not having the time to attend review meetings. I believe this is a symptom of the process and tools used to support the demand planning process (i.e. planners are spending too much time collecting and rationalizing data). There are tools available now that provide the ability for all participants in the process including sales, marketing, finance and even customers to simultaneously input their information into a single data model that is ‘pre-rationalized’ so that demand planners can spend more of their time actually planning.

Well said Bob. I quite agree with you. Another important point I would like to highlight is to correct the sales data during every planning cycle. When the business will come back to its usual pace, the demand planners would be able to use this corrected sales data for forecasting. Otherwise, if they miss to correct the data, it will result in wrong output. It is not easy to be a demand planner!

I agree with the comments stated here. A couple items pop to mind as I read this.
1- Know your market and your products. They react differently than other products. Don't assume correlations that aren't there.
2- Don't assume the economy is the blame for all volatility. Make sure that you understand volatility or changes to demand and the underlying cause.
3- Demand planners that don't have time to review cannot incorporate the latest information into forecasts. If time is scarce, figure out ways to get what you need from those closest to the demand.
4- Don't assume demand from distributors/wholesalers is the same as "true" or underlying demand. Distributors increase/decrease inventory for a variety of reasons. While this should be part of your forecasting process already, be aware that the economy impacts them, too.

I would like to share my forecasting and demand planning knowledge through this forum. In general, in most of the MNCs, the forecast is done at value and volume levels. Forecasting responsibility generally lies with Marketing and Sales Department. Initially, corporate office fixes the value target for the year based on last year sales value with expected percentage growth. The top management always expects to meet or exceed the value target. The value target has broken down to volume (brand wise) using the latest yield and expected growth in a particular brand. Brand volume is further broken into SKU wise through top down approach based on historical data. The trend in historical data will help to project month wise data. This will become the target or plan for the year.
Earlier forecasting techniques were based on historical trends as the consumer’s behavior was predictable. In recent trends, the consumer behavior is erratic and unpredictable, the dependency on historical data and trends has reduced to some extent and market dynamics play an important role in formalizing the forecast number. However, the historical data is used for base line data projection and this will guide the marketing and sales team to add value by incorporating their market insight, seasonality, competition, promotion, price change and other parameters.

The marketing / sales team forecast is for short period (ie 3 months) and these figures are discussed in forum on monthly basis where sales, marketing, finance, production and purchase participate to decide on forecast volume and their roles are given below.
 Marketing and sales will take the responsibilities on distribution of products through channels and marketing strategies like promotion, media plan etc.
 Finance team calculates the value based on forecast figures and analyzes the excess or shortfall of revenue. If finance team realizes the higher shortfall of forecast value against the target plan, the same will be highlighted to top management. Marketing and sales in consultation with top management will decide to run a special promotion to promote the sales, increase the prices of the product, make up the loss against another brand or plan new product launch to make up the financial loss.
 Production team ensures the availability of required stocks at right time from manufacturing site to distribution centre.
 Purchase department is responsible for availability of raw materials and promotional items in time to production department.

Forecasting is done through the package which uses various statistical tools like Moving Average, Weighted Moving Average, ARIMA (Auto Regressive Integrated Moving Average), Winter, Holtz methods.
Demand shaping is done through internal marketing strategy whereas demand sensing is through obtaining data and demand signals from external customers like distributors, retailers. Hence, analyzing demand sensing data is more critical before using it in forecast. For example, obtaining forecast volume from distributors and reaches marketing through sales team, might be distorted due to Bullwhip effect.

To understand the latest market dynamics, the company is relying on CRM module to capture sales volume and other data from the distributors. Here IT plays a major role in integrating the data from distributors to the ERP system. This helps marketing and sales team to analyze the sales trend and proactively take the appropriate decision while forecasting the volume.

The demand planner role varies as per company strategy. In general, forecasting is done by sales and marketing. Demand planner will use the forecasting data at SKU / Distribution centre level and compute net requirement after considering opening stock, stocks in transit. This is called as DRP (Distribution Requirement Plan). Demand Planner will meet the supply team and agree upon final dispatch plan from factory to distribution centre. This is called as Agreed Supply Plan (ASP). The demand planner will also monitor the performance of forecast accuracy by comparing against actual and high variations are highlighted to stakeholder for proper control. Also, demand planner measures the performance of the supply team by comparing the actual receipt of stocks against the agreed quantity and time. This enables demand planner to exercise proper control over the marketing and supply team.

I think what Vikram meant is that forecasts and their natural errors are less important when you have the ability to reduce the leadtime and use this flexibility. The shorter the leadtime, the easier you can react on demand fluctuations - within certain limits, of course.
I have implemented a two way system: on the mid-term horizon, statitical forecasting with sales data as basis, on a short-term horizon though the development of the order book. That way, you can plan and reserve capacities, but at the same time have some flexibility when it comes to calling off from supplier/plant.

Post a comment

(If you haven't left a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Thanks for waiting.)

Please key in the two words you see in the box to validate your identity as an authentic user and reduce spam.

Subscribe to this blog's feed

Follow us on

Blogger Profiles

Infosys on Twitter