The 7 Habits of Highly Effective Forecasters

As a consequence of being in a forecasting consultancy we have picked up an in-depth understanding of the planning and forecasting functions of the firms. One conclusion we have drawn is that irrespective of distribution, dimension or division, most firms have similar struggles from similar function weaknesses. What we also noticed is that for demand planning and forecasting, the best practices are likewise impartial to size, sector and regional differences. This article lists out those practices or habits we find most effective when it comes to optimizing these functions.

1.Recognize the difference between the plan and forecast
The first and foremost of these habits is to separate the planning and forecasting functions. Very often companies lump the two together, while there is significant interdependence, they are inherently different processes. Demand Forecasting provides to the company an estimation of potential future demand, planning lays out the necessary steps to deliver that future demand. What is most often done is having a forecast built from the plan. Forecasting is a statistical task, planning is a resource based response to the same. To have the two contribute most effectively, segregation is a necessity.

2.Set up the forecasting process
The next best practice is to have a proper cyclic forecasting process in place, following which planning can be carried out. Our recommendation is Prepare, Execute, Evaluate, Correct and then Plan. The first is to understand needs and requirements, as well as the environment to create an initial forecast. Next is to execute by defining the ranges as dictated by the prepared forecast. This is then evaluated through comparison to the actual ranges. Lastly, corrections are made based on the results from evaluation to obtain a revised forecast. This is then presented as the final forecast for which planning is carried out.

3.Deploy statistical methods for baseline forecast
A third practice that helps firms best incorporate these functions is to have a wide range of statistical models at hand to shape out the forecasts. Very often, firms end up modelling their data on simplistic analytic techniques because Excel cannot handle more sophisticated methods. Consequently, their forecasting is nowhere a robust as it should be or could be had they an advanced forecasting system in place. Advanced systems employ a wide range of advanced statistical models and can often choose the best fit, resulting in a much more accurate forecast. 

4.Ensure continuous data cleansing
Next what we have noticed is that for a robust baseline forecast, two components are vital. One is data correction based on market and customer input and the other is data cleansing based on external events. As always, customer is King. A factor that can significantly increase the accuracy of forecasts is to involve what the customer, through market response analyses . A  forecast ignoring this input can easily be blindsided. Quick responsivity to demand shaping events is a huge asset. Likewise, properly carried out data cleansing also contributes to strengthening the forecast. Acknowledging and correcting outliers helps justify current performance as well as avoid undue influence on future forecasts.

5.Identify the user and purpose of forecast
Fifth amongst the habits is the creation of forecasts based on two parameters – who it is for and what purpose it is intended to serve. For instance, when marketing asks for a forecast, it is more likely what they need is a general idea of which products need to be promoted more whereas what logistics would need is a more accurate number for optimum distribution and warehousing. However, even for the same function then numbers change depending on intent. For instance, sales may need forecasts to determine the targets they need to meet. They may also need forecasts to decide what would be the right product mix. One assigns them parameters, the other, empowers them to focus their energy appropriately. If these attributes aren’t decided that forecast remains just a number, an uninsightful one at that.

6.Continuous improvement and not blame game
Up next is the practice to not blame forecasting for business woes. A forecast, at the end of the day is nothing more than an assessment. One that is decided by considering all the various inputs. If there is an error, what should be read is where was the information lacking. A review must be carried out of why the planner couldn’t focus and correctly forecast the influence of the element that had tripped them up. If it was a lack of information, steps should be carried out to prevent repetition. If it was an external event, its influence should be identified and corrected before the next forecast.

7.Collaborative team effort
The last of the 7 habits is to have proper communication channels in place. A forecast is a collaborative team effort, where everyone could and should pitch in. However more than merely changing  the forecast to reflect their needs, each contributor should provide an explanation for their carrying out of the same. This reasoning from across the board should ideally be visible to all involved parties, so that the resulting forecast is as robust as can be.

This is an inexhaustive list of what we consider best practices and recommend as best habits. Comment below if you think we have missed any out.


  1. SUSHIL KUMAR says:

    Hi Rishi
    Cud not have agreed more.

    1. rishi says:

      Thank you Sushil. Glad you find this information useful. For updates on our weekly demand planning resource article, follow our Linkedin page – and like us on Facebook

  2. Doris Choo says:

    Thank you for publishing this highly Effective Forecasters. Most informative and relevant. I read another article by Michael Gilliland at SAS Institute; ” Worse Practices ” that interestingly view your 7 habits by looking at it from another perspective – misunderstanding of the basics.

    1. Thank you for your comment. It is hard to argue with these tenets of effective forecasting; however, as I mentioned in my post to the article sadly once human greed and lack of teamwork, which endeavours to create win-win situations, get mixed into the equation, these effective practices are difficult in principle to achieve.

      We, at Anamind, welcome your comments and will respond to each one.

      Steve Miller

  3. This is a very well-written article and is very strategically sound. The unfortunate reality is that several of these best practices get derailed by petty corporate politics. If we want to truly speak about a one number plan, then we need to align performance incentives so that the organization focuses on ‘one source of truth’ and pulls on the tug-of-war rope IN THE SAME DIRECTION.

    Thank you for a good article that captures the essence of what could be. With 40+ years of experience I sadly look at it with a jaundiced eye and say, this could be done but will it be done.

    For what anyone thinks this opinionated comment is worth, the final winner of the competitive battle for survival is the organization that truly embraces the spirit of valuing all associates and their opinions as worth considering and getting away from trying to find fault with forecasts that can never be perfect. These organizations can and will utilize their intellectual capital in a positive way that reinforces the veracity of the plan and makes sure it is a stretch plan and it is met.

  4. Generally concur with your observations. But, in #3 the issue is not the spreadsheet paradigm of Excel that limits the use of ‘advanced’ techniques. In fact the PEERForecaster.xla Addin ( is fully capable of the advanced State Space forecasting methodologies (including ARIMA and family of trend/seasonal exponential smoothing models). The problem is that most demand planners do not realize that forecasting for the supply chain requires a relational database to adequately reconcile forecasts within a collaborative planning process.. A spreadsheet is a flat file (with lots of cell technology), but is not a database and cannot support the fast moving changes in ‘bigger data’ applications.

Leave a Comment

Your email address will not be published. Required fields are marked *