Should we have frozen forecasts? I’m not talking predicting cold weather or demand for French fries.  I’m talking about the concept of having a horizon for which forecast changes are not allowed.

From time-to-time I hear the suggestion that “we should freeze forecasts and not allow any changes within current month” (or some other period).  I advocate against this practice.

The reason usually given for not allowing forecast changes in near horizon is that production schedules are locked in. Having a Supply Plan that is locked in near-term horizon is quite necessary.  Plants need stable schedules for purposes such as procuring materials and scheduling labor.  Frozen and slush periods in the supply plan are necessary for stable production schedules.

Frozen Period – A time horizon for which changes in the Supply Plan are only made for emergencies. Emergency changes within the frozen period often require additional approval by high-level authority. The Frozen Period may be as little as one or two weeks, depending on factors such as local labor policies and material lead-times.

Slush Period – A time horizon for which automated software is not allowed to change the Supply Plan.  This allows supply planners to arrange production to meet scheduling constraints without the software refresh destroying their work. The Slush Period is typically a few weeks longer than the Frozen Period. The Slush Period should be at least enough time to allow planners to dial in one cycle of production. 

While slush and frozen periods are necessary for Supply Planning, they are detrimental to Demand Planning.  Demand Planning should always present the most current expectation of demand.  Supply Planning needs to know about significant increases or decreases in demand, even if there is not enough lead-time to make changes in production. 

Changes in demand during the Frozen Period for Supply Planning will impact the projected inventory at the end of the Frozen Period.  This often means that Supply Planners will need to adjust the plan in the Slush Period. But if demand changes are not allowed, Supply Planners will not have an accurate picture of projected inventory levels.

Increasing Demand

As an example, Supply Planning probably needs to know when an item that has been averaging 3,500 units per week suddenly has unexpected incremental demand that increases it to 5,500 cases per week.  Not allowing any changes in to forecasts in the next three weeks would mean that inventory at the end of week three would likely be about 6,000 units less than Supply Planning had projected.

Figure 1 demonstrates the impact on inventory of not changing the forecast.  Changing the forecast for weeks 1-3 gives one shock to the system when the forecast is increased. By not increasing the forecast for weeks 1-3, there will be repeated shocks to the system as actual sales get posted day-by-day in the first three weeks.

Figure 2 shows what supply plan may have looked like before the demand was increased.  Production minimum runs of 10,000 units every two to three weeks are sufficient to support demand.  The next production is planned for week 3.

Figure 3 shows what supply plan might looked like after the demand is increased, assuming production schedule is locked for the first three weeks. If the forecast had not been increased for the first three weeks, the expected shortage of 2,076 cases in week 2 would not be apparent. Updating the forecast gives supply chain planners the opportunity to see the resulting supply risk and consider mitigation strategies.

When to Change Forecasts

Rather than having a frozen period that prohibits any near-term changes to forecasts, I recommend considering whether the proposed change is significant. The Demand Plan for an item is usually a point estimate representing the mid-point or most likely value of a probabilistic variable.  Changing the demand plan for slight changes simply introduces noise into the system. 

It would not be challenging to create a statistical formula to determine the probability and expected value of supply risk for a given demand change (it would be a similar calculation to the one I developed for Expected Obsolescence). However, an experienced Demand Planner should have sufficient grasp of their products to understand when a proposed change in demand is within the range of normal variability and when it is a significant departure from what was previously forecasted.  What is significant depends item-by-item on factors such as inventory levels, forecast volatility, etc.

I do not like the concept of a frozen period in which no changes are allowed to the Demand Plan. Instead, I would agree with and support this statement: “we only allow changes to the Demand Plan within the near-horizon for changes that are significant.”