This is the first article in a series on measuring demand variation.
Quantifying demand variation is definitely one of the more contentious topics among inventory managers and business analysts. It makes sense. After all, inaccurately measuring the variation of demand has real business consequences such as tied-up cash, delivery failure, or even both!
Many manufacturers estimate demand variation rather than measure it. Unfortunately, there’s a number of additional consequences that come with estimating demand variation. Of course, first, there is the real risk of overestimating or underestimating the necessary amount of safety stock as a result of emotion-driven “intuition”. Additionally, even if the estimate is exactly correct, it is not repeatable, teachable, or verifiable. Getting it right is akin to pinning the tail on the donkey. It’s a matter of luck. Sooner or later you’re going to miss by a mile.
While, measuring demand variation can seem daunting, it can and should be done utilizing a reliable, standardized approach.
3 Types of Demand Variation
Demand variation can get complex. Before diving in, first, it’s good to familiarize ourselves with the 3 key demand variation categories. Not only can demand vary from day to day, but average daily demand can also vary over time. Any combination of various micro and macro-economic forces can drive changes in average daily demand. Seasonal effects on demand, can further complicate matters.
- Daily Demand Volatility: Demand variation from one day to another. This daily volatility often seems, and typically is somewhat random.
- Daily Demand Shifts: Changes in average daily demand experiences changes. Daily demand shifts can sometimes be predicted because they are caused by some observable – but often uncontrollable – factors, such as consumer spending habits.
- Seasonality: Regular demand shifts positively correlated with routine social or environmental events including, but not limited to holidays, industry planning periods, and government budgeting cycles. Demand is significantly, but routinely higher or lower during these periods.
Demand patterns are graphed in a similar fashion to how we graph sawtooth curves. Rather than graph every unique order quantity, we aggregate demand within distinct, recurring, larger units of time. The time period might be a single day or even a week. So, to graph demand, we simply calculate the total of all unique demand signals within each defined, uniform time period. The total demand for each period will be a discrete data point on the graph. Graphing this way, marginalizes the minutia that might otherwise sabotage analytical proficiency at gaining valuable insights.
Graphing real demand can be invaluable to gaining operational insights, but only if we understand how demand patterns impact visual analysis, when graphed. Consider the four graphs in Figure 1. Each graph, accounts for 4 weeks of demand with an identical daily average of 500 units and monthly average of 10,000 units. The yellow area accounts for average daily demand, the blue line is actual daily demand, and the red line is average weekly demand for a 5-day work week running from Monday to Friday.
- Demand Pattern 1. Chart 1 appears fairly stable. There is little daily or weekly variation.
- Demand Pattern 2. In contrast to Chart 1, Chart 2 is very volatile. Data points tend to be much higher and much lower than the demand pattern the first chart, and the difference in demand is greater from day to day and week to week.
- Demand Pattern 3. Chart 3 illustrates a scenario where there is high daily variation, while weekly variation is zero and predictable at 2,500 units per week. In this example, all demand is concentrated on Wednesday, while the remaining work days have no demand.
- Demand Pattern 4. Chart 4 illustrates a situation where the entire month’s demand of 10,000 units is concentrated in a single day, with zero demand on the remaining 19 work days. Considering the stark difference between the first day (and week) with the second day (and week), you could justifiably say there is high variation. Yet, the subsequent days (and weeks) have no variation.
If it’s not readily apparent, while graphing demand over the appropriate time horizon is much better at accurately illustrating meaningful variation, in most situations, it simply isn’t feasible to plot and visually analyze a graph for every individual inventory item. That’s okay. Again, just as with inventory sawtooth curves, a proven process and established formulas will help to overcome this analytical challenge.
Impact of Internal Chaos
Most of the time, demand variation is the result of external market factors, such as competitor and customer behavior. However, a significant degree of variation can also be the result of poor internal planning, coordinating, and scheduling. Be vigilant in identifying and putting an end to poor internal practice and performance.
That said, if demand really is extremely volatile, Heijunka (level-loading) and kanban are extremely simple and powerful lean tools for stabilizing internal production quantities even when the external environment is extremely unruly.
In part 2 of this series on demand variation, we learn the specific steps in how to calculate demand variation as well as explore what to do when you lack data, and simply can’t quantify demand.