A forecast is a prediction about the value of a time series variable at some time in the future. For instance, one might want to estimate next month’s sales or demand for a product item. A time series is a sequence of numbers recorded at equally spaced time intervals; for example, unit sales recorded every month.

The objectives you pursue when you forecast depend on the nature of your job and your business. Every forecast is uncertain; in fact, there is a range of possible values for any variable you forecast. Values near the middle of this range have a higher likelihood of actually occurring, while values at the extremes of the range are less likely to occur. The following figure illustrates a typical distribution of forecast values.

Illustrating a forecast distribution of forecast values

Point forecasts

The most common use of forecasts is to estimate a sequence of numbers representing the most likely future values of the variable of interest. For instance, suppose you are developing a sales and marketing plan for your company. You may need to fill in 12 cells in a financial spreadsheet with estimates of your company’s total revenues over the next 12 months. Such estimates are called point forecasts because you want a single number (data point) for each forecast period. Smart Demand Planner’ Automatic forecasting feature provides you with these point forecasts automatically.

Interval forecasts

Although point forecasts are convenient, you will often benefit more from interval forecasts. Interval forecasts show the most likely range (interval) of values that might arise in the future. These are usually more useful than point forecasts because they convey the amount of uncertainty or risk involved in a forecast. The forecast interval percentage can be specified in the various forecasting dialog boxes in the Demand Planning Software.  Each of the many forecasting methods (automatic, moving average, exponential smoothing and so on) available in Smart Demand Planner allow you to set a forecast interval.

The default configuration in Smart Demand Planner provides 90% forecast intervals. Interpret these intervals as the range within which the actual values will fall 90% of the time. If the intervals are wide, then there is a great deal of uncertainty associated with the point forecasts. If the intervals are narrow, you can be more confident. If you are performing a planning function and want best case and worst case values for the variables of interest at several times in the future, you can use the upper and lower limits of the forecast intervals for that purpose, with the single point estimate providing the most likely value. In the previous figure, the 90% forecast interval extends from 3.36 to 6.64.

Upper percentiles

In inventory control, your goal may be to make good estimates of a high percentile of the demand for a product item. These estimates help you cope with the tradeoff between, on the one hand, minimizing the costs of holding and ordering stock, and, on the other hand, minimizing the number of lost or back-ordered sales due to a stock out. For this reason, you may wish to know the 99th percentile or service level of demand, since the chance of exceeding that level is only 1%.

When forecasting individual variables with features like Automatic forecasting, note that the upper limit of a 90% forecast interval represents the 95th percentile of the predicted distribution of the demand for that variable. (Subtracting the 5th percentile from the 95th percentile leaves an interval containing 95%-5% = 90% of the possible values.) This means you can estimate upper percentiles by changing the value of the forecast interval. In the figure, “Illustrating a forecast distribution”, the 95th percentile is 6.64.

To optimize stocking policies at the desired service level or to let the system recommend which stocking policy and service level generates the best return, consider using Smart Inventory Optimization.   It is designed to support what-if scenarios that show predicted tradeoffs of varying inventory polices including different service level targets.

Lower percentiles

Sometimes you may be concerned with the lower end of the predicted distribution for a variable. Such cases often arise in financial applications, where a low percentile of a revenue estimate represents a contingency requiring financial reserves. You can use Smart Demand Planner in this case in a way analogous to the case of forecasting upper percentiles. In the figure, “Illustrating a forecast distribution” , the 5th percentile is 3.36.

In conclusion, forecasting involves predicting future values, with point forecasts offering single estimates and interval forecasts providing likely value ranges. Smart Demand Planner automates point forecasts and allows users to set intervals, aiding in uncertainty assessment. For inventory control, the tool facilitates understanding upper (e.g., 99th percentile) and lower (e.g., 5th percentile) percentiles. To optimize stocking policies and service levels, Smart Inventory Optimization supports what-if scenarios, ensuring effective decision-making on how much to stock given the risk of stock out you are willing to accept.

Our customers have usually settled into one way to manage their service parts inventory. The professor in me would like to think that the chosen inventory policy was a reasoned choice among considered alternatives, but more likely it just sort of happened. Maybe the inventory honcho from long ago had a favorite and that choice stuck. Maybe somebody used an EAM or ERP system that offered only one choice. Perhaps there were some guesses made, based on the conditions at the time.

The Competitors

Too seldom, businesses make these choices in haphazard ways. But modern service parts planning software lets you be more systematic about your choices. This post demonstrates that proposition by making objective comparisons among three popular inventory policies:  Order Up To, Reorder Point/Order Quantity, and Min/Max.  I discussed each of these policies in this video blog.

• Order Up To. This is a periodic review policy where every T days, on-hand inventory is tallied and an order of random size is placed to bring the stock level back up to S units.
• Q, R or Reorder Point/Order Quantity. Q, R is a continuous review policy where every day, inventory is tallied. If there are Q or fewer units on hand, an order of fixed size is placed for R more units.
• Min, Max is another continuous review policy where every day, inventory is tallied. If there are Min or fewer units on hand, an order is placed to bring the stock level back up to Max units.

Inventory theory says these choices are listed in increasing order of effectiveness. The first option, Order Up To, is clearly the simplest and cheapest to implement, but it closes its eyes to what’s going on for long periods of time.  Imposing a specified passage of time in between orders makes it, in theory, less flexible. In contrast, the two continuous review options keep an eye on what’s happening all the time, so they can react to potential stockouts quicker. The Min/Max option is, in theory, more flexible than the option that uses a fixed reorder quantity because the size of the order dynamically changes to accommodate the demand.

That’s the theory. This post examines evidence from head-to-head comparisons to check the theory and put concrete numbers on the relative performance of the three policies.

The Meaning of “Best”

How should we keep score in this tournament? If you are a regular reader of this Smart Forecaster blog, you know that the core of inventory planning is a tug-of-war between two opposing objectives: keeping inventory lean vs keeping item availability metrics such as service level high.

To simplify things, we will compute “one number to rule them all”: the average operating cost. The winning policy will be the one with the lowest average.

This average is the sum of three components: the cost of holding inventory (“holding cost”), the cost of ordering replenishment units (“ordering cost”) and the cost of losing a sale (“shortage cost”). To make things concrete, we used the following assumptions:

• Each service part is valued at $1,000.
• Annual holding cost is 10% of item value, or $100 per year per unit.
• Processing each replenishment order costs $20 per order.
• Each unit demanded but not provided costs the value of the part, $1,000.

For simplicity, we will refer to the average operating cost as simply “the cost”.

Of course, the lowest average cost can be achieved by getting out of the business. So the competition required a performance constraint on item availability: Each option had to achieve a fill rate of at least 99%.

The Alternatives Duke it Out

A key element of context is whether stockouts result in losses or backorders. Assuming that the service part in question is critical, we assumed that unfilled orders are lost, which means that a competitor fills the order. In an MRO environment, this will mean additional downtime due to stockout.

To compare the alternatives, we used our predictive modeling engine to run a large number of Monte Carlo simulations.  Each simulation involved specifying the parameter values of each policy (e.g., Min and Max values), generating a demand scenario, feeding that into the logic of the policy, and measuring the resulting cost averaged over 365 days of operation. Repeating this process 1,000 times and averaging the 1,000 resulting costs gave the final result for each policy.  

To make the comparison fair, each alternative had to be designed for its best performance. So we searched the “design space” of each policy to find the design with the lowest cost. This required repeating the process described in the previous paragraph for many pairs of parameter values and identifying the pair yielding the lost average annual operating cost.

Using the algorithms in Smart Inventory Optimization (SIO^TM) we made head-to-head-to-head comparisons under the following assumptions about demand and supply:

• Item demand was assumed to be intermittent and highly variable but relatively simple in that there was neither trend nor seasonality, as is often true for service parts. Daily mean demand was 5 units with a large standard deviation of 13 units. Figure 1 shows a sample of one year’s demand. We have chosen a very challenging demand pattern, in which some days have 10 to even 20 times the average demand.

Figure 1: Daily part demand was assumed to be intermittent and very spikey.

​​

• Suppliers’ replenishment lead times were 14 days 75% of the time and 21 days otherwise. This reflects the fact that there is always uncertainty in the supply chain.

And the Winner Is…

Was the theory right? Kinda’ sorta’.

Table 1 shows the results of the simulation experiments. For each of the three competing policies, it shows the average annual operating cost, the margin of error (technically, an approximate 95% confidence interval for the mean cost), and the apparent best choices for parameter values.

Table 1: Results of the simulated comparisons

For example, the average cost for the (T,S) policy when T is fixed at 30 days was $41,680. But the Plus/Minus implies that the results are compatible with a “true” cost (i.e., the estimate from an infinite number of simulations) of anywhere between $39,890 and $43,650. The reason there is so much statistical uncertainty is the extremely spikey nature of demand in this example.

Table 1 says that, in this example, the three policies fall in line with expectations. However, more useful conclusions would be:

1. The three policies are remarkably similar in average cost. By clever choice of parameter values, one can get good results out of any of the three policies.
2. Not shown in Table 1, but clear from the detailed simulation results, is that poor choices for parameter values can be disastrous for any policy.
3. It is worth noting that the periodic review (T,S) policy was not allowed to optimize over possible values of T. We fixed T at 30 to mimic what is common in practice, but those who use the periodic review policy should consider other review periods. An additional experiment fixed the review period at T = 7 days. The average cost in this scenario was minimized at $36,551 ± $1,668 with S = 343. This result is better than that using T = 30 days.
4. We should be careful about over-generalizing these results. They depend on the assumed values of the three cost parameters (holding, ordering and shortage) and the character of the demand process.
5. It is possible to run experiments like those shown here automatically in Smart Inventory Optimization. This means that you too would be able to explore design choices in a rigorous way.

In almost every business and industry, decision-makers need reliable forecasts of critical variables, such as sales, revenues, product demand, inventory levels, market share, expenses, and industry trends.

Many kinds of people make these forecasts. Some are sophisticated technical analysts, such as business economists and statisticians. Many others regard forecasting as an important part of their overall work: general managers, production planners, inventory control specialists, financial analysts, strategic planners, market researchers, and product and sales managers. Still, others seldom think of themselves as forecasters but often have to make forecasts on an intuitive, judgmental basis.

Because of the way we designed Smart Demand Planner, it has something to offer all types of forecasters. This design grows out of several observations about the forecasting process. Because we designed Smart Demand Planner with these observations in mind, we believe it has a style and content uniquely suited for turning your browser into an effective forecasting and planning tool:

Forecasting is an art that requires a mix of professional judgment and objective, statistical analysis.

It is often effective to begin with an objective statistical forecast that automatically accounts for trends, seasonality, and other patterns.  Then, apply adjustments or forecast overrides based on your business judgment. Smart Demand Planner makes it easy to execute graphical and tabular adjustments to statistical forecasts.

The forecasting process is usually iterative.

You will likely decide to make several refinements of your initial forecast before you are satisfied. You may want to exclude older historical data that you find to no longer be relevant.  You could apply different weights to the forecast model that put varying emphases on the most recent data. You could apply trend dampening to increase or decrease aggressively trending statistical forecasts.  You could allow the Machine Learning models to fine-tune the forecast selection for you and select the winning model automatically.  Smart Demand Planner’s processing speed gives you plenty of time to make several passes and saves multiple versions of the forecasts as “snapshots” so you can compare forecast accuracy later.

Forecasting requires graphical support.

The patterns evident in data can be seen by a discerning eye. The credibility of your forecasts will often depend heavily on graphical comparisons other business stakeholders make when they assess the historical data and forecasts. Smart Demand Planner provides graphical displays of forecasts, history, and forecast vs. actuals reporting.

Forecasts are never exactly correct.

Because some error always creeps into even the best forecasting process, one of the most useful supplements to a forecast is an honest estimate of its margin of error.

Smart Demand Planner presents both graphical and tabular summaries of forecast accuracy based on the acid test of predicting data held back from development of the forecasting model. 

Forecast intervals or confidence intervals are also very useful.  They detail the likely range of possible demand that is expected to occur.  For example, if actual demand falls outside of the 90% confidence interval more than 10% of the time then there is reason to investigate further.  

Forecasting requires a match of method to data.

One of the major technical tasks in forecasting is to match the choice of forecasting technique to the nature of the data. Features of a data series like trend, seasonality or abrupt shifts in level suggest certain techniques instead of others.

Smart Demand Planner’ Automatic forecasting feature makes this match quickly, accurately and automatically.

Forecasting is often a part of a larger process of planning or control.

For example, forecasting can be a powerful complement to spreadsheet-based financial analysis, extending rows of figures off into the future. In addition, accurate sales and product demand forecasts are fundamental inputs to a manufacturer’s production planning and inventory control processes. An objective statistical forecast of future sales will always help identify when the budget (or sales plan) may be too unrealistic. Gap analysis enables the business to take corrective action to their demand and marketing plans to ensure they do not miss the budgeted plan.

Forecasts need to be integrated into ERP systems
Smart Demand Planner can quickly and easily transfer its results to other applications, such as spreadsheets, databases and planning systems including ERP applications.  Users are able to export forecasts in a variety of file formats either via download or to secure FTP file locations.  Smart Demand Planner includes API based integrations to a variety of ERP and EAM systems including Epicor Kinetic and Epicor Prophet 21, Sage X3 and Sage 300, Oracle NetSuite, and each of Microsoft’s Dynamics 365 ERP systems. API based integrations enable customers to push forecast results directly back to the ERP system on demand.

The result is more efficient sales planning, budgeting, production scheduling, ordering, and inventory planning.

​In a highly configurable manufacturing environment, forecasting finished goods can become a complex and daunting task. The number of possible finished products will skyrocket when many components are interchangeable. A traditional MRP would force us to forecast every single finished product which can be unrealistic or even impossible. Several leading ERP solutions introduce the concept of the “Planning BOM”, which allows the use of forecasts at a higher level in the manufacturing process. In this article, we will discuss this functionality in ERP, and how you can take advantage of it with Smart Inventory Planning and Optimization (Smart IP&O) to get ahead of your demand in the face of this complexity.

Why Would I Need a Planning BOM?

Traditionally, each finished product or SKU would have a rigidly defined bill of materials. If we stock that product and want to plan around forecasted demand, we would forecast demand for those products and then feed MRP to blow this forecasted demand from the finished good level down to its components via the BOM.

Many companies, however, offer highly configurable products where customers can select options on the product they are buying. As an example, recall the last time you bought a personal computer. You chose a brand and model, but from there, you were likely presented with options: what speed of CPU do you want? How much RAM do you want? What kind of hard drive and how much space? If that business wants to have these computers ready and available to ship to you in a reasonable time, suddenly they are no longer just anticipating demand for that model—they must forecast that model for every type of CPU, for all quantities of RAM, for all types of hard drive, and all possible combinations of those as well! For some manufacturers, these configurations can compound to hundreds or thousands of possible finished good permutations.

There may be so many possible customizations that the demand at the finished product level is completely unforecastable in a traditional sense. Thousands of those computers may sell every year, but for each possible configuration, the demand may be extremely low and sporadic—perhaps certain combinations sell once and never again.

This often forces these companies to plan reorder points and safety stock levels mostly at the component level, while largely reacting to firm demand at the finished good level via MRP. While this is a valid approach, it lacks a systematic way to leverage forecasts that may account for anticipated future activity such as promotions, upcoming projects, or sales opportunities. Forecasting at the “configured” level is effectively impossible, and trying to weave in these forecast assumptions at the component level isn’t feasible either.

Planning BOM Explained

This is where Planning BOMs come in. Perhaps the sales team is working a big b2b opportunity for that model, or there’s a planned promotion for Cyber Monday. While trying to work in those assumptions for every possible configuration isn’t realistic, doing it at the model level is totally doable—and tremendously valuable.

The Planning BOM can use a forecast at a higher level and then blow demand down based on predefined proportions for its possible components. For example, the computer manufacturer may know that most people opt for 16GB of RAM, and far fewer opt for the upgrades to 32 or 64. The planning BOM allows the organization to (for example) blow 60% of the demand down to the 16GB option, 30% to the 32GB option, and 10% to the 64GB option. They could do the same for CPUs, hard drives, or any other customizations available.  

The business can now focus their forecast at this model level, leaving the Planning BOM to figure out the component mix. Clearly, defining these proportions requires some thought, but Planning BOMs effectively allow businesses to forecast what would otherwise be unforecastable.

The Importance of a Good Forecast

Of course, we still need a good forecast to load into an ERP system. As explained in this article, while ERP  can import a forecast, it often cannot generate one and when it does it tends to require a great deal of hard to use configurations that don’t often get revisited resulting in inaccurate forecasts.  It is therefore up to the business to come up with their own sets of forecasts, often manually produced in Excel. Forecasting manually generally presents a number of challenges, including but not limited to:

• The inability to identify demand patterns like seasonality or trend
• Overreliance on customer or sales forecasts
• Lack of accuracy or performance tracking

No matter how well configured the MRP is with your carefully considered Planning BOMs, a poor forecast means poor MRP output and mistrust in the system—garbage in, garbage out. Continuing along with the “computer company” example, without a systematic way of capturing key demand patterns and/or domain knowledge in the forecast, MRP can never see it.

Extend ERP  with Smart IP&O

Smart IP&O is designed to extend your ERP system with a number of integrated demand planning and inventory optimization solutions. For example, it can generate statistical forecasts automatically for large numbers of items, allows for intuitive forecast adjustments, tracks forecast accuracy, and ultimately allows you to generate true consensus-based forecasts to better anticipate the needs of your customers.

Thanks to highly flexible product hierarchies, Smart IP&O is perfectly suited to forecasting at the Planning BOM level so you can capture key patterns and incorporate business knowledge at the levels that matter most. Furthermore you can analyze and deploy optimal safety stock levels at any level of your BOM.

Those of you who track hot topics will be familiar with the term “digital twin.” Those who have been too busy with work may want to read on and catch up.

What is a digital twin?

While there are several definitions of digital twin, here’s one that works well:

A digital twin is a dynamic virtual copy of a physical asset, process, system, or environment that looks like and behaves identically to its real-world counterpart. A digital twin ingests data and replicates processes so you can predict possible performance outcomes and issues that the real-world product might undergo. [Source: Unity.com]. For additional background, you might go to Mckinsey.com.

What is the difference between a digital twin (hereafter DT) and a model? Primarily, a DT gets connected to real-time data to maintain the model as an up-to-the-minute representation of the system you are working with.

Our current products might be called “slow-motion DT’s” because they are usually used with non-real-time data (though not stale data, since it is updated overnight) and applied to problems like planning the next quarter’s raw material buys or setting inventory parameters for a month or longer.

Are people using digital twins in my industry?

My impression is that the penetration of DT’s may be highest in the aerospace and nuclear industries. Most of our customers are elsewhere: in manufacturing, distribution, and public utilities such as transportation and power. Soon we’ll be offering new products that come closer to the strict definition of a DT that is connected intimately to the system it represents.

DT Preview

Most users of Smart Inventory Optimization™ (SIO) run the application periodically, typically monthly. SIO analyzes current demand for inventory items and recent supplier lead times, converting these into demand and supply scenarios, respectively. Then users either interactively (for individual items) or automatically (at scale) set inventory control parameters that will provide the long-term average performance they want, balancing the competing goals of minimizing inventory while guaranteeing a sufficient level of item availability.

Smart Supply Planner (SSP) operates in a more immediate way to react to contingencies. Any day could bring an anomalous order that spikes up demand, such as when a new customer places a surprising initial stocking order. Or a key supplier could experience a problem at its factory and be forced to delay shipment of your planned replenishment orders. In the long run, these contingencies average out and justify the recommendations coming out of SIO. However, SSP will give you a way to react in the short run to seize opportunities or dodge bullets.

At its core, SSP operates like SIO in that it is scenario driven. The differences are that it uses short planning horizons and uses real-time initial conditions as the basis for its simulations of inventory system performance. Then it will provide real-time recommendations for interventions that offset the disruption caused by the contingencies. These would include cancelling or expediting replenishment orders.

Summary

Digital twins let you try out plans “in silico” before you implement them in the factory or warehouse. At their core are mathematical models of your operation but connected to real-time data. They provide a “digital sandbox” in which you can try out ideas and get immediate predictions of how well they will work. Much more than a spreadsheet, DT’s will soon be the key tool in your inventory planning toolbox.