How long should it take for a demand forecast to be computed using statistical methods?  This question is often asked by customers and prospects.  The answer truly depends.  Forecast results for a single item can be computed in the blink of an eye, in as little as a few hundredths of a second, but sometimes they may require as much as five seconds.  To understand the differences, it’s important to understand that there is more involved than grinding through the forecast arithmetic itself.   Here are six factors that influence the speed of your forecast engine.

1) Forecasting method.  Traditional time-series extrapolative techniques (such as exponential smoothing and moving average methods), when cleverly coded, are lighting fast.  For example, the Smart Forecast automatic forecasting engine that leverages these techniques and powers our demand planning and inventory optimization software can crank out statistical forecasts on 1,000 items in 1 second!  Extrapolative methods produce an expected forecast and a summary measure of forecast uncertainty. However, more complex models in our platform that generate probabilistic demand scenarios take much longer given the same computing resources.  This is partly because they create a much larger volume of output, usually thousands of plausible future demand sequences. More time, yes, but not time wasted, since these results are much more complete and form the basis for downstream optimization of inventory control parameters.

2) Computing resources.  The more resources you throw at the computation, the faster it will be.  However, resources cost money and it may not be economical to invest in these resources.  For example, to make certain types of machine learning-based forecasts work, the system will need to multi-thread computations across multiple servers to deliver results quickly.  So, make sure you understand the assumed compute resources and associated costs. Our computations happen on the Amazon Web Services cloud, so it is possible to pay for a great deal of parallel computation if desired.

3) Number of time-series.  Do you have to forecast only a few hundred items in a single location or many thousands of items across dozens of locations?  The greater the number of SKU x Location combinations, the greater the time required.  However, it is possible to trim the time to get demand forecasts by better demand classification.  For example, it is not important to forecast every single SKU x Location combination. Modern Demand Planning Software can first subset the data based on volume/frequency classifications before running the forecast engine.  We’ve observed situations where over one million SKU x Location combinations existed, but only ten percent had demand in the preceding twelve months.

4) Historical Bucketing.  Are you forecasting using daily, weekly, or monthly time buckets?  The more granular the bucketing, the more time it is going to take to compute statistical forecasts.  Many companies will wonder, “Why would anyone want to forecast on a daily basis?” However, state-of-the-art demand forecasting software can leverage daily data to detect simultaneous day-of-week and week-of-month patterns that would otherwise be obscured with traditional monthly demand buckets. And the speed of business continues to accelerate, threatening the competitive viability of the traditional monthly planning tempo.

5) Amount of History.  Are you limiting the model by only feeding it the most recent demand history, or are you feeding all available history to the demand forecasting software? The more history you feed the model, the more data must be analyzed and the longer it is going to take.

6) Additional analytical processing.  So far, we’ve imagined feeding items’ demand history in and getting forecasts out. But the process can also involve additional analytical steps that can improve results. Examples include:

a) Outlier detection and removal to minimize the distortion caused by one-off events like storm damage.

b) Machine learning that decides how much history should be used for each item by detecting regime change.

c) Causal modeling that identifies how changes in demand drivers (such as price, interest rate, customer sentiment, etc.) impact future demand.

d) Exception reporting that uses data analytics to identify unusual situations that merit further management review.

The Rest of the Story. It’s also critical to understand that the time to get an answer involves more than the speed of forecasting computations per se.  Data must be loaded into memory before computing can begin. Once the forecasts are computed, your browser must load the results so that they may be rendered on screen for you to interact with.  If you re-forecast a product, you may choose to save the results.  If you are working with product hierarchies (aggregating item forecasts up to product families, families up to product lines, etc.), the new forecast is going to impact the hierarchy, and everything must be reconciled.   All of this takes time.

Fast Enough for You? When you are evaluating software to see whether your need for speed will be satisfied, all of this can be tested as part of a proof of concept or trial offered by demand planning software solution providers.  Test it out, and make sure that the compute, load, and save times are acceptable given the volume of data and forecasting methods you want to use to support your process.

Managing spare parts inventories can feel impossible. You don’t know what will break and when. Feedback from mechanical departments and maintenance teams is often inaccurate. Planned maintenance schedules are often shifted around, making them anything but “planned.”   Usage (i.e., demand) patterns are most often extremely intermittent, i.e., demand jumps randomly between zero and something else, often a surprisingly big number. Intermittency, combined with the lack of significant trend or seasonal patterns, render traditional time-series forecasting methods inaccurate. The large number of part-by-locations combinations makes it impossible to manually create or even review forecasts for individual parts.   Given all these challenges, we thought it would be helpful to outline a number of do’s (and their associated don’ts).

1. Do use probabilistic methods to compute a reorder points and Min/Max levels
   Basing stocking decisions on average daily usage isn’t the right answer. Nor is reliance on traditional forecasting methods like exponential smoothing models. Neither approach works when demand is intermittent because they don’t take proper account of demand volatility. Probabilistic methods that simulate thousands of possible demand scenarios work best. They provide a realistic estimate of the demand distribution and can handle all the zeros and random non-zeros. This will ensure the inventory level is right-sized to hit whatever service level target you choose.
    
2. Do use service levels instead of rule-of-thumb methods to determine stocking levels
   Many parts planning organizations rely on multiples of daily demand and other rules of thumb to determine stocking policies. For example, reorder points are often based on doubling average demand over the lead time or applying some other multiple depending on the importance of the item. However, averages don’t account for how volatile (or noisy) a part is and will lead to overstocking less noisy parts and understocking more noisy parts.
    
3. Do frequently recompute stocking policies
   Just because demand is intermittent doesn’t mean nothing changes over time. Yet after interviewing hundreds of companies managing spare parts inventory, we find that fewer than 10% recompute stocking policies monthly. Many never recompute stocking policies until there is a “problem.” Across thousands of parts, usage is guaranteed to drift up or down on at least some of the parts. Supplier lead times can also change. Using an outdated reorder point will cause orders to trigger too soon or too late, creating lots of problems. Recomputing policies every planning cycle ensures inventory will be right-sized. Don’t be reactive and wait for a problem to occur before considering whether the Min or Max should be modified. By then it’s too late – it’s like waiting for your brakes to fail before making a repair. Don’t worry about the effort of recomputing Min/Max values for large numbers of SKU’s: modern software does it automatically. Remember: Recalibration of your stocking policies is preventive maintenance against stockout!
    
4. Do get buy-in on targeted service levels
   Inventory is expensive and should be right-sized based on striking a balance between the organization’s willingness to stock out and its willingness to budget for spares. Too often, planners make decisions in isolation based on pain avoidance or maintenance technicians’ requests without consideration of how spending on one part impacts the organization’s ability to spend on another part. Excess inventory on one part hurts service levels on other parts by disproportionally consuming the inventory budget. Make sure that service level goals and associated inventory costs of achieving the service levels are understood and agreed to.
    
5. Do run a separate planning process for repairable parts
   Some parts are very expensive to replace, so it is preferable to send them to repair facilities or back to the OEM for repair. Accounting for the supply side randomness of when repairable parts will be returned, and knowing whether to wait for a repair or to purchase an additional spare, are critical to ensuring item availability without inventory bloat. This requires specialized reporting and the use of probabilistic models.  Don’t treat repairable parts like consumable parts when planning.
    
6. Do count what is purchased against the budget – not just what is consumed
   Many organizations will allocate total part purchases to a separate corporate budget and ding the mechanical or maintenance team’s budget for parts that are used. In most MRO organizations, especially in public transit and utilities, the repair teams dictate what is purchased. If what is purchased doesn’t count against their budget, they will over-buy to ensure there is never any chance of stockout. They have literally zero incentive to get it right, so tens of millions in excess inventory will be purchased. If what is purchased is reflected in the budget, far more attention will be paid to purchasing only what is truly needed. Recognizing that excess inventory hurts service by robbing the organization of cash that could otherwise be used on understocked parts is an important step to ensuring responsible inventory purchasing.

​

 What is the wiggle effect? 

It’s when your statistical forecast incorrectly predicts the ups and downs observed in your demand history when there really isn’t a pattern.  It’s important to make sure your forecasts don’t wiggle unless there is a real pattern.

Here is a transcript from a recent customer where this issue was discussed:

Customer: “The forecast isn’t picking up on the patterns I see in the history.  Why not?” 

Smart:  “If you look closely, the ups and downs you see aren’t patterns.  It’s really noise.”  

Customer:  “But if we don’t predict the highs, we’ll stock out.”

Smart: “If the forecast were to ‘wiggle’ it would be much less accurate.  The system will forecast whatever pattern is evident, in this case a very slight uptrend.  We’ll buffer against the noise with safety stocks. The wiggles are used to set the safety stocks.”

Customer: “Ok. Makes sense now.” 

The wiggle looks reassuring but, in this case, it is resulting in an incorrect demand forecast. The ups and downs aren’t really occurring at the same times each month.  A better statistical forecast is shown in light green.

Microsoft Dynamics 365 F&SC and AX can manage replenishment by suggesting what to order and when via reorder point-based inventory policies.  A challenge that customers face is that efforts to maintain these levels are very detailed oriented and that the ERP system requires that the user manually specify these reorder points and/or forecasts.  As an alternative, many organizations end up generating inventory policies by hand using Excel spreadsheets or using other ad hoc approaches.

These methods are time-consuming and both likely result in some level of inaccuracy.  As a result, the organization will end up with excess inventory, unnecessary shortages, and a general mistrust of their software systems. In this article, we will review the inventory ordering functionality in AX / D365 F&SC, explain its limitations, and summarize how Smart Inventory Planning & Optimization can help improve a company’s cash position.   This is accomplished by reduced inventory, minimized and controlled stockouts.   Use of Smart Software delivers predictive functionality that is missing in Dynamics 365.

Microsoft Dynamics 365 F&SC and AX Replenishment Policies

In the inventory management module of AX and F&SC, users can manually enter planning parameters for every stock item. These parameters include reorder points, safety stock lead times, safety stock quantities, reorder cycles, and order modifiers such as supplier imposed minimum and maximum order quantities and order multiples. Once entered, the ERP system will reconcile incoming supply, current on hand, outgoing demand, and the user defined forecasts and stocking policies to net out the supply plan or order schedule (i.e., what to order and when).

There are 4 replenishment policy choices in F&SC and AX:  Fixed Reorder Quantity, Maximum Quantity, Lot-For-Lot and Customer Order Driven.

• Fixed Reorder Quantity and Max are reorder point-based replenishment methods. Both suggest orders when on hand inventory hits the reorder point. With fixed ROQ, the order size is specified and will not vary until changed. With Max, order sizes will vary based on stock position at time of order with orders being placed up to the Max.
• Lot-for Lot is a forecasted based replenishment method that pools total demand forecasted over a user defined time frame (the “lot accumulation period”) and generates an order suggestion totaling the forecasted quantity. So, if your total forecasted demand is 100 units per month and the lot accumulation period is 3 months, then your order suggestion would equal 300 units.
• Order Driven is a make to order based replenishment method. It doesn’t utilize reorder points or forecasts. Think of it as a “sell one, buy one” logic that only places orders after demand is entered.

Limitations

Every one of F&SC / AX replenishment settings must be entered manually or imported through custom uploads created by customers.  There simply isn’t any way for users to natively generate any inputs (especially not optimal ones). The lack of credible functionality for unit level forecasting and inventory optimization within the ERP system is why so many AX and F&SC users are forced to rely on spreadsheets for planning and then manually set the parameters the ERP needs.  In reality, most planners end up manually set demand forecasts and reordering.

And when they can use spread sheets, they often rely on wide rule of thumb methods that results in using simplified statistical models.  Once calculated in the spread sheet these must be loaded into F&SC/AX.  They are often either loaded via cumbersome file imports or manually entered.   Because of the time and effort, it takes to build these, companies do not frequently update these numbers.

Once these are set in place, organizations tend to employ a reactive approach to changes.  The only time a buyer/planner reviews inventory policy is annually or at the time of purchases or manufacturing.   Some firms will also react after encountering problems with inventory levels being short (or too high).  Managing this in AX and F&AS requires manual interrogation to review history, calculate forecasts, assess buffer positions, and to recalibrate.

Microsoft recognizes these constraints in their core ERPs and understands the significant challenges to customers.  In response Microsoft has positioned forecasting under their AI Azure stack.  This method is outside of the core ERPs.  It is offered as a tool set for Data Scientists to use in defining custom complex statistics and calculations as a company wishes.  This is in addition to some basic simple calculations as a starting point are currently in their start up phases of development.  While this may hold long term gains, currently this method means customers start from near scratch and define what Microsoft currently called ‘experiments’ to gauge demand planning.

The bottom line is that customers face large challenges in getting the Dynamics stack itself to help solve these problems.  The result is for CFOs to have less cash available for what they need and for Sales Execs to have sales opportunities unfilled and a potential loss of sales because the firm can’t ship the goods the customer wants.

Get Smarter

Wouldn’t it be better to simply leverage a best of breed add-on for demand planning; and a best of breed inventory optimization solution to manage and balance costs and fulfilment levels?  Wouldn’t it be better to be able to do this on a daily or weekly basis to make your decisions closest to the need, preserving cash while meeting sales demand?

Imagine having a bidirectional integration with AX and F&AS so this all operates easily and quickly.   One where:

• you could automatically recalibrate policies in frequent planning cycles using field proven, cutting-edge statistical models,
• you would be able to calculate demand forecasts that account for seasonality, trend, and cyclical patterns,
• You would automatically leverage optimization methods that prescribe the most profitable stocking policies and service levels that consider the real costs of carrying inventory and stock outages, giving you a full economic picture,
• You could free up cash for use within the company and manage your inventory levels to improve order fulfillment at the same time as you free this cash.
• you would have safety stocks and inventory levels that would account for demand and supply variability, business conditions, and priorities,
• you’d be able to target specific service levels by groups of products, customers, warehouses, or any other dimension you selected,
• you increase overall company profit and balance sheet health.

Extend Microsoft 365 F&SC and AX with Smart IP&O

To see a recording of the Microsoft Dynamics Communities Webinar showcasing Smart IP&O, register here:

https://smartcorp.com/inventory-planning-with-microsoft-365-fsc-and-ax/

A statistical forecast of zero can cause lots of confusion for forecasters, especially when the historical demand is non-zero.  Sure, it’s obvious that demand is trending downward, but should it trend to zero?  When the older demand is much greater than the more recent demand and the more recent demand is very low volume (i.e., 1,2,3 units demanded), the answer is, statistically speaking, yes.  However, this might not jive with the planner’s business knowledge and expected minimum level of demand.  So, what should a forecaster do to correct this? Here are three suggestions:

1. Limit the historical data fed to the model. In a down trending situation, the older data is often much greater than the recent data.   When the older much higher volume demand is ignored, the down trend won’t be nearly as significant.  You’ll still forecast a down trend, but results are more likely to be line with business expectations.

2. Try trend dampening. Smart Demand Planner has a feature called “trend hedging” that enables users to define how a trend should phase out over time. The higher the percentage trend hedge (0-100%), the more pronounced the trend dampening.  This means that a forecasted trend will not continue through the whole forecast horizon.  This means the demand forecast will start to flatten before it hits zero on a downtrend.

3. Change the forecast model. Switch from a trending method like Double Exponential Smoothing or Linear Moving Average to a non-trending method such as Single Exponential Smoothing or Simple Moving Average. You won’t forecast a downtrend, but at least your forecast won’t be zero and thus more likely to be accepted by the business.