Company

Recent Posts

Categories

Archives

About Demand Planning LLC

Demand Planning LLC, based in Boston MA, is a consulting boutique comprised of seasoned experts with real-world supply chain experience and subject-matter expertise in demand forecasting, S&OP, Customer planning, and supply chain strategy.

We provide process and solutions consulting, as well as customized training across a variety of industries.

Through our knowledge portal DemandPlanning.Net, we offer a full menu of training programs through in-person and online courses, as well as a variety of informational articles, downloadable calculation templates, and a unique Demand Planning discussion forum.

  • 11Jan

    As a consultant, I get asked this question a variety of times.  Why should we improve forecast accuracy?  What is it worth to us?  What is the business case to spending money to improve the process and implement some expensive demand planning software?

    There is value in this for sure – the evidence comes from the fact that year-after-year companies continue to hire more demand planners and invest millions of dollars in putting together processes and software packages.  Now to dissect this, let us look at the Service-Cost Balance model that we preach in our workshops.

    Improving the demand forecast directly gets to two things that are important to companies:

    1.  Increasing the top-line – Improve customer fulfillment and delivery and increase the level of Sales.  Nothing works like reputation or service performance.  People flock to your company products and services if they are satisfied with what you offer.  The converse will be painful – bad news spread like wild fire.

    2.  Improve the bottom-line through process efficiency – Optimize inventories by trimming the level of inventories, cutting obsolescence and producing just enough to meet your demand.

    3.  Eliminate process costs – Expediting, over-time that may otherwise be required to fulfill volatile and unpredicted customer demand.

    Now if we want to quantify the benefits from forecast accuracy, it is difficult to quantify the first item.  What is our lost sales?  What is the effect of one point reduction in fill rates?  Did it cause lost sales?  Did we lose a customer?

    It is a little easier to look at the savings in inventory through improvements in forecast accuracy.

    At any time, your inventory level is comprised of Safety Stock to meet demand and supply uncertainties and the inventory needed

    But your average inventory = Safety Stock + Lead time demand or (order quantity if you are using the min-max method to deploy).

    We can find a relationship between the amount of safety stock you can decrease for each percent point reduction in Forecast Error, since forecast error figures directly in the Safety Stock calculation.

    Safety Stock = Forecast Error * Service Level Quotient * Square Root (Lead Time)

    From simple calculus, the effect of one unit reduction in Forecast Error is just the product of Service Level Constant * Sqrt (Lead-time).  So at 98% service level with a lead time of say two months, the effect of one unit reduction in Forecast Error is 2.90 Units.

    What is the impact of a % point reduction in MAPE?  The answer depends on the level of your current forecast quality.

    However, the relationship of inventory reduction to your current MAPE Level is not linear.   The calculation on the reduction in inventory is a function of your current level of forecasting.  This determines the quantity of inventory reduced for each percent point improvement in Forecast error.

    If you are at 50% MAPE on average currently, the reduction would be 2% of inventories for each point improvement in forecast error.  IF you are at 25% MAPE currently, the improvement will be 4% for each point reduction.

    This is actually a conservative estimate since it considers only the safety stock component of the total inventory.

    If your forecast is an over-estimate of the Lead time demand, then you carry additional inventory not accounted for in the above calculation.  This will also introduce obsolescence risk.

    So a lower bound is to just use the reduction in safety stock but being aware that a better forecast will also allow you to deploy better to the lead time demand.  Better forecast can also help you reduce production and logistics costs such as over-time and expediting costs.

    So if you have $200MM in inventory and let us assume that 25% of that total is safety stock then you have $50MM in safety stock.

    A reduction of 10 point MAPE will result in a reduction of $10MM in safety stock.  10 point MAPE reduciton =20% reduction in Safety Stock or $10MM assuming currently your average MAPE is 50%.

    Again this does not consider any reductions in the inventory carried to cover lead time demand (particularly due to forecast bias) or obsolescence that results from unsold inventory.  Obsolescence is a bigger risk since you lose all of the inventory investment.

    Next question is what does the reduction of $10MM in inventory really mean to you?  This completely depends on the inventory carrying cost for your company.

    Inventory carrying cost can be any where between 12% to 40% per year.  So in financial terms, a $10MM reduction in inventory could result in bottom line improvement of $1.2MM (at 12% carrying cost) to $4MM.

    Now we are talking real dollars……………..

    Let us try to save some of those costs by bettering the demand plans!

    Tags: , , , , , ,

  • 23Nov

    This has been a recurring challenge and a potential land mine when it comes to the raw number crunching in demand planning.  Zeroes and Nulls……….

    Is the Null the same as a zero?

    Although excel formatting can code a zero as a dash, can a zero be interpreted as a Null?  If so when and when not?

    Similarly, including some may also be bad for the health of your demand forecast.

    Leading zeroes – Will you include them in developing a statistically modeled forecast.

    How about nulls in the middle of the data? These either show up as nulls, .dots, or some times as zeroes.

    Ok.  Now that I have asked too many questions, I will also propose some answers for you.

    Generally Nulls can never be treated as zeroes.  They are different things.  Nulls mean nothing, the absence of anything.  Nulls mean no data or no observation.  If you average a series with nulls, the nulls count in either the numerator or the denominator.  Zeroes are different.  If you average them, they will have no contribution to the numerator but will count as an observation in the denominator so you will have your average reduced with the presence of zeroes.

    At least in demand forecasting, we can coin the following rules:

    1.  Leading Zeroes can be interpreted as Nulls. 

    At times a product may be slated to launch in a specific month and hence the system may start recording zeroes as data if the launch is delayed.  Leaving them in may result in a poor statistical forecast.

    2.  Nulls in the middle can be interpreted as zeroes.

    Some systems may record nulls if there is no demand activity.  However if the nulls occur in the middle of a time series history, I would recommend they be treated as zeroes.   Most intermittent demand data is characterized by zero sales volume frequently.  If you leave them as nulls, this will inflate your average and generally result in a upwardly biased demand forecast.

    Imagine this scenario:

    90 Null Null 90 Null 90 Null Null 90 Null Null Null 90.

    If you ignore the null, then your demand forecast will be 90 per month if you use the average as the model to forecast. If you interpret the null to be zero, then your average will be 30 units a month.  If the customer has a three month requirement of 90 units and orders only once in every three months, then 30 per unit seems more likely as a forecast.

    3.  Trailing zeroes and Nulls

    Exclude or include?  What do you do with these?

    They look harmless to me if you have a decent exponential smoothing engine. Do you agree?

    We will discuss our approach in detail in our Demand Planning and Sales Forecasting Tutorial workshops scheduled for Feb 2013 in Dallas, TX.  We go into the nuts and bolts of many practical challenges that demand planners face.  This is why our workshop is considered to be the most practical and hands-on when it comes to Training for Demand Forecasters.

    Ignoring these nulls and zeroes can be perilous and lead you down the wrong path.  This may affect both your forecasting and inventory setting.  More dangerous to ignore these if you are calculating safety stock parameters.

    You can see more info on our workshop at http://demandplanning.net/demandplanning_tutorialCA.htm. I am told that we have just a handful of seats left for the early bird quota.

    Have a great holiday!

    Tags: , , , , , , , ,

  • 06Mar

    If all works well, then it is a perfect world.  You carry just the right amount of inventory to service your customers at 99% and get away with very minimal working capital.  Obviously, your low cash-to-cash cycle should result in larger portion of your Gross Margins go to your Net Margins………..

    Excess inventories happen as a matter of fact:

    • Forecasting problems – not knowing what the customers need.  This may also result in some obsolescence.
    • Forecast Bias – Just keeping the forecasts high generally on everything.
    • Sudden Demand reduction due to market place volatility or losing a key customer.
    • Economies of Scale in production – Higher lot sizes are way too attractive to resist.

    Excessive inventory can also be carried as a price Hedge.  Steel prices are expected to rise and quantities may even be in short supply. So you buy up and keep more of it.

    Life time Buys – Rare earth materials or a supplier that is close to a single source is facing financial difficulties.

    Utilities some times carry spare parts inventory for the next 100 years. Perhaps one or two ancient grids use these parts. If we stock out of these parts, the Utility has no choice but to scrap the old grid and build a new one. The opportunity cost may actually outweigh many times over the cost of carrying these parts.
    Excess inventory can also result from supplier uncertainty. If supplier does not meet schedule or if the lead time is time varying over a period, you have to carry more inventory to meet the uncertainty in supply.

    There is perhaps another reason but really a different version of the price hedge. IF suppliers offer a quantity discount, then that ends up lowering your cost of production with the consequent higher price to pay on the inventory carrying cost.

    The punch here is the lowered cost per unit from the discount that applies to your consumption as well. This may result in ordering and carrying a quantity much higher than the dictated EOQ.  Now here comes the distinction between items above the COGS line and items below.  If a quantity discount is offered, this lowers the COGS and boosts the Gross Margin.   You may have to order more than your calculated EOQ to avail the discount.

    Tags: , , , , ,