Category Archives: Forecasts

8 Key Design Considerations for Optimizing Your Demand Planning Process: Part II


Today’s guest post is from Josh Peacher, a Senior Consultant in the Operations Practice of Archstone Consulting, A Hackett Group Company.

In the first installment, we focused on defining the 4 basic design considerations for optimizing your organization’s demand planning process. These considerations included:

  1. Utilization of time series forecasting and exception management to drive a base forecast
  2. Selecting the right software tool for your business
  3. Identifying a set of core metrics and KPIs that help to identify opportunities and drive accountability
  4. Effectively leveraging external information to elicit a more accurate forecast

These design considerations are foundational in nature and effectively addressing each will ensure that your organization’s demand planning process has a solid base. However, to truly move the needle towards world class performance, a set of more advanced considerations must be applied.

5. Drive Towards a Consensus Demand Plan

A formal demand planning process should conclude with an aligned set of forecast numbers that the entire organization understands and can speak to. This doesn’t necessarily mean that a “One-Number” forecast must be reached as this can be very difficult and cause a whole set of different issues. However, organizations should look to align on a set of numbers and be prepared to speak to and manage to the gaps. Key participants in the consensus demand plan conversation include Sales and Account Teams, Finance, Supply Planning, and Demand Planning. Each of these groups will bring a different perspective and set of information to the discussion resulting in a more informed final demand plan.

6. Identify the Right Level of Detail

When defining the appropriate level of detail to forecast at, leading companies strike a balance between importance to the business and complexity of the process. The diagram below defines a general set of guidelines for identifying the appropriate level of forecast detail based on the situation. As a general rule of thumb, the more important and complex the set of items is to the business, the higher the required level of detail and rigor.

Complexity vs. Importance

7. Ensure Adequate Resources

As I mentioned in the first installment, demand planning is commonly an overlooked element of supply chain planning. This often leads to an insufficient allocation of resources by the organization. Demand planning is an arduous process that requires a high level of dedication and attention. More times than not, I see organizations that have failed to realize this and leave their demand planning team without the necessary bandwidth to perform effectively. The net effect is a less accurate forecast, poor demand signals trickling through the system, and a higher turnover rate. A few simple rules of thumb to ensure that your organization is not falling into this trap include the following:

  • Install dedicated analysts for demand planning.
    This will ensure that demand planners are focusing on value-add activities and have the right information on hand to make informed decisions.
  • Make sure that your demand planners aren’t wearing too many organizational hats.
    It’s an odd phenomenon but demand planners often end up taking on responsibilities that are well outside of their job scope and not essential to their core function. The best way to decipher this is just to simply ask them where their pain points are. Trust me … they will tell you!
  • Understand which segments are the most critical and complex to the business and distribute them across your demand planner resources.
    Ideally, each of your demand planners will have a portfolio of demand responsibilities that are evenly distributed amongst the four quadrants of the above diagram.

8. Define your Organization Process Model

Too often I have seen organizations operating in an environment of chaos because they lack a defined process and cadence for their demand planning cycle. You may believe that you have a process in place, but can you articulate what it is? Can the demand planning resources in your organization define the calendar of events that make up the process? Many times what people believe to be a process is actually floating tribal knowledge and tends to vary depending on who you ask within the organization. Without a well-defined process, it’s difficult to hold others accountable and overall performance tends to suffer. An optimal process must be defined for each organization based upon it’s unique set of variables and constraints. However, the list below is a set of monthly activities that can be found in most leading company processes.

  • Prepare Data –
    Cleanse and gather all required data for the demand planning process (internal and external)
  • Generate Initial Forecast –
    Generate both the base statistical forecast and manage exception SKUs manually
  • Incorporate Market Intelligence –
    Collaborate with trade partners and external contacts to incorporate quantitative and qualitative data into the forecast (e.g., POS Data, Customer Forecast, Promotional Calendars, Pull-Forward Buys)
  • Consensus Reconciliation Meeting –
    Meet with sales and finance to reconcile the bottoms up forecast with top down financials and sales forecasts
  • Refine and Publish Final Forecast –
    Make final adjustments to forecast before transmitting to ERP
  • Monitor Performance –
    Monitor forecast for large anomalies and diagnose root cause of error

Thanks, Josh!

8 Key Design Considerations for Optimizing Your Demand Planning Process: Part I


Today’s guest post is from Josh Peacher, a Senior Consultant in the Operations Practice of Archstone Consulting, A Hackett Group Company.

Demand Planning was once an overlooked element of supply chain management. However, more and more companies are beginning to understand how essential this component is to overall operational well-being. After all, a demand forecast is the genesis of the supply chain process. If poor demand signals are being sent through the system, it becomes extremely difficult to manage raw material and finished goods inventories, execute an efficient manufacturing process, effectively service customers, and ultimately drive an accurate financial forecast. So if your organization hasn’t already taken a long, hard look at improving its demand planning process, it’s time to begin. As a starting point for your journey, let’s take a look at the 8 key design considerations for optimizing your demand planning process. In this first installment, we’ll focus on the 4 most basic design considerations and then move to more advanced principals in the second installment.

1. Start with Statistical Forecasting and Exception Management

  • Statistical forecasting should always drive the original forecast. A simple set of formulas such as exponential smoothing, weighted moving average, and Holt-Winters can deliver more accurate, reliable, and efficient forecasts across the entire sku base than manual forecasts. This can often be a change management challenge for many organizations as demand planners feel a pride of ownership over their forecast and have trouble with relinquishing control to a set of arithmetic functions. This is where exception reporting comes into play.
  • Exception reporting utilizes a set of pre-defined criteria to identify skus that are not ideal candidates for statistical forecasting. Since the strength of statistical forecasting comes from identifying patterns in demand history, highly erratic and/or variable skus are not good candidates and require manual intervention of the forecast. While exception criteria are customizable, common filters include frequent zero demand periods, high variance between last 6 months history and next 6 months forecast, high variance in month-over-month demand history, and frequent shortages. Exception reporting is also an excellent way for demand planners to prioritize their time across the sku set and focus their efforts on the skus that truly require attention.

2. Select the Right Software Tool

In today’s environment of sku proliferation and real time information, it’s become a necessity to utilize a demand planning tool to assist with the demand planning process. Software solutions such as Manugistics, SAP APO, and Logility all have their strengths and weaknesses. Key criteria to evaluate when selecting a solution include:

     
  • Customer service reputation of the provider
  • The tool’s ability to handle forecasting nuances (i.e., 5-4-4 calendar recognition and promotional forecasts)
  • Transparency and reliability of the generated statistical forecast
  • Forecast performance reporting and exception reporting capabilities
  • Flexibility to forecast at multiple levels (e.g., sku, customer, category, business unit)

3. Track the Right Metrics

Demand planning metrics should serve two purposes:

  1. Identify improvement opportunities and
  2. Drive accountability.

The appropriate metrics will vary based on the characteristics of the industry and company in question. However, a few core, agnostic metrics are routinely found in leading organizations. These include:

     
  • WAPE (Weighted Absolute Percent Error) – In my opinion, WAPE is the most balanced and telling measure of forecast error. Some professionals will advocate for MAPE. However, MAPE doesn’t effectively account for volume as the forecast error % for each period is treated equally.
  • BIAS – Bias is similar to forecast error. However, bias provides a measurement of whether your forecast tends to be above or below actual demand thus signaling a forecasting over/under “bias”.
  • Period-over-Period Error Trend – You’ll want to understand whether your demand planning process is improving or digressing. Measuring the forecast accuracy over time will also help to identify meaningful changes occurring in the business.

4. Leverage the Correct Data

Statistical forecasting and exception management will help to get a reasonably accurate forecast. However , to drive forecast error down to best-in-class levels, demand planners must leverage external information.

As the graphic above shows, there is an abundance of information that demand planners could call upon to help them adjust their forecast. The real art of demand planning is knowing which of these data sources to use and when. Over time, your organization will get a sense for which information streams are most relevant and can begin to build a rules-based process around the use of external information.

Thanks, Josh! We look forward to Part II.

Can We Harness the Wisdom of Crowds in Supply Chain Forecasting?

How Do We Harness the Wisdom of Crowds in Supply Chain Forecasting? A little over two years ago, I posed this question to you. I got a few responses, mostly private, who were thankful that I pointed out that you cannot just blindly follow the wisdom of just any old crowd, because expert judgements often demonstrate logically inconsistent results, but not a lot of advice on how we could successfully approach the task of integrating the wisdom that crowds could provide in our supply chain processes. The reason that we wanted to tackle this problem is because it is true that teams of forecasters often generate better results (and decisions) than individuals as long as the teams include a sufficient degree of diversity of information and perspectives.

But this is the caveat. The wisdom of crowds only holds if the crowd is large enough to contain the necessary diversity of information and perspectives in a statistically significant way. In other words, you will need a lot of people, and these people will need to be from diverse backgrounds and possess diverse skill-sets. But even this might not be enough in some situations.

As pointed out in this great blog post over on the World Future Society blogs by Thomas Frey from last August on The False Wisdom of Crowds, the decision between flying on a plane piloted by a single AI or the combined intelligence of 3,000 people is not as simple as you think. While it is true that the combined IQ and skill-set of 3,000 people is much greater than any AI on the planet, as you Next Generation Trekkies aware of the quick adaptability of the Borg will be quick to point out, it is also true that if all 3,000 of these people are farmers from the MidWest, it is likely the case that not one of them will know how to fly a plane! In contrast, the AI might be the best autopilot software in the industry, successfully used problem-free on tens of thousands of flights. The only way you’d beat that is if you had a collective of 3,000 of the best airline pilots in the industry. But the statistical likelihood of selecting that crowd from the global population is astronomical.

As Drew Curtis, founder of Fark.com, points out, “Crowds are dumb. The reality is that, while people are very good at knowing what they personally want, they are generally very bad at understanding the truths of the world around [them]. If you want proof, consider the examples given by Thomas and Drew, which include:

  • In the ’50’s, it was common knowledge that if a nuclear bomb went off in your city, you’d be safe if you simply learned to “duck and cover”.
  • Until 2007, it was a well-known fact that real-estate was a great investment where you would virtually never lose money.
  • Only once percent of Web comments have any value and the rest are just garbage.

In other words, diversity is not enough. You need expertise. And you need the right expertise. But as pointed out in SI’s post from 2010 and Thomas’ blog post, ‘social influence’ diminishes the range of opinions and tends to lead crowds in the direction chosen by the most respected and/or socially powerful individuals. So you have to gather data from a “blind crowd” that cannot see each other.

In other words, when you put it all together you need:

  • diversity, as addressed in our previous post,
  • privacy, as partially addressed in our previous post,
  • expertise, as demonstrated by Thomas’ blog, and
  • statistical significance, as not adequately addressed yet.

Taken together, these requirements pose a bit of a problem, which is made clear in Thomas’ post where he quotes a recent study by McKinsey and Company that calculated an immediate shortage in the US of almost 200,000 people with analytical expertise and 1.5 million managers and decision makers with the skills to understand and make decisions based on the analyses provided by the analytical experts. Overall, we’re starving for expertise in Supply Chain, as evidenced by the fact that less than 10% of companies truly employ advanced sourcing techniques! The average company just doesn’t have enough people to meet the diversity, expertise, and statistical significance required to guarantee that a crowd decision will be better than the decision of their “leading expert” in that area. And since most firms don’t want to share expertise, sourcing processes, or suppliers, especially where strategic or high-value categories are concerned, they’ve essentially cut-off external sources of expertise. The result: beyond non-strategic / low-value categories they would be willing to hand off to a GPO, their chances of truly harvesting the wisdom of crowds for many Supply Management processes are low, at best — and this leads us to wonder if we really can harness the wisdom of crowds in supply chain forecasting in practice.

New Thoughts? Comments? Criticisms?

Supply Chain Planner — Here are Three Solutions to Nearly Every Problem

A recent piece over on Supply Chain Cowboy on Three Silver Bullets to Solve Nearly Every Supply Chain Fire simultaneously enthralled and shocked me because I cringe every time I hear that air freight is one of the three solutions to your current supply chain dilemma, as it is a prime indicator of a major supply chain issue — specifically, lack of planning.

But there are ways to avoid the issue. The first one is:

Supply Chain Forecasting Systems

A good, modern, supply chain forecasting system is the best way to figure out not only what you are going to need, but when you are going to need it and when you are going to have to get the orders in and production started in order to meet shipping deadlines and avoid the need for air freight.

The second way to avoid the issue is:

Supply Chain Visibility

(Near) real-time visibility into where your stuff is from your suppliers, their suppliers, and their raw-material suppliers. All delays have ripple effects, and the best way to prevent a hiccup, or disruption, that will force you to have to use air-freight is to have real-time visibility all the way through your supply chain so you can be aware of a potential issue as soon as it happens.

And the third silver bullet, I’m sad to say, is:

Standby Air Freight

Good forecasting will significantly reduce the number of emergencies and the number of times you have to ship air-freight to meet a deadline, and good supply chain visibility will reduce this number even further as you will be able to order from secondary suppliers or ship through back-up carriers when hiccups or disruptions do arise to meet the deadlines laid out in your forecasting system. That being said, no technology will completely eliminate the need. There will always be unexpected events that will cause interruptions at the last minute where the only recovery option is to air freight reserve stock. If the Super Panamax ship gets delayed a week in port because of customs issues after your cargo is loaded, there’s nothing you can do. Or if a second tier supplier gets cut off because of a civil uprising and you have to arrange for the first tier supplier to get replacement product from another second tier supplier further away, there may be no other way to get the product fast enough. That being said, the number of instances where there is no way but up should be few and far between with good supply chain planning and visibility systems.

How Not to Excel at Forecasting

Simply put, use Microsoft Excel. It’s appalling that a recent survey by ToolsGroup and the Global Market Development Centre (GDMC) found that even though two-thirds of companies in the consumer goods supply chain consider demand volatility and forecast accuracy a high businesses priority, half still rely on Excel spreadsheets for forecasting.

Relying on Excel for forecasting is like relying on:

  • a Longship to get you across the Atlantic
  • your first guess on Let’s Make a Deal to be the right one
  • a shareholder proxy getting on the ballot at a Fortune 500
  • Florida surviving a hurricane season without any major city suffering damage
  • the price of fuel going down and staying down for an upcoming series of spot buys
  • natural resource supply to be consistent and predictable year-over-year
  • a flip of a fair coin to come up heads seven times in a row

Now, it’s true that:

  • the Vikings did make it across the Atlantic in a Longship, but a single storm could sink it
  • the first door you pick, with one-in-three odds, could be the right one, but the odds are actually twice as good if you switch
  • an activist shareholder can sometimes get a proxy on the ballot if he or she has enough time and money, but as pointed out by John Gillespie and David Zweig in Money for Nothing (How the Failure of Corporate Boards is Ruining American Business and Costing Us Trillions), examples are few and far between
  • even though no storms made landfall in Florida in 2011, this is Not a common occurrence
  • gas prices did consistently drop in the USA between September 2008 and December 2008, but have been otherwise steadily rising for the last five years
  • in some years the rice, sugar, and corn crops are almost the same as in the previous year, but given the increase in hurricanes, tsunamis, droughts, and other natural disasters in recent years, this is not a common occurrence
  • yes, heads can come up seven times in a row when flipping a fair coin, but the chances of this happening are less than 1%

In other words, you can forecast with Microsoft Excel, but your chances of doing well, especially given that 90% of spreadsheets have non-trivial errors (and collectively cost enterprises billions, as Fidelity and Fannie Mae found out), are (vanishingly) small (as the complexity of the forecast increases). One has to remember that there’s no intelligence behind a spreadsheet and they are just a source of peril that can cost your organization millions without anyone noticing.