“Guests, like fish, begin to smell after three days” is an adage attributed to Benjamin Franklin. Today we might adapt that old adage to say something about demand disruption. We are still dealing with so much demand uncertainty even several months into the pandemic, and that stinks like old fish! And there is no end in sight.
No matter what industry you are in, demand uncertainty has moved in and isn’t moving out any time soon. Many industries are seeing lower demand and huge uncertainty, including anything to do with away from home eating and entertainment. For example, the continued daily uncertainty about restaurants being opened or closed complicates planning for foodservice suppliers, suppliers of suppliers, and growers.
Other industries, such as retail food manufacturers, are outsourcing more production in response to increased demand and consumer hoarding. Whether demand is up or down, many companies are trimming product lines to improve profitability and streamline manufacturing. Regardless of whether your demand is up or down, the disruption means that supply chain processes which worked for years and decades now need continued adaptation to changing circumstances.
How do you plan when there is no reliable track record for demand? Suppliers to the disrupted industries need at least three things to survive the crisis:
- Demand planning processes that do not rely on history
- Supply planning flexibility to react to unexpected demand changes
- Sales and Marketing initiatives to find new outlets for product when demand unexpectedly drops.
This is a Supply Chain blog, so I’ll share some ideas about demand planning and supply planning. I’ll leave the Sales and Marketing initiatives to people who are experts in those areas.
Demand Planning Strategies
Below are six strategies to help Demand Planning navigate through these times. Someone else might have a slightly different list, but I offer these as suggestions that every Demand Planning organization should consider during this disruption.
Increase Collaboration
In the absence of valid history, demand planners need to rely on increased collaboration. Perhaps you previously had a consensus process that only reviewed some items with your commercial team. Now you need to start reviewing them all.
Ideally, Demand Planners should already be informed of market decisions such as promotions, new distribution, etc. Now you will also need additional information such as how many customer locations are open or closed in various geographic regions
Consolidate input from marketing, sales, and customers into forecasts by item, customer, and geography. Evaluate the results of these forecasts relentlessly to see which ones are adding value.
Increase Frequency of Planning Cycles
Although monthly and quarterly trends may have been enough to go by in the past, the situation is far more volatile now. You need to be able respond to changes in daily and weekly trends.
As things change daily, you will need a rapid reporting capability. You don’t want to overreact to new information, but you must capture and understand new information as quickly as it is made available. Ad hoc planning and the running of multiple scenarios will be critical in understanding and evaluating the latest information. Delayed in forecast updates might be costly, so you need to be able to update plans frequently and efficiently.
Use Point of Sale Information
If it is available, use Point-of-Sale (POS) data for consumer demand to build better forecasts. POS data is the purest form of consumer demand. Depending on your industry you may be able to use Nielson data for retail sales. Or you may be able to have your customers share their data for sales via a collaborative program that will give you access to downstream POS data.
Using POS data will help protect you from the Bullwhip Effect that distorts the demand signal because of actions in the supply chain between you and your customer. However, there is still the risk that end consumers are hoarding your product relative to their normal ordering patterns.
Use Predictive Analytics and Causal Modeling
In addition to POS data, consider whether there are additional metrics available that can be early warning indicators of changes in demand. Daily sales, weekly sales, and open orders all might be used to monitor whether demand is tracking as expected. You might use additional Predictive Analytics based on factors such as whether certain regions are easing or tightening social distance rules.
Even in normal times, Causal Modeling of these independent variables will often do better at forecasting future periods than the univariate statistical models demand planners typically rely on. However, the regression analysis to establish the relationships between all the possible independent variables and the actual demand for each item might require a significant amount of computing power. Machine Learning has the potential to generate better forecasts from all the data, but unless you have already implemented it, do not expect results in time to help with the current crisis.
Well-trained demand planners should be able to develop models based on POS data and other independent variables. However, the effort to build these models amidst the current disruptions may require a temporary increase to your demand planning team.
Update Demand Segmentation
Perhaps you have previously done demand segmentation where you consider different strategies for high volume vs. low volume and high volatility vs. low volatility. But now the disruption of demand means it is time to update demand segmentation (and update it frequently).
When doing Demand Segmentation, volatility is typically measured by the Coefficient of Variation (COV). The strategies suggested below for Supply Planning go back to the basics of Demand Segmentation and how to manage items with high COV.
Instead of using history, you will need some way of estimating demand volatility in the current situation. One possible approach would be to use high, low, and best guess estimates for each item to estimate a probability distribution for demand. COV can then be calculated based on standard deviation and mean of the estimated probability distribution.
Prepare History Overrides
Be very wary of statistical black boxes and best-pick forecast models based solely on history. At first, these forecasts are unlikely to react quickly to a disruption. But then coming out of the disruption, they may overreact to recent history. A well-trained demand planner with good market intelligence will be able to tune stats to the post-covid reality much faster than an automated best-pick forecast model.
Seasonality may be the same post-covid as it was pre-covid, while base forecast level and trend may be different. Demand planners will need to evaluate trends in detail by product, customer, and geography. The input from product experts mentioned above will be critical in developing the market intelligence to make these evaluations. Once the demand planners have a comfort level of the expected new patterns, they can adjust history from before and during the disruption. The adjusted history (or history overrides) will then drive statistical forecasts to align with the new expectations.
However, demand needs to stabilize into new patterns before adjusted history can be finalized. How does one know when demand has stabilized enough to start using history again? One suggestion may be to use XmR control charts to track forecast at item level before, during, and after the disruption. These charts can help determine when forecast accuracy is back to a stable pattern. I’ll explore that concept more in a future blog.
Supply Planning Strategies
Demand planning can only go so far in addressing uncdertainity. The primary tools for managing uncertain times are in the supply planning arena. Production systems will need to be more flexible. Here are four suggestions that may help create the flexibility needed.
Increase Order Lead Time
When demand is significantly reduced, it becomes much more challenging to schedule production effectively. Longer lead-time on customer orders will improve your ability to effectively group similar items to minimize setups and changeovers without driving up unnecessary inventory. You might want to consider a two-tiered approach to inventory where certain high-volume, stable items are kept in stock and available to order, while extending customer order lead time on items that have more uncertainty in demand or are difficult to schedule.
Adjust Safety Stock
Traditional safety stock calculations will increase safety stock with increased demand uncertainty. However, lower volume means more capacity is available, so production may be able to respond more quickly, reducing the amount of safety stock needed.
You might want to increase inventory on stable items that have steady demand, while decreasing inventory on items that have more uncertainty in demand. This will help to increase production flexibility and protect priority items and priority customers while mitigating the risk of obsolescence.
Run Smaller Lot Sizes
Maybe your production lot size is based on an Economic Lot Size calculation to balance changeover costs with inventory carrying cost and risk of obsolescence. Whatever methodology you have used to determine minimum production runs, it needs to be re-evaluated. Increased demand uncertainty means increased inventory costs and increased risk of obsolescence. If setup/changeover costs are unchanged, your economic lot size will be smaller. Smaller lot sizes may also be forced by customers uncertainty, and credit risk. Smaller lot sizes may increase manufacturing costs, but this will be offset by reduced inventory and obsolescence risk.
Use Demand Sensing
Demand Sensing helps identify the actual customer order trends and improves the near-term forecast. It is a form of predictive analytics that typically is focused on the near-term (maybe only a few days into the future). As such, it is really a supply planning tool, not demand planning.
Like Machine Learning, Demand Sensing takes time to implement and tune before it yields useable results. Unless you have already started working on it, do not expect results in time to help with the current crisis. If you have already started, accelerating and expanding the use of Demand Sensing can improve your responsiveness to demand changes in the current uncertainty.
Conclusion
No matter what industry you are in, you are having to adapt to new ways of working. The additional workload to adapt may also require additional resources on temporary basis. I’ve given some suggestions that might help adapt to the ongoing uncertainty, and there are some additional resources listed below. Do you have any other suggestions do you have that have helped you adapt to the ongoing disruptions?
Additional Resources
Nielson.com, July 22,2020. 3 Key Challenges Impacting Your Demand Planning Now
Sunil Bharadwaj, Journal of Business Forecasting, volume 39 issue 2, summer 2020. How Supply Chain Functions Must Evolve Post Lockdown
Eric Wilson, demand-planning.com, May 18, 2020. 7 Demand Planning Tips for Surviving Post-covid Uncertainty
Brent Johnstone, mdm.com, July 16, 2020. 59 Questions for Distributors to Assess Your Pandemic Response