The New Era of Demand Planning

When the COVID-19 pandemic began to take hold, leaders at Velux – a skylight manufacturer that experiences relatively predictable supply and demand patterns in normal times – anticipated a dip in demand for their products. In fact, in the hardest hit markets like Spain and Italy, demand dropped to zero.

Then, the unexpected happened.

As people found themselves stuck at home for longer than anyone imagined – and anticipating a summer with limited or no vacations or travel – they craved more daylight in their suddenly shrunken worlds. During lockdown, demand for what this 75-year-old, family-owned Danish company calls “windows for roofs” went through the roof.

As Phillip Melchior, head of global planning at Velux, told SAP Innovation Evangelist Tom Raftery’s Digital Supply Chain podcast in 2020, it’s been a real roller coaster ride – from demand falling by half to increasing by 50% in the company’s four biggest markets during the first several weeks of the pandemic. Meanwhile, the company’s stocked products were at record lows (earlier in the pandemic, Velux itself had shut down its factories for two weeks) and its suppliers were struggling to deliver.

But blindly building more skylights, as supplies became available, wouldn’t solve the problem. Velux serves customers in many countries with varying standards for window dimensions. What the company needed was a sort of crystal ball: a clearer picture of how demand was likely to shift in its markets around the world.

Post-pandemic supply chain practices prioritize risk management

Like many other organizations, Velux discovered that traditional demand forecasting (which relies on historical data to less frequently plot supply chain strategies) fell short during this period of ongoing, rapidly changing (and unprecedented) upheaval. While companies had weathered previous supply chain disruptions (a tsunami in Japan, floods in Thailand, a volcanic eruption in Iceland), those breakdowns were localized, temporary, and fixable. What COVID-19 has made clear is that longer lasting, global supply chain hits – particularly when accompanied by extreme and unprecedented swings in demand – require a different demand planning approach, one that is rooted in current realities and uses more agile and predictive modeling.

To deal with the successive supply and demand shocks, Velux amped up its demand planning processes. It turned a once-monthly integrated planning exercise (involving sales, operations, finance, and logistics) into more frequent meetings – at least weekly and sometimes semi-weekly or on-demand – that incorporated more timely data. The company’s embrace of sales and operations planning, an integrated business management process that drives organizational consensus to balance supply and demand, was invaluable and became a more continuous process. And it enabled the company to not only ride this roller coaster, but also capitalize on the increased demand.

Businesspeople using wireless devices in the office

While this pandemic may seem like a once-in-a-lifetime event, veteran futurists tell us this is just the beginning. Disruptions will become more prevalent, persistent, and punishing. The speed, frequency, scope, and scale of catastrophes from global climate change and cyberterrorism to rolling pandemics and more are likely to increase, with few obvious patterns to guide decision-making.

More data – and more collaboration

To ride these waves, organizations need to integrate more – and more recent – data into demand planning. The process will need to become not only more frequent but also more forward-looking and automated, which has become feasible thanks to the self-learning capabilities of artificial intelligence (AI) systems as well as the availability of relevant demand signals. In contrast to traditional demand forecasting, a new approach involves integrating and analyzing data indicative of current trends likely to impact demand. By analyzing information like point-of-sale data, weather forecasts, social media feeds, costs and pricing, competitive intelligence, macroeconomic indicators (like GDP, interest rates, inflation, and unemployment rates), and healthcare data, organizations can better anticipate and prepare for shorter-term or unexpected shifts.

Understanding the correlations of influencing factors also provides the basis for much better long-term planning of investments like pricing, promotions, and marketing. This process is often referred to as “long-term demand sensing.”

“There is a big transformation in the demand planning space,” says David Simchi-Levi, professor of engineering systems at Massachusetts Institute of Technology (MIT) and director of the MIT Data Science Lab. “At a high level, companies are starting to use multiple sources of data to better understand and predict future demand.”

In order to make these changes, however, the demand planning function will have to become more expansive. Supply chain, manufacturing, logistics, and sales experience will continue to be valuable, but these must be complemented by data science and analytics expertise to succeed in transforming intuition and data into actionable insight. This new approach to demand planning won’t be just for the supply chain function anymore. It will be an enterprise capability, integrated into overall strategic planning and requiring greater collaboration with partners and customers as well as with the machines preparing demand predictions behind the scenes.

This new approach to demand planning won’t be just for the supply chain function anymore. It will be an enterprise capability, integrated into overall strategic planning.

There are challenges to overcome, but some organizations are realizing that they already have access to much of the data they need to do a better job at short-term demand planning. With mature analytics capabilities to generate insight from relevant data and the ability to make decisions and changes based on that demand intelligence, they can have a significant impact on creating a more sustainable, long-term business plan as well. “It’s all about leveraging data and analytics that you have,” says Simchi-Levi, “and it turns out you can do that much faster than anyone thought.”

By applying analytics and machine learning to the many shifting variables that drive demand – and doing so more frequently – organizations can keep pace with and respond more effectively to the marketplace. Such an approach reduces errors in long-term forecasts, increases efficiency, and – ultimately – boosts customer satisfaction with better availability or improved product design or configuration.

Companies that have begun incorporating and analyzing demand at a more granular and more frequent level are able to make the necessary changes to meet near-term demand with greater precision.

While skylight maker Velux relied on traditional spreadsheet analytics for its initial COVID-19 response, the company’s short-term demand planning approach could be improved, planning head Melchior noted on Raftery’s podcast. Melchior said that one of his company’s next steps is to deepen their capabilities with the application of machine learning to better anticipate unusual fluctuations in demand.

Woman interacting with wall screen

From forecasts to foresight

The experience of COVID-19 forever changed the nature of demand planning. The process can no longer be wrapped up in a group’s knowledge or intuition but must be driven by leading indicators. The days of gathering once a month to agree on a demand forecast based on last year’s data are over. As Paul A. Myerson, instructor of management and decision sciences at Monmouth University, wrote recently, “Companies that still primarily rely on using historical demand data to create forecasts are, in effect, driving while looking in the rear-view mirror. We know the result of that.”

The goal of modern demand planning is not consensus but sensing – using a variety of current data and advanced analytics to anticipate and respond to near-term changes in and the long-term trajectories of demand patterns. This new generation of demand planning is not a static, one-off exercise but an ongoing, data-driven practice. And it’s a foundational element for the more resilient supply chains required for an uncertain and volatile future.

Gain key insights by subscribing to our newsletter

Subscribe to our newsletter for up-to-date insights on technology information and connect with us on Twitter.

“In a steady state, what you did yesterday is important,” says Jonathan Wright, global leader and vice president of supply chain consulting at IBM Global Business Services. “In a volatile environment, what you did yesterday is next to useless. You need continuous intelligent planning – planning every week or even every day, and not just based on yesterday.” Companies in the fast-moving consumer goods category, for example, now combine their historical data models with real-time demand signals to improve forecast accuracy, says Wright, who works with a number of companies in this area.

The pandemic made clear the need for greater visibility and granularity into key demand drivers to enable organizations to quickly adjust their plans. Organizations that have begun to incorporate analytics are able to generate more frequent and accurate demand predictions. “Instead of executives spending time generating their own forecast and arguing about which forecast is more accurate,” says Simchi-Levi, “they let data and analytics generate a single forecast and an effective supply to satisfy demand and devote their time to strategic thinking.”

This approach sounds like common sense, but what became quickly apparent in 2020 was how few companies had achieved the mature digitization in their supply chains to make the most of such a demand planning approach. The fact was hidden by human middleware: many supply chain professionals responded to upheaval with overtime heroics that essentially made up for the deficiencies of traditional forecasting and planning. Such heroics are not sustainable, however.

Workers checking packages on conveyor belt in warehouse

Beyond the bullwhip

Ongoing mismatches between demand and supply has marked the COVID-19 pandemic.

What began with shortages of pasta and hand soap is now showing up as a lack of semiconductors necessary to make new cars and lumber to build new homes. Short-term behavioral shifts across the spectrum – from hoarding and panic buying on one end to a complete collapse then resurgence in demand for big ticket items on the other – illustrated the bullwhip effect writ large and clear. The extreme variations in demand exposed supply chain weaknesses upstream.

When lockdowns led to a crash in car sales, auto manufacturers cut their orders for semiconductor chips (of which some of today’s cars require more than a thousand). Savvy chip makers, noting surges in demand for consumer electronics, shifted their production to booming markets. Thus, when consumers got the itch to go new car shopping late last year, car makers couldn’t meet the resurgent demand.

“When demand surged in many product categories, manufacturers struggled to shift from supplying one market segment to supplying another, or from making one kind of product to making another,” Willy C. Shih, the Robert and Jane Cizik Professor of Management Practice in Business Administration at Harvard Business School, wrote in the Harvard Business Review.

“What happened in the last 18 months accelerated a process that companies thought would take five years to complete.”

David Simchi-Levi, professor of engineering systems at MIT

Consider the U.S. groceries market as another example: food makers with products and factories optimized for institutional buyers struggled to adjust to the drop in demand from restaurants and cafeterias at the pandemic’s onset, and the accompanying escalation in demand from stuck-at-home consumers. Few picked up on the earliest signals of this shift, but those with the flexibility to change their packaging were able to change their approach.

Horizon scanning lets leaders evaluate the risks of outlier events and prepare for their potential impact

Similar situations that played out in other industries pointed to the need to adjust demand planning for future disruptions. The pandemic “provides us perhaps with a foretaste of what a full-fledged climate crisis could entail in terms of simultaneous exogenous shocks to supply and demand, disruption of supply chains, and global transmission and amplification mechanisms,” noted a recent McKinsey Quarterly article.

As a result, companies from large appliance makers and high-tech manufacturers to consumer-packaged goods giants have been evolving their demand planning approach over the last year, according to Simchi-Levi.

That observation matches what supply chain leaders and analysts have been telling us: those manufacturers that aren’t already adopting a more modern demand planning practice will likely pursue a more proactive and predictive process within the next two years, to develop more resilient supply chains with the capability to resist or even avoid the impact of supply chain disruption – and the ability to quickly recover.

Much recent attention has focused on the supply side of the equation. But a key ingredient will be a more informed and up-to-date understanding of actual demand.

Close-up of woman's face and glasses

The role of data

Smarter demand planning that uses more and better data will soon be a mandatory part of an integrated business planning process. It will include tracking data signals indicative of trends on both supply and demand sides, simulating impacts on the supply chain, adjusting demand plans, and executing on them. The goal is to ensure that supply chains not only survive extreme disruption but are better aligned with emerging signals of demand from day to day.

Collecting more and better data sources that have direct and indirect impact on customer demand is where much of the focus is right now. That gives organizations a better idea of what’s actually happening – and likely to happen.

The good news is that the data to accomplish this demand planning process is all out there: point-of-sale data, social media feeds, market trends, macroeconomic indicators like inflation or unemployment, healthcare data, and more. For example, if you’re a luggage manufacturer, you can access current travel data to inform your demand planning processes. Over the past year, that data certainly proved more informative than historical sales trends.

In the retail and consumer-packaged goods industries, accurate forecasts of short-term buyer demand – a couple of months, or even a week out – are critical to ensure products are available to consumers when they most need them. Companies can combine customer mobility information, shopper sentiment, point-of-sale data, and current orders along with historical data, and then use machine learning to create more frequent and accurate demand forecasts. 

Consider how quickly and profoundly consumers’ snack-buying patterns changed last year. Leaders at Mondelēz International (which owns the Oreo, Triscuit, and Cadbury brands) had to determine how to meet the growing demand for cookies, crackers, and chocolate flying off supermarket shelves. “Going into the crisis, we thought all bets were off. There was no good historical data and no clear path forward,” Orkun Ozturk, who heads the demand and supply planning team for Mondelēz in Europe, said in an August 2020 webinar.

The company’s leaders concluded that market intelligence from its special situation management teams (including a global team and one located in each country and plant) would be more valuable than the usual forecasting data.

Smarter demand planning, using more and better data, will soon be a mandatory part of an integrated business planning process that includes tracking data signals indicative of trends on both supply and demand sides.

“It’s very important to have real-time data and use it quickly,” Ozturk said. “The last thing you want is to get it late or get it wrong.”

Using machine learning to analyze point-of-sale data, syndicated epidemiological data, social media, and local economic and mobility information to predict changing demand patterns can enable companies to better anticipate quick shifts in demand. By feeding demand signal data into machine models at the Mondelēz COVID-19 command center, the company was able to eliminate historical biases and respond more quickly to changes in the marketplace. Using AI to analyze a wide variety of demand signals (for example, the impact of stimulus checks on cookie demand) has helped Mondelēz reduce out-of-stock items and retain millions of dollars by avoiding lost sales and supply chain inefficiencies.

Those organizations that want to effectively do this kind of demand planning synchronize their processes to perform long-, medium-, and short-term planning. In the past, short-term planning was done, at most, once a month. Now some companies are revisiting demand plans weekly or even daily.

This kind of frequency is trickiest for business-to-business companies who may be two tiers deep in a five-tier supply chain, but they too are seeking greater visibility into demand signals from their business customers (and their customers) for better demand planning.

Woman brainstorming

Process changes for better demand planning

Companies that do not have all the data they need to better anticipate shifts in demand will need to locate or collect it. Then they must synchronize it. Collection and integration of more and better data for demand predictions, however, is just a start. To get better at anticipating and meeting demand, manufacturers need to make a number of process adjustments.

First, they will need to redefine demand planning itself. Monthly demand planning is likely a thing of the past. Most companies will need to plan, replan, and plan again based on their ongoing analysis of demand signals and determine how to respond to sudden shifts informed by supply chain signals. And they’ll need to do so at a more refined level.

They will also have to adopt systems capable of planning at different levels of granularity (at the product SKU level, for example, versus the product family level). Leading manufacturers are looking to leverage new technology-enabled approaches to better sense and respond to demand shifts. By leveraging AI and machine learning across broader and richer data sources, they can better analyze demand signals, generate insights, and drive scenario planning to optimize their supply chains on the fly. And the cost of these technologies is decreasing, making this automation more accessible to a wide variety of companies and making it easier to generate a strong return on investment, says Jim Kilpatrick, global supply chain network operations leader for Deloitte Consulting.

At IBM, Wright has been working with companies to create intelligent workflows for demand sensing that are self-learning and self-correcting. With automation, organizations can collect and analyze more current data to provide the hyper-localized visibility necessary to expand or contract manufacturing and distribution as needed.

Synchronizing demand intelligence within the supply chain will also be necessary. It would be of little value if, say, a washing machine manufacturer determines there will be a global spike in demand for its products, but its tier-three suppliers don’t hear about it for a month. Or ever.

Multi-enterprise collaboration will also be very important, leading manufacturing partners to invest in peer-to-peer networks to share demand forecasts and inventory levels.

Implementing these new systems will all require significant organizational and cultural change for many manufacturers, most of whom still put a great deal of human effort and ingenuity into the demand planning processes. They will need to not only adopt these systems, but also become adept at using them – and trusting them.

Then – most importantly – companies must be able to adjust business or manufacturing processes to respond to demand signals.

Power tool in use, sparks flying

Manufacturing adjustments sparked by demand signals

When Honeywell saw its demand for box scanning equipment surge along with e-commerce activity, it was in a tough spot. The sensors required to produce them had become scarce. So, as the company’s chief supply chain officer told Bloomberg, Honeywell redesigned its scanners to accommodate different sensors. Likewise, Caterpillar, seeing booming demand for its heavy machinery, has had to rework its manufacturing practices and materials specifications to meet demand amid shortfalls in hard-to-source resins required to produce them.

In some cases, however, it may not make financial sense to invest in new capabilities or approaches to meet a surge (or drop) in demand. There will remain an art and a science to demand planning. The science is the analytics. The art is in determining how to apply the insight that the analysis produces.

Some of this will take time. But the future of demand planning will not wait – nor does it have to. Manufacturers can integrate and analyze more demand signal data as a start, a process that it turns out can take mere weeks or days and not months. Some companies have seen significant impact simply from assembling relevant data into a single depository.

This data-driven demand planning approach can become the newest (and perhaps most important) category of collaborative enterprise planning for manufacturers since the vast majority of their costs are tied up in their supply chains.

“What happened in the last 18 months accelerated a process that companies thought would take five years to complete,” says Simchi-Levi. “There have been so many significant disruptions to the supply chain. It had a huge impact on business and people. But there is a silver lining. It showed us what the future of the supply chain and demand planning is. And it taught us that the future is here.”