Clarkston Consulting
Skip to content

Adapting Traditional Demand Forecasting to the New Normal

Contributors: Scott Shaw

We yearned for 2020’s unprecedented challenges to come to an end at the onset of the new year. However, the disruptions have rolled right into 2021 without faltering in magnitude. What does differentiate this new year, is that companies have more experience living through the disruptions. The learnings from 2020 should continue to be applied in a meaningful way to key business processes including demand forecasting. Traditional forecasting has been broken for quite some time and COVID-19 should be the final push that drives companies away from antiquated methods. The below strategies highlight how to join the leading organizations who have been investing in forecasting improvements long before the pandemic.

Value Beyond Demand Forecasting Tools

The complexity of the current environment makes it significant for demand planners to shift time from managing data to understanding the drivers of data. Demand forecasting tools can help reducing manual efforts and correlating alternative data sets to companies’ historical data. However, a demand forecasting tool alone will not effectively keep up with today’s dynamic environment. A demand planning tool needs to be complemented by analytics.

Seek Alternative Data Sets for Forecast Modeling Inputs

Traditional forecasting that relies heavily on historical sales over time to predict future demand can be unreliable in a volatile environment. A more dependable way to navigate today’s uncertainty is to leverage unconventional data sets and historical sales data – including 2020 data. Although there is an end in sight with vaccinations in progress, we are still living through the COVID-19 disruption with vaccines not projected to meet the masses of the United States until mid-year. This signifies that 2020 data is still powerful and should not be overlooked.

Alternative data sets for demand forecasting today can include data on COVID-19 reopening cycles, near real time sales data, Natural Language Processing (NLP) of social media, and IoT data. Determining the drivers of this data and how future demand can act in a similar manner may be the answer when historical sales patterns can become solely unreliable. Although there is an end in sight, the United States has until late summer before vaccines will reach 70-85% of the population per the New York Times. This signifies that 2020 data is still relevant as the COVID-19 disruption will remain through most of 2021.

Need help finding the right data sets? Depend on Clarkston’s analytics subject matter experts with experience finding, analyzing, and applying data sets to drive results. For example, Clarkston predicted competitor drug adoption for a leading biopharmaceutical company by extracting data from internal and external sources. Alternative data sets included prescribing patterns, competitor information, publishing data, information on doctors and institution information.

Ensemble Forecast Modeling Leads to Simplicity Through COVID-19 Pandemic

In today’s dynamic environment, combining multiple simple models through ensemble modeling is a more robust approach than having one complex model. Ensemble modeling combines predictions from different models to suggest an estimate. The models are combined and weighted based on importance. The inputs and models should continuously be analyzed to ensure accuracy of estimates. Having more transparency and simplicity makes it easier to regularly analyze the models and ultimately improve forecast accuracy.

Predicting demand in 2021 for nonseasonal products can combine the following models on top of historical sales:

  • 2008 economic recession by looking at how businesses performed during this crisis. It is projected that there is likely to be an economic recession so should consider this model
  • Short-term impact of COVID-19 by looking at the impact of infection rates, increased hospital beds and the number of the vaccines administered
  • Long-term impacts of COVID-19 by determining habit changes that will remain post-pandemic

For instance, consider applying ensemble modeling to elective surgeries that were halted due to COVID-19. HBR published that a study predicts the post-pandemic backlog for elective surgeries exceeds one million cases for orthopedic surgery alone. To forecast the demand for the number of elective surgeries in 2020 combine the following models:

  • Performance of elective surgeries in the 2008 economic recession should be assessed. A study from the American Hospital Association showed that about one-third of hospitals had seen either a moderate or significant decrease in elective procedures. This can be used in a model to predict how elective surgeries would perform if we enter a recession this year.
  • Then a model should be developed based on the short-term impacts of COVID-19. When there was a spike in infection rates how were elective surgeries impacted in 2020? When there was a spike in COVID-19 cases leaving hospitals overwhelmed elective surgeries were put on hold. What number of COVID-19 cases led to this halt and will COVID-19 cases reach this level in 2021?
  • Lastly, will COVID-19 impact long term habits for elective surgeries? Will alternatives to surgery such as rehabilitation be considered more frequently, will patients rethink surgeries going forward if symptoms subsided after waiting or will the patient surgical care experience be simplified?

These three models will be weighted to determine the demand forecast for elective surgeries in 2021.

Preparing for Future Black Swan Events & Approach for Highly Seasonal Products

Beyond the COVID-19 pandemic and for highly seasonal products, companies should consider investing further in advanced analytics. Advanced analytics solutions include Machine Learning, Artificial Intelligence and Data Science. Clarkston used advanced analytics techniques for a large US energy company to forecast incoming call volumes to their customer service center. This resulted in a 5% decrease in average forecast error from the original forecast.

Leave Antiquated Processes and Solutions Behind

Companies have seen their forecast accuracy degrade in 2020 which can lead to dissatisfied customers, stockouts, overstocking, and inaccurate production schedules. To avoid this risk, companies need to incorporate alternative data sets, ensemble modeling and analytics expertise into their demand forecasting process. Antiquated processes and solutions will not sustain the turbulent world we live in today.

Although Covid-19 may have exposed the fragility of the world’s supply chains, it accelerated the adoption of technologies and best practices that will have long term positive impacts for companies. If you need assistance with your demand forecasting for the new year and beyond, contact one of our subject matter experts today.

Subscribe to Clarkston's Insights

  • I'm interested in...
  • Clarkston Consulting requests your information to share our research and content with you.

    You may unsubscribe from these communications at any time.

  • This field is for validation purposes and should be left unchanged.

Additional contributions by Mikayla Doane and Courtney Loughran

Tags: Supply Chain Planning & Execution