As we are winding up another year full of disruptions and opportunities, businesses are already hard at work planning for the next year and beyond. Both in the budgeting and the S&OP processes, Demand Planning is usually the starting point and plays an essential role in establishing a resilient planning process. Many planning organizations have embraced scenario planning on the demand side as part of their resilient supply chain initiatives. Doing this manually results in more effort and can make the process of coming up with the ‘Approved’ plan even longer.
On the supply chain planning side, there is usually a single supply plan that corresponds to the ‘Approved’ demand, and the plan usually does not account for uncertainty on both the demand and supply sides. Of course, many of the market dynamics and operational realities are left for the S&OE process to resolve.
However, sometimes the S&OE lead times (usually 1-4 weeks horizon) may not be sufficient to make optimal decisions. At that point, everyone is left to make do with the short-term constraints such as budget, warehousing capacity, production resources, and raw materials already planned for. As one of the enablers of resilience in supply chain planning, Scenario Planning has been an emerging theme.
Embedding scenario planning capability in the S&OP process, both on the demand and supply planning processes, can help supply chain planning leaders create more agility in S&OP decision-making. As the types of scenarios get increasingly complex, having a digital platform and an advanced planning platform automatically generating different scenarios and demand sensing capabilities based on external data are becoming ‘must-haves’ for enabling fast and agile decision-making. Since scenario planning is still a ‘user-in-the-loop’ process, having a flexible and easy user interface is also essential in embedding it in strategic, tactical, and operational planning.
Taking this a step further, if the scenarios also come with associated probabilities, wouldn’t that be even better? Sounds good in theory, but how exactly would that work in practice? This vision is now feasible with the current advanced planning and AI/ML technologies.
In this blog, I’ll share some examples of probabilistic demand-planning scenarios that CPG manufacturing companies can automate with the help of new technologies. On the demand side, in addition to seasonal effects, variations in demand can be due to any or all of the following reasons, and each business should address the most impactful aspects in terms of potential revenue for their market dynamics when building their demand scenarios.
Changes in the Existing Retail Partner POS Locations and Distribution Channels
On the upside, retailers are opening more POS locations; on the downside, retailers closing stores or going out of business might significantly impact demand in the upcoming months. In 2023, at least 20 retailers announced closing stores, adding up to at least 2800 store closures. Forecasting models can account for the impact of the changing distribution channels by leveraging AI/ML models that learn from recent periods; to enable that, getting POS data from retail partners helps a lot, coupled with planogram data capturing display locations such as freezers, refrigerators, and shelves.
Also, through digital collaboration with the retail partners on the tactical plans, CPG companies can get more visibility into the future demand to plan their resources accordingly and meet their service level commitments. Without external data, the analytical models can generate year-over-year growth by channel to arrive at a range of growth values. Now, with the advances in AI models, potentially unstructured data such as news alerts and social media sentiment about retail partners can also captured and turned into valuable insights.
New Sales Channels and Accounts
Getting a heads up from potential demand from account executives working on new B2B contracts would help, wouldn’t it? However, not all deals may come to fruition, right? Assigning probabilities to new deals or rankings (HOT, WARM, COLD) and getting that data into a forecasting model through digital integration with CRM systems could help! These inputs can provide min/max values for the new business volume not part of the historical transactions. If the sales teams are overly optimistic, consistent bias can also be monitored and adjusted for by the models based on the observed conversion rates.
Changes in Product Portfolio
All CPG companies invest in developing new eco-friendly products such as green beauty to keep up with changing consumer preferences for more natural ingredients and sustainable packaging; demand is also shifting from existing products to new ones. As new products are launched, there are likely to be cannibalizing effects on the existing products; varying ranges for these effects, coupled with marketing initiatives and social media interactions, can drive different demand scenarios for new products and existing products.
AI/ML technologies available today model these impacts and even create Best/Worst/Most Likely demand scenarios with probabilities based on past occurrences.
Impact of Promotions
In the pre-pandemic era, about one-third of CPG products sold in the U.S. were sold on promo, and over the years, the effectiveness of the promos has been on a declining trend. During the pandemic, promotional activity decreased due to supply chain challenges and inventory constraints. Increases or decreases in promotional budgets and the types of promotional vehicles deployed impact demand. As the supply chain issues are resolved and availability increases, there will be an opportunity to ramp up the promotions on CPG products once again.
AI-driven scenario planning around promotional spending will be one of the critical capabilities contributing to optimizing topline growth and profitability. At the category or product group level, the types of offers being offered, featured in the promotions, and how they are displayed in the stores or the websites can all impact the sales of specific products.
Even if not all the variables are known in advance, and as the price elasticities continue to evolve in the post-pandemic era, it is wise to keep working with ranges of promotional effectiveness, covering best, worst, and most likely case scenarios generated automatically. There are also pre- and post-promo impacts that can be modeled in terms of ranges, and these can also impact the featured products or other related product sales in the periods leading up to or after the event.
Impact of Weather
We have been experiencing more and more extreme weather events across the globe in the last 30 years. In 2023, the U.S. experienced the sixth ‘warmest’ January on record. There were also very high levels of rainfall, and California alone experienced ten storms that caused floods and landslides. February was also an extreme month, with winter storms in Southern California while the rest of the country experienced a hot winter. These events and many others that followed impacted consumer behavior and supply chain networks. Some of these directly impacted the consumption of some products, and some impacted the availability due to supply chain disturbances.
As the temperature prediction models indicate, global warming effects will continue to impact us in the upcoming years and beyond. As these deviations from seasonal norms become more common, businesses need to consider these when generating baseline forecasts and future demand scenarios. As with all the other demand-shaping events, such anomalies must first be addressed in creating the baseline demand by normalizing the sales history. Advanced forecasting systems using AI/ML models can automatically perform such normalization activity. At the point of scenario planning, AI models can help generate different forecast scenarios for normal temperatures and extreme hot and cold cases. Of course, the impact of weather is not just limited to temperature; rainfall, snow storms, amount of sunshine, etc, can also influence the consumption of some products.
Impact of Price Changes
In 2023, inflation and rising prices were a concern for consumers worldwide. On average, CPG prices have risen 13% over 2022. With the onset of recession, consumers are also concerned about how to best allocate their wallet spending, leading to demand shifts across product categories. The same concerns will carry over to 2024 as well, which will lead to different pricing scenarios for CPG manufacturers. Certain product categories and brands are more sensitive to price changes than others. Brands need to consider the varying levels of recession across countries while making pricing changes in each country.
This is an area where AI/ML models can help as well. Changes in demand in response to changing prices can be modeled using price index elasticity models, considering how prices in each category have been changing in comparison to the competitors and how that has affected the demand.
The creation of demand scenarios corresponding to different price changes by brand, category, and region can be automatically generated to enable faster decision-making, continuously considering the latest trends.
Driving S&OP Decisions with Probabilistic Planning
Taking into consideration all of the work that has to be done on the historical transaction data to arrive at a baseline forecast that is unbiased and normalized for all the different demand drivers discussed above, as well as accounting for the probabilistic distribution to arrive at a range of values rather than a single deterministic forecast, doing this manually can be a daunting task. Luckily, we now have the technology to automate sales data pre-processing and generating forecasts with various possibilities.
The planners can define the scenarios most relevant for their business and have the advanced planning systems do the rest. Of course, demand planning by itself does not drive any decisions. The next step is to reconcile these demand scenarios with supply plans and assess the areas where there might be problems in raw material availability, production and storage capacities, or logistics constraints.
To enable resilience, AI-driven inventory targets and production plans must cover the range of probabilistic demand scenarios on an ongoing basis, constantly reconciling plans against forecasts and actual production against the targets and automatically generating prescriptive plans to meet business goals covering the range of possibilities.
Ultimately, to drive data-driven decision-making, all the pros and cons of these scenarios, in terms of revenue uplift and cost implications, need to be addressed as well and presented in a user-friendly manner so business leaders can quickly assess the impact of changing dynamics both on the demand and the supply side and make smart decisions.