You’re reading this right now because you likely know there’s something you could improve about your demand planning and forecasting process. In the end, it’s all about gaining the know-how and adopting the perspective you need in order to level up. We’re here to provide you with a set of tools, insights, step-by-step techniques, and executable strategies that will help you:
- Crystalize your outlook of the CPG landscape, including the residual effects of the coronavirus pandemic, including its effect on the global supply chain;
- Take action and realize quantifiable benefits across your supply chain, such as bridging data gaps or adding transparency to gain consumer trust;
- Achieve your demand planning goals, including raising accuracy, lowering error, and mitigating risk;
- Hit important KPIs, such as increased service levels and new distribution; and other important steps unique to your organization’s growth.
Without further ado, here’s a list—we like to call it definitive—of the 21 best practices for demand planning and forecasting professionals in the CPG industry (and beyond).
1. Know why you’re here.
This is always a solid place to begin any assessment and learning endeavor: What are your goals? Perhaps you:
- Are ready to push your S&OP process to the next level, making it more efficient and effective.
- Struggle with forecast accuracy and want to learn more about bias and error benchmarking.
- Are reeling from the impact of the Covid-19 pandemic and want to make sure you don’t repeat the same mistakes twice.
- Desire a better understanding of consumer purchasing patterns and the variables that influence demand.
- Want to learn about the availability of advanced tools that are built to handle the complexities of modern-day demand planning, including artificial intelligence.
- Like many CPG executives, simply feel like you’ve been shooting in the dark lately when it comes to supply chain planning, from the manufacturing floor to the S&OP meeting.
Whatever it is, write it down, record it, share it. Seek advice. And rest assured, you’ve come to the right place.
2. Act with a real sense of urgency.
It’s vital to understand what’s impacting the daily lives of people around the world, even on the most basic levels. What’s the day-to-day like for consumers around the world—economically, socially, professionally, personally? How have recent events changed their lives, and what impact is it likely they have on their future? How have consumer purchasing patterns changed given the circumstances? Exploring and following these informational trails is key.
Consumption-driven demand forecasts and the algorithms that support them should be informed by a deep knowledge and analysis of these variables.
The aforementioned questions do not necessarily need to be answered this very second. Rather, the most important aspect to take from this consideration is the fact that Covid-19 has changed the way the world, and therefore the consumer, functions—perhaps permanently. At the very least, the pandemic has accelerated changes that were already in motion, such as the solid emergence of DTC-native brands and widespread adoption of eCommerce channels, each with its own unique fulfillment practice and typical buyer journey.
For example, three models in the customer purchasing journey in the digital grocery eCommerce landscape include:
- Pure-play/ship-to-home (via a service like Amazon, the customer is enabled to complete their entire purchase online);
- Last-mile (services like Instacart complete the at-home delivery); and
- Click-and-collect (the customer picks up a purchase they began online)
The lesson here is that consumer behavior and demand patterns are inherently volatile—perhaps more than ever before. The consumer can be fickle because life, and the conditions that influence it, can be fickle. Demand can drastically change in an instant, and it’s your job to help your organization keep up as much as possible by staying on the pulse of consumer purchasing patterns.
Your sense of urgency should stem from this very fact. When the variables that impact consumption change, how quickly can your company adapt to new, in-motion realities? What technologies or processes does your company have in place that enable such agility?
If you’re not asking these questions now, your sense of urgency needs to be dialed up.
3. Examine your current demand planning and forecasting processes.
Now it’s time to take a good look in the mirror, so to speak. As much as time and resources allow, execute an internal (or external) assessment of your current demand planning and forecasting process, from top to bottom. Ask questions—lots of ‘em.
Mirror, Mirror: 17 Essential Demand Planning Assessment Questions
These questions should set you off on the right path of exploration. We’ll dive deeper into each of them below and examine plenty more, too.
- How much time does a typical demand forecasting process take and with what frequency does this occur?
- Who’s involved in your demand planning or S&OP process, and how much time and resources does it take to accomplish both individual and team goals, as well as align on a demand forecast?
- What tools are you using to forecast demand? When were they implemented, and what has changed since?
- Are these tools effective? Are they holding you back? For example: If you’re using spreadsheets or a comparable system, can it handle the amount of data and advanced mathematics you wish to employ?
- Understand the methods you use to calculate demand forecasts.
- Is your data clean or is it incomplete, mismatched, or even erroneous?
- Is data siloed between your teams, thereby creating misalignment(s) in the demand planning process from the get-go?
- Are you relying on historical sales data or are you implementing data that more closely reflects the present-day consumer, such as POS data or other market signals?
- What kinds of external datasets would you ideally like to include in your demand forecasting calculations, and how can you obtain them?
- Are you calculating forecasts at the SKU-, customer-, and warehouse- levels?
- How consistent are your predictions? What are your average accuracy and mean deviation?
- Have you been able to identify key drivers of demand?
- How accurate are your sales and inventory predictions?
- Have promotions had a direct increase on sales? How about competitive spend?
- Are you measuring for error and bias?
- How agile are your models? Are they trained to quickly recover from unprecedented spikes or dips in demand, such as the waves we experienced during March and April of 2020?
- Are your forecasts dynamic and insightful? And are they actionable? In other words, are you able to optimize working capital and realize cost savings based on the recommendations your data or demand planning provides?
4. Identify your top-line needs.
Asking key questions is just the start; you also need to do the work to answer them.
Once you’ve had the chance to thoroughly examine your demand planning and forecasting process, it’s important to identify the areas that need the most attention and begin to consider potential solutions.
In practice, a thought exercise might look like this:
- Where do my inefficiencies lie? Am I constantly wasting capital by holding too much inventory? Is this the result of inaccurate demand forecasts, or something else?
- Perhaps you see that you might have some work to do on the forecast accuracy front— balancing supply with demand is the battle to gain an upper hand in.
- Or perhaps you’re spending too much on repositioning costs and want a system that helps you prioritize production based on which SKUs have higher margins, as well as give you granularity as to warehouse supply levels and demand near those ship-from locations. How can you get that done?
- Or maybe your sales team is spending far too long on their projections and you want a cloud-based system to automate that process and free your salespeople up… the list goes on, as it should.
Don’t stop until you’ve determined your most pressing needs. Then:
5. Prioritize them.
6. Commit to embracing a fresh demand planning philosophy.
In our guide “The New Rules of Demand Planning,” we detail why a fresh approach to demand planning is vital in today’s CPG industry. Below is a breakdown of the ways in which you can take on this new perspective—a list of demand planning commandments for 2020 and beyond.
7. Embrace the consumer.
To understand the future, you must embrace the present. No longer should you treat historical shipments, or supply, as the end-all-be-all of your forecasting and demand planning strategy. Forecasts should be driven by the consumers and the variables that influence demand.
8. Embrace (better) data.
Data is the foundation of your demand forecasts. Your data should possess two basic qualities: cleanliness and robustness—and in that order, too. “Bad” or “dirty” data describes information that’s tainted by error: incomplete and inaccurate. While you may have all the data your servers can handle, some of it may be unusable because it contains mistakes. Fix these first.
9. Embrace dynamism.
Dynamic data and insights do not just appear; they must be scientifically ascertained. The analysis of two or more datasets, as outlined in our guide, The New Rules of Demand Planning and Forecasting, can yield transformational insights that can be acted on. They include:
- Historical sales and seasonality. Raw day-to-day sales patterns contain a treasure chest of valuable information, including the relationship between seasonal trends and long-term growth.
- Historical inventory. This information is vital to understanding the impact of inventory on sales, and the relationship there. For example, analyzing historical inventory levels—quantity on hand, quantity on order, weeks of supply, service levels, etc. more—is key to understanding how insufficient inventory causes a company to miss out on potential sales.
- Product transitions, maturity. Data about a product’s lifespan, such as the information trail of SKU numbers and/or UPC codes, is vital to the process of establishing and analyzing time relationships between SKUs and UPCs. Other aspects include the size of a product’s label or package, or any other bit of data that’s usable: If it has the potential to impact the business, it should be incorporated in this analysis.
- Obsolescence and discontinued products. Data that details product sunsets is key to understanding the demand of comparable products. It’s essential to understand the reason as to why a product is discontinued—component innovations, new claims, product enhancements, poor performance, for example. These details are essential to better predictions.
- New products and distribution data. Product launches, increased shipments, as well as distribution gains are vital to forecasting modeling. It's tough to accurately predict demand for new products unless it is known what there is to predict, so this data should be removed from future forecast calculations. Distribution change models can better predict the sales impact of distribution gains in the base business (pipelines from everyday consumption), distribution losses, and new product distribution gains. It's also vital to understand in-store dynamics, such as the number of facings, shelf placement, multiple placements, shopper marketing activities, and competition placement.
- Marketing investment data. Analyzing the numerical relationships between promos (own and competitive) and changes in sales volume and velocity (lift/drop) is vital to enhancing demand forecasting algorithms and accuracy.
10. Embrace automation.
Automation is all about letting a process work for you—and for your team—in the background. It’s like putting your forecasting on cruise control. While the technology does the work, you and your teams can focus on other important tasks and goals. As they say, time is money.
As an example, learn how automating demand forecasts has been incredibly beneficial to The Fishin’ Company, the largest importer of tilapia in the world and one of the largest importers of frozen fish. Automation has helped Fishin’ Co. streamline their S&OP process, making it more efficient. They’ve also been able to increase in accuracy using Demand Planning AI, helping them to save millions in yearly inventory holding costs.
11. Embrace the moment.
12. Embrace Demand Planning AI.
Simply put, CPG companies must arm themselves with the tools that enable them to quickly and adeptly maneuver their business through supply disruptions and drastic changes to consumer behavior. Now more than ever, organizations should be looking to incorporate the kinds of technological advancements available to them in order to prepare for unforeseen or obstacles that may arise and throw off commercial plans at a moment’s notice.
Demand Planning AI is a tool that was engineered to meet this moment—specifically for CPG companies. Machine-learning models are programmed to deeply and continuously analyze the complexities and ever-evolving dynamics of the CPG industry. Statistically speaking, these dynamics represent countless internal and external datasets, including historical shipments, point-of-sale data, and other variables that impact demand. Obtaining this amount of information is tough enough, to say nothing of cleaning it and analyzing it. This is the job of an AI, not a human being (or even an army of them).
Additionally, Demand Planning AI will automatically adapt (rather than react). What does this mean and why is it a vital element of an effective and efficient demand planning process?
Well, take Covid-19, for example, which brought about fast, unprecedented changes to consumer purchasing patterns. In March and April of 2020 especially, many CPG companies struggled to produce and/or hold enough inventory and meet unprecedented spikes in demand, resulting in lost sales. Furthermore, in some cases, by the time some companies caught up, demand unexpectedly ceded back to normal (or less elevated) levels, causing capital to be tied up in excess inventory and warehousing costs.
Demand Planning AI is effective because of its ability to proactively treat Covid-19, and other abnormal events, for what it is at the most basic level: data. The pandemic is an outlier event that brought about unprecedented changes to consumer purchasing behavior and the CPG industry at large. And its impact will continue to reverberate for the foreseeable future. Ultimately, however, the pandemic is a small dataset in a pool of variables.
As a result, Demand Planning AI can help CPG companies to recover quickly from unprecedented events like a global pandemic. Read here to learn how Demand Planning AI corrected itself during the peak of Covid-19 to help a category leader optimize inventory and meet demand in just a three-week turnaround.
13. Assess the data you’re currently using to forecast demand.
The “assess” portion of this best practice is actually a two-step process. First, you should determine the data you’re actually using to calculate demand forecasts. If it’s internal data, determine exactly what you’re working with. How granular of an understanding do you have into your historical shipments? How far back do your numbers go? How much data do you have? And so on.
The second part of this process is to determine which data you wish you could include in your calculations. Ask yourself: What’s missing? Some datasets you might be well aware of but don’t have access to (or the time or technological capabilities to obtain it); others you simply might not know about.
14. Know that when it comes to data, size is no longer a barrier.
The size of your organization’s data pool should not present an obstacle to any part of your demand forecasting process. In the past, this may have been a hindrance to companies, but this is no longer the case at all. This sort of technological infrastructure (enterprise-grade servers, e.g.) is widely available. In short, bring on the data! The more the merrier.
15. Understand why “clean” data is essential.
Consider the adage, Garbage in, garbage out. If you’re inputting “dirty” data into your demand forecast calculations, you’re likely to end where you began—on the wrong foot.
When your data isn’t clean—meaning that it’s incomplete, inaccurate, not up-to-date, or mistake-laden—then your demand forecasts are relying on erroneous data to begin with. This is a recipe for disaster.
16. Automate the data-cleaning process (and free up the humans).
Whether your company has 5 SKUs, 2 warehouses, and 20 ship-to locations, or you distribute 10,000 SKUs worldwide, the data-cleaning process is much, much easier using machine-learning algorithms than human power. For instance, data engineers can create and apply mapping algorithms to large sets of raw data, thereby automating the cleaning process and completing it in a fraction of the time it would take a human to accomplish similar tasks. Manual data analysis is incredibly time-consuming and heavy on process. This may include labor-intensive tasks like "sourcing, scrubbing, collating, and formatting activities that leave little time for actual analysis." ("Machine Learning in Supply Chain Planning—When Art & Science Converge," Journal of Business Forecasting, Spring 2019)
17. Know which datasets are vital to sales and inventory predictions.
18. Understand the availability of different types of datasets.
Taking a fresh perspective on demand planning and forecasting means not relying solely on historical sales data (shipments at the SKU-level by customer, ship-to, and ship-from locations). This data alone cannot account for rapid shifts in consumer behavior and other microeconomic indicators.
And in today’s economic landscape, where eCommerce is booming and DTC-native brands are gaining ground on market share from legacy brands, it’s vital to have a complete picture of the variables that can drive demand. Major datasets include:
- Point-of-sale (POS) data
- Internal data
- Geographical patterns
- Category, health, and ingredient trends
- Own and competitive marketing and trade investment (promotions, e.g.)
- Search engine data (Google.com, Bing.com, more)
- Social media data (Facebook, Instagram, Twitter, Snapchat, TikTok, LinkedIn, and more)
- Weather patterns
- Seasonality (including slow-moving products with regular, albeit intermittent, demand)
External datasets, including some of the above, are widely available, sometimes for a fee and sometimes not. A few free online repositories are: Google’s Public Data Explorer directory, Data.gov, Google Trends, the U.S. Census, GitHub (for developers), FiveThirtyEight, Kaggle’s datasets, and this compendium from Eastern Michigan University. And, of course, there are free resources from market intelligence firms like IRI Worldwide (its Covid-19 economic indicators page is well worth a visit), SPINS, and Nielsen.
19. Embrace data science.
20. Demand granularity.
21. Use AI and Automation to realize efficiencies in your S&OP Process.
Understanding Data Science
Data science has become one of the most sought-after applications of the 21st Century—a time when there is more ascertainable data produced than ever before. Simply put, data science is the framework through which robust analytics can be executed.
The discipline itself is a riveting cross-section of expertises, combining some level of mathematics and statistical aptitude, computer science (knowledge of database languages like SQL or Python; data governance and infrastructure), and certain domains of professional knowledge, such as criminology, sociology, and political science—or, in this case, the Consumer Packaged Goods industry and supply chain. But it's where these three disciplines overlap and are applied that truly illustrates the impact data science can have on a CPG company. This includes data engineering, machine learning, and research and analysis.
The work of data scientists is incredibly interdisciplinary, which is part of what makes it so effective.
The Benefits of Applied Data Science on Forecast Accuracy
Over time, forecasting methodologies have of course changed. While many forecasting fundamentals remain, technological advancements and the modernization/digitization of commerce and the supply chain have completely changed the demand planning game.
It's essential for any CPG company to be able to implement, test, iterate, and hone a number of algorithms that can efficiently and effectively analyze data. These consumption-based models should be able to predict demand with high accuracy across a number of time horizons; measure for error (WMAPE); and forecast sales at the SKU-level by customer, warehouse, ship-from, and ship-to locations.
The best demand planning platforms or AI technology go beyond insights, however, providing recommendations that pinpoint exactly where and how companies can realize efficiencies across the supply chain and business at large. For instance, the best AI platforms should have the ability to transform raw data into actionable insights, which themselves become benefits. This includes:
- Priority rankings of SKUs based on profit margins, enabling optimal manufacturing output and warehouse supply levels.
- Recommendations regarding the most efficient ways to transfer or reposition inventory based on demand levels by ship-to location for optimal production output and inventory levels.
- Deep learning models self-adjusting to new market realities, taking variables into account immediately.
And, this entire process should be automated—zero-touch forecasts available at the click of a button.
As we outline in our Ultimate Demand Planning guide, the process of streamlining S&OP meetings is all about realizing the benefits of automation and forecast accuracy. While the right demand planning process varies for each company, the collaborative goal of reaching a consensus and planning according to projections remains relatively the same. Automating your demand planning process, as well as implementing an consumption-driven artificial intelligence tool to achieve high levels of accuracy, can help you realize some of the following benefits:
- Increased efficiencies, from cross-functional teams getting on the same page faster to quicker decision-making.
- Data visibility, including a complete picture of sales and inventory data that's unified across the organization (to say nothing of access considerations).
- Invigorated sales teams & efficiency. When demand forecasts are automated on an AI platform, that means there is less pressure on sales teams and other departments to manually tabulate and analyze forecasts.
- Increased confidence. When executives are able to rely on accurate forecasts over a period of time, they tend to believe in them and the methodologies behind them. And for good reason. And when those forecasts translate raw data into actionable insights, leadership can make supply chain decision based on clear-cut information and recommendations. This might include increased marketing spend based on lift or investment in products with higher profit margins and velocity.
And there you have it. Twenty-one of the very best demand planning and forecasting practices. Now, though we do cover a ton of ground, this is not totally exhaustive—we'd need a library for that. And yet, for executives or demand planning professionals who wants to make headway in terms of understanding the ways in which they can use cutting-edge technology and methodologies to realize efficiencies across their supply chain, this guide will keep them plenty busy.