What is Supply Chain Forecasting? + 9 Best Methods (2024)

Supply chain forecasting and weather forecasts have more than one thing in common.

Both make predictions based on past and present information. Both use hard data and, at times, intuition to varying degrees of accuracy. And in both cases, something that didn’t appear on the radar can leave you feeling caught out and unprepared—whether that’s without an umbrella or without the inventory needed to fill an order.

Understanding how to properly forecast your supply chain needs is critical to ensuring your ecommerce store’s success. Getting it right can lead to better supplier relationships, increased customer satisfaction, and more capital to grow and scale your business.

Learn from supply chain management, fulfillment, and shipping experts to find out how supply chain forecasting can make or break your store’s next quarter—and what you can do to get and stay ahead. 

Table of Contents:

1. What is supply chain forecasting?
2. Why is supply chain forecasting important?
3. 9 supply chain forecasting methods
4. What is the best method of supply chain forecasting?
5. Supply chain forecasting challenges
6. Supply chain forecasting trends
7. Next steps to take with supply chain forecasting

What is supply chain forecasting? 

Supply chain forecasting consists of looking at past data about product demand to inform business decisions around planning, budgeting, and stock inventory. It can help a business prevent loss, especially during the holidays.

As its name implies, supply chain forecasting is based largely on analyzing supply. But customer demand also plays into it. Factors such as seasons, supply chain trends, the economy, and global events can all lead to spikes or sluggish sales, which can affect inventory control.

Why is supply chain forecasting important?

You don’t have to be a regular reader of the Journal of Supply Chain Management to know that timing means everything.

“To deliver orders fast and inexpensively, you need to have inventory in stock,” says Kristina Lopienski, director of content marketing at ShipBob, a global logistics platform that fulfills ecommerce orders for direct-to-consumer brands.

“Tracking inventory velocity over time involves being able to monitor bestsellers and stay ahead of production—even as demand changes.” 

Key factors may include:

• Turnover rate of products
• Lead times needed for each supplier or product
• Freight transit times
• Warehouse receiving times
• The cost of storage 

“If supply chain forecasting isn’t accurate down to a couple of weeks, it can cause costly ripple effects that will zap the profitability of an entire quarter or half year,” says Leandrew Robinson, general manager of mesh logistics with shipping and software company Auctane (which includes ShipStation, ShippingEasy, ShipWorks, and ShipEngine).

Products that arrive late to your warehouse or shipping center won’t make it to customers in time. This doesn’t just damage your brand’s reputation, it leads to a loss in sales and digs into customer acquisition costs (CAC). If you don’t have it in stock or it’s on backorder, your customers will go elsewhere.

“Many brands go out of stock during their biggest sales of the year, so they’re spending money on ads to create demand to then find themselves unable to convert that demand. This drives CAC way up and negatively affects brand affinity,” says Adii Pienaar, founder of Cogsy, a forecasting operations platform for DTC companies. “Fast-growing brands tend to duct tape their operational issues as they arise, but patching up problems won’t scale.”

On the flip side, inventory arriving before you need it can lead to increased warehouse costs or losses if products have a short shelf life. It also ties up capital, which could otherwise be used to scale or improve different aspects of your business.

And if you order the wrong amount or the wrong products? You may be left with deadstock. 

“Stale inventory sits in a warehouse gathering dust and accumulating fees,” says Nicholas Daniel-Richards, co-founder of ShipHero, which offers warehouse management software and shipping solutions. “The only way to salvage such situations is by selling at cost or at steep discounts, or selling in bulk to clearance houses.”

9 supply chain forecasting methods

There are two forecasting methods:

• Quantitative forecasting, which uses historical data to estimate future sales. These methods work largely on the assumption that the future will mirror the past, and involve complex mathematical formulas, which are typically performed by computer software.
• Qualitative forecasting, which is best for when data is nonexistent or hard to come by. 

Explore the different methods below to see what’s best for you. 

1. Moving average forecasting

One of the simplest methods for forecasting, this method examines data points by creating an average series of subsets from complete data.

As it’s based on historical averages, moving average forecasting doesn’t take into account that recent data may be a better indicator of the future and should be given more weight. It also doesn’t allow for seasonality or trends. As a result, this method is best for inventory control for low-volume items.

A bookstore might use a three-month moving average for predicting demand for a steady-selling book, where each month’s forecast is based on sales from the past three months. This wouldn’t work for seasonal items like calendars, which sell out at certain times.

• Pros: Easy
• Cons: Doesn’t allow for seasonality or trends
• Best for: Low-volume items
• Type: Quantitative

2. Exponential smoothing

Picking up where average forecasting leaves off, this method takes into account historical data, but gives more weight to recent observations. It’s similar to adaptive forecasting, which takes into account seasonality.

Variations on exponential smoothing including Holt’s forecasting model (sometimes called trend-adjusted exponential smoothing or double exponential smoothing) and the Holt-Winters method (also known as triple exponential smoothing), which factors in both trends and seasonality.

For instance, a fast-fashion retailer might use exponential smoothing to forecast clothing sales because it lets them focus on the latest trends, adapting quickly to consumer preferences.

• Pros: Easy; takes historical and recent data into account
• Cons: Can be prone to lag, causing forecasts to be behind
• Best for: Short-term forecasts or non-seasonal items 
• Type: Quantitative

3. Auto-regressive integrated moving average (ARIMA)

One method that fits within the ARIMA category is Box-Jenkins. ARIMA stands for auto-regressive integrated moving average. It uses time series data based on past performance to better understand the data set or to predicate future trends. Costly and time-consuming, this time series forecasting method is also one of the most accurate, although it’s best suited for forecasting within timeframes of 18 months or less.

ARIMA could be used by an ecommerce brand to forecast sales in the 18 months leading up to a major product launch. In this way, the brand can allocate marketing spend and prepare the supply chain.

• Pros: Very accurate
• Cons: Costly; time-consuming
• Best for: Time frames of less than 18 months
• Type: Quantitative

4. Multiple aggregation prediction algorithm (MAPA)

A relatively new method that’s specifically designed for seasonality, MAPA smooths out trends to help prevent over or under estimating demand. Although not nearly as popular as Holt’s or Holt-Winters, research has shown it performs better.

With its ability to handle seasonality, MAPA is beneficial for forecasting fashion sales, which may be influenced by multiple seasonal patterns, including spring and summer collections, autumn collections, and cyclical trends.

• Pros: Prevents over and underestimating
• Cons: Still relatively new; not as proven
• Best for: Seasonal items
• Type: Quantitative

5. Bottom-up forecasting

This method estimates a company’s future performance and how it will work “up” to revenue. It considers a brand’s suppliers’ production schedules, then layers key growth assumptions and scheduled marketing events onto this solid foundation. This method results in a more accurate forecast compared to a top-down approach, with brands only ordering stock that will actually sell, in turn preventing the unnecessary tying up of capital.

“Brands can then bring this forecast to their suppliers to negotiate a discounted unit price or better ongoing terms,” says Adii. “Any predictability brands can offer manufacturers becomes leverage in the conversation. This way, brands lower their cost of goods sold and spend less to make each dollar of revenue. As a result, they become more profitable without raising prices.”

• Pros: More accurate forecast compared to traditional top-down approach (which fails to optimize for unit economics)
• Cons: Errors at the micro level may become amplified as they approach the macro level
• Best for: Scaling merchants
• Type: Quantitative

6. Historical analogies

Historical analogy forecasting predicts future sales by assuming a new product will have a sales history parallel to a present product (either one sold by you, or one sold by a similar competitor). A comparative analysis, it has poor accuracy in the short term, although may be more accurate in the medium and long term.

A company might predict the success of a new product by comparing it to a previous one. When launching a new video game, a publisher may compare it to a previous title with similar themes and market conditions to predict sales. It’s a good baseline estimate, but it doesn’t take into account market dynamics or consumer tastes.

• Pros: May be more accurate in the mid- to long term
• Cons: Poor accuracy in the short term
• Best for: Similar items
• Type: Qualitative 

7. Sales force composition

Sometimes called “collective opinion,” this method relies on the personal insights and opinions of experienced managers and staff, gathered as a team exercise. Panels of this nature typically have a poor to fair accuracy.

When introducing a new product line, a company can use the sales force composite method. The sales team, drawing upon direct customer interaction, may provide insights that are not obvious from quantitative data alone. However, this method is subject to biases and can vary significantly based on the sales team’s optimism or pessimism.

• Pros: Fairly easy to collect
• Cons: Poor to fair accuracy
• Best for: When quantitative methods aren’t feasible
• Type: Qualitative 

8. Market research

For an ecommerce brand to gauge the potential success of an upcoming product or feature, it could conduct online surveys or analyze previous customer feedback. The target market’s direct input can help tailor product offerings more effectively to meet their needs.

This research may include surveying, polling, or conducting focus groups for your target demographic.

• Pros: Provides insights into your target demographic
• Cons: Can be time and/or money intensive
• Best for: New product launches 
• Type: Qualitative 

9. The Delphi method

In this technique, individual questionnaires are sent to a panel of experts, with responses aggregated and shared with the group after each round, until they reach a consensus. Since the panel doesn’t collaborate, bias is eliminated from the process. 

This is considered one of the most effective and dependable qualitative methods for long-term forecasting. 

• Pros: Unbiased
• Cons: Slow, long process, leading to risk of subject matter expert dropouts
• Best for: Long term planning 
• Type: Qualitative

What is the best method of supply chain forecasting?

If you’re relying on spreadsheets, Adii says that using a moving average that focuses on recent sales velocity is your best bet. But if you’re using programmatic software, time-series methodologies are the most relevant, with the most popular ones being ARIMA, CNN-QR, Deep-AR, and Prophet.

“Their accuracy depends on the type of retail data they’re forecasting,” he says. “The best option here is to compare statistical significance and confidence levels of all those algorithms and pick the one that’s strongest for your data.”

But regardless of what method of supply chain forecasting you use, there will be inherent errors due to assumptions, so it’s impossible to achieve 100% accuracy—although you’ll generally find that much like the weather, short-term forecasts are more accurate than long-term forecasts.

There is one thing our experts agreed on, though: Qualitative methods rely on the opinions of consumers and market or industry experts, which are ultimately subjective and less accurate.

“The strongest method of supply chain forecasting is quantitative and trend forecasting based on hard data and analysis,” says Nicholas. 

Supply chain forecasting challenges

Changing regulations

Events of the past few years have made common knowledge what supply chain experts have long known: the global logistic network is incredibly vulnerable to political instability, natural disasters, and regulatory changes, all of which are now happening with increasing frequency and severity. 

Adii says this has caused brands to start diversifying their supply chains by working both on- and off-shore. 

“Building a supply chain to meet decentralized demand will be key to growth,” he says, noting that many merchants don’t just sell directly on Shopify—they also sell products on marketplaces such as Amazon and Etsy and natively on social media platforms.

 “There will be a shift from ‘supply chain management’ to ‘demand chain management,’” he predicts, adding that Cogsy is currently building a tool to give manufacturers more visibility and predictability in how the brand generates demand and sells.

Product returns

Free returns are now considered a cost of doing business, but they’ve also changed how customers shop. It’s not unusual for online shoppers to order multiple sizes, colors, or products, find the right fit, and then return the rest. 

Between Thanksgiving and January alone, millions of returns are made every year, amounting to over $171 billion in goods. Making returns easy is good customer service, but it can complicate supply forecasting. The percentage of products being returned and the reasons those returns happen can vary widely based on the product category you sell and seasonality.

Trends and changing demand patterns

Trends and fads come and go. Without sufficient stock, you can miss out on a surge in demand altogether.

For ecommerce merchants with brick-and-mortar locations, managing these demands can be even more complex, as customers can suddenly change the channels they shop on, making it difficult to predict where to stock inventory.

Matt Warren, CEO of Veeqo—which helps support ecommerce merchants in their omnichannel inventory management—says this is why retailers are increasingly turning to a hybrid online/offline approach. He cites the case of one of Veeqo’s clients, a large US fashion retailer with a big physical retail footprint:

“They used Veeqo to turn each of their stores into a mini-fulfilment location, allowing them to optimize delivery times for online customers,” Matt says. “They can also seamlessly marry stock level data with all their online/offline sales data, which enables a more sophisticated demand forecast. It’s the kind of innovative, hybrid online/offline approach to commerce that the industry has been talking about for a while.”

Seasonality of products

“Not factoring in seasonality and current events is one of the biggest mistakes I see ecommerce merchants making when it comes to supply chain forecasting,” says Leandrew. “It’s hard to react to a booming holiday sales period a few weeks before.”

Supplier or manufacturer lead time

Prior to founding Veeqo, Matt ran an online luxury watch retailer. His experience taught him that predicting demand was only ever half the battle.

“Each supplier—and sometimes each individual SKU—needs a different lead time,” he says.

In addition, it’s important to take into account warehouse and shipping lead times, which may be affected by overseas holidays.

Chinese New Year may slow fulfillments from China, while holiday peaks may cause delays or congestion at ports, slowing deliveries. This is where building strong relationships and communications with your suppliers becomes vital.

Siloed data

Matt also cautions that siloed data can affect the accuracy of supply chain forecasting. 

“Too many merchants use different software for different parts of their business. Add in working across multiple websites, marketplaces, and fulfillment locations and you can see where the headache comes from,” he says. “It’s worth either investing in all-in-one software to unify your sales and inventory data or putting the hard yards in to pull it all together via spreadsheets.” 

Skewed data

“Brands can’t create accurate forecasts with skewed data,” says Adii. “Merchants can infuse real-time data into their forecasting process to have a better idea of where they stand and where they can expect to be in the future. With better data in hand, they can chart a path that ensures they get there." 

Adii says that in order for data to be accurate, merchants need to avoid common inventory forecasting mistakes by:

1. Shortening the time it takes to update data in their systems
2. Avoiding changing SKU IDs for the same product
3. Taking inventory stock levels into account when completing demand forecasting
4. Identifying limited edition products to interpret their data accordingly
5. Linking demand for all versions of the same product
6. Analyzing each channel separately

Historical data isn’t enough 

“Quantitative methods that rely on historical data only are not reliable in fast and hyper-growth environments where most of our ecommerce customers are operating,” says Kristjan Vilosius, CEO and co-founder of Katana, which offers supply management software for makers and manufacturers. He makes the point that you’re better at making sense of events after they’ve happened.

“Investing in tracking and early warning systems and finding ways to make the supply chain management leaner and less dependent on stock levels is often a better investment, rather than trying to find the best forecasting methods,” he says.

Adii agrees, noting that many brands report struggling with the time it takes to create operational plans—which can cause delays in taking action, hindering a brand’s ability to capitalize on opportunities and mitigate risks.

 “The challenge with using time-series forecasting methodologies is that historical data often lags, especially in high-growth environments,” he says. “At Cogsy, we believe in additional future plans, such as marketing events, and assumptions or growth modeling, on top of a baseline forecast that was created by analyzing historical data. This creates the most holistic perspective on future demand.”

Supply chain forecasting trends

AI-assisted management 

This trend refers to the use of AI that actively assists in decision-making, not just by analyzing large datasets but also by learning from past decisions to improve future outcomes. 

Gartner refers to this as “Actionable AI,” defining it as the use of data to influence systems learning as a means of influencing decision-making processes based on how they adapt to new situations in real-time. As a result, AI can support supply chain operations in a more nuanced, context-aware way, making it a co-pilot in decision-making instead of just a data analysis tool.

Having “control tower” visibility

AI is being used to improve visibility across the supply chain, acting as a “control tower” for operations, as described by KPMG Global. 

Real-time visibility means being able to see beyond your own company and into your broader network of suppliers, partners, and logistics. Through predictive foresight and agility, AI-driven visibility tools can help you anticipate problems, respond to changes quickly, and collaborate better.

With these tools, you could track shipments, predict potential delays, and provide alternative logistics solutions proactively—plus facilitate collaboration with suppliers by sharing demand forecasts and inventory levels, enhancing overall efficiency.

Advanced analytics 

IDC’s supply chain survey highlighted advanced analytics and AI as the most important technologies for supply chains over the next three years. Modern supply chains are complex and generate a lot of data. AI can process, analyze, and extract meaningful insights from the data.

Artificial intelligence-driven demand forecasting tools could be used to predict future sales patterns based on market trends, economic indicators, and historical sales data, so you can adjust production schedules, inventory levels, and shipping logistics accordingly.

Next steps to take with supply chain forecasting 

When it comes to determining the best forecasting methods to use, you’ll need to consider a number of factors:

• What is the lifespan of the products? Are they perishable or can they remain on shelves in a warehouse indefinitely?
• How often are the products sold?
• How are sales affected by different seasons, months, and special sales events?
• What are the warehouse fees associated with a particular item?
• By what date do you need to reorder inventory for each product?
• What are your standard reorder points?
• Do you require safety stock?

“Supply chain forecasting shouldn’t be guesswork, but that’s the reality for many ecommerce merchants today. Online merchants need to understand the difference that real-time data and app integrations could make on their inventory replenishment capabilities,” says Nicholas.

“It’s the difference between being in-stock or out-of-stock, it’s the difference between having stale inventory or not, and it’s the difference between running a successful supply chain or not.”

Working with supply chain, inventory, shipping, and fulfillment experts can help keep you safe in stormy weather and simplify this process.

A full logistics service provider, the Shopify Fulfillment Network can help you build a resilient supply chain, with a vast network of strategically located fulfillment centers nationwide.

Veeqo, Katana, ShipHero, ShipBob and ShipStation are just some of Shopify Plus’ management and shipping partners who can help.

Illustration by Diego Blanco