11 Sep How to Forecast Demand in Retail (2021)
Demand forecasting has become vital to the survival (and growth) of many retailers in the last few years. That’s because top-performing retailers simply cannot rely on inaccurate, decades-old approaches to forecasting demand.
In order to optimize their inventory investments and maximize GMROII, today’s retailers need accurate demand forecasts for every SKU in every store. But achieving this level of SKU prediction requires a big data analytics approach to demand forecasting — designed for the digital era of retailing.
In order to understand how to take full advantage of this new technology, we will first explain what demand forecasting actually is, the most common methods of demand forecasting in retail, key constraints that need to be accounted for, and modern examples of accurate demand forecasting.
What is Retail Demand Forecasting?
To understand retail demand forecasting, you must first understand demand and forecasting as separate concepts.
Demand is an economics concept that describes the willingness of consumers to purchase a specific product at a specific price.
Forecasting is a statistical process that uses existing data to predict future performance.
Therefore, demand forecasting is a statistical prediction of the willingness of consumers to purchase specific products, at a specific price, in a specific time frame. Put plainly, if retailers could predict this accurately, they can easily minimize costs and maximize profits by:
- Setting accurate financial and merchandise targets
- Choosing optimal prices for every SKU
- Stocking enough inventory to maximize sales without overstocking
This may sound like something that retailers have been doing forever — but this is not the case. That’s because many retailers start with the wrong question when attempting to forecast:
“How much do I need to sell?”
Starting with this question leads to bad decision after bad decision.
For example, most retailers order too much of the wrong products and then rely on discount promotions and clearance pricing to fix the resulting inventory problems. Or they order too little to satisfy true demand, which leads to out of stocks and lost sales.
Those problems rarely occur when retailers start by asking the right question:
“What is the true customer demand for all of our products?”
Starting here leads to much better decisions.
By focusing on demand, retailers can get the right amount of the right products to their stores, cutting costs, improving their sales and profits in the process.
But to make sure you’re focused on the right question, we need to clear up to major misconceptions with demand forecasting:
- Sales forecasting vs. demand forecasting
- Forecast vs. plan
Sales Forecast vs. Demand Forecast
A common misconception is that demand forecasts and sales forecasts are the same.
While sales forecasting tools can be sophisticated, they rely mostly on past sales to extrapolate trends into the future.
These tools cannot calculate the sales you lost because your stores ran out of stock. For example, if you sold 100 units because you only had 100 units in stock, the sales forecast won’t reflect the 250 units you could have sold. (Your true past demand.)
Sales forecasting creates a self-defeating prophecy, because you will continue to understock high performing products. Demand forecasting techniques go beyond simple trend extrapolation, accounting for hundreds of factors that influence demand for each SKU in every channel (price, events, product families, assortment, product cannibalization, etc.).
Demand Forecast vs. Demand Plan
People often use demand planning and demand forecasting interchangeably. Though they are closely related, retailers must distinguish between forecasts vs. plans.
Demand forecasting is, as mentioned above, the process that generates forecasts of customer demand. This process will analyze industry trends, variables that affect a product’s demand, sales histories, and more in order to project how consumer demand will look like in the future for any given product.
Planning in retail, on the other hand, is the process that converts those forecasts into actions. Breaking down financial budgets into categories, and stores. The demand forecast informs plans for more accurate product assortment decisions, promotional strategies, and more.
Demand planning can be a confusing term that combines both terms, but in most cases when retailers refer to demand planning they mean forecasting demand.
The Importance of Forecasting Demand in Retail Big Data Analytics
Inventory — even above real estate, merchandising or advertising — is the biggest investment retailers make.
In most cases, retailers must purchase their inventory months before their products start selling. And inventory-driven costs extend well beyond payments to your suppliers. There also are costs associated with running distribution centers and transporting stock to the stores. Furthermore, once your inventory is in stores, it generates carrying costs, as well as costs from shrink and clearance pricing. And finally, inventory ties up cash flow, and takes up shelf space from other potential merchandise — carrying a high opportunity cost.
That’s why retail key performance indicators are dominated by inventory-based metrics like Gross Margin Return on Investment and In Stock Percentage.
Understanding how much of every SKU needs to be stocked in every store and distribution center is mission-critical for modern retailers.
Underestimating demand results in empty shelves, rainchecks, dissatisfied customers, and lost revenue and profit.
Overestimating demand generates costs of its own. Having too much inventory on-hand ties up capital that could have been used to buy better-selling products and lowers GMROI. And just as bad, excess inventory will require expensive clearance discounts that can eliminate profit margins.
Therefore, accurately estimating consumer demand is the essential first step to optimizing inventory investments.
Demand forecasting drives most of the business’ performance metrics by reducing inventory costs, increasing turnover, and improving the customer experience. Demand forecasting can also drive more evidence-based decisions throughout the organization. With better insights into true consumer demand, retailers can realign their assortment, merchandising, and promotion strategies to maximize revenue and profit.
Demand Forecasting Methodologies
Generations of business school graduates have learned how to use demand forecasting methods that fall into three general categories: qualitative forecasts, time-series forecasts, and causal modeling.
Qualitative forecasts are built up from market research and expert predictions. Public opinion surveys, for example, provide a sense of consumer confidence. More focused surveys can reveal consumers’ purchase intentions. Consultants, analysts, and other experts provide insights based on their work within the industry. A retailer’s own leadership provides insights based on their years of experience that inform qualitative forecasts.
Time-series forecasts rely on analyses of existing SKUs’ historical sales. These techniques identify trends, short-term seasonality, medium-term cycles, and growth rates. Analysts extend these historical factors to predict sales in the future. Unfortunately, time-series forecasting techniques assume that past performance is indicative of future results. This may be true when forecasting the next few weeks but becomes less true the further out you look.
Causal modeling uses simulations to create more accurate demand predictions. Analysts use regression techniques to create equations that correlate various inputs to sales outputs. Leading indicators like new housing starts, for example, will impact future appliance sales. Store-specific factors like the proximity of competitors have measurable impacts on sales. Internal factors, such as headcount trends or overall ad spend, also have quantifiable impacts.
Analysts combine all of these equations to produce a predictive demand forecast model. Running that model once may not produce the correct result since causal models are sensitive to initial conditions and their built-in assumptions. Much like meteorologists forecasting a hurricane’s path, retail analysts must run their model multiple times to create a consensus forecast.
Producing basic time-series forecasts may take a few hours while surveys and other qualitative techniques can take weeks to complete. Developing sophisticated causal models can be a year-long process.
Not to mention that not all products behave the same, and not all stores exhibit the same geo-demographics. As we’ll see in the next section, however, it doesn’t matter how fast these tools yield forecasts if they aren’t forecasting demand accurately.
Image Credit: SupplyChainDigest
Important Demand Forecasting Constraints to Consider in the Digital Era
Traditional forecasting methods struggle in the face of today’s dynamic retail market.
In the past, retailers could reserve SKU-level forecasting for the most important products and cover the rest of the assortment in category-level (or sub-category-level) forecasts.
That approach won’t work in today’s dynamic retailing environment.
Optimizing your inventory investment requires a much more nuanced understanding of each SKU’s demand. Demand varies based on the store’s location, the time of year, and the impact of other SKUs in the assortment. On top of that, even smaller retailers must address the multi-channel reality of today’s shopping environment. Increasingly tech-savvy customers comparison shop not just against competitors, but even against a retailer’s other sales channels.
But those factors are just the tip of the iceberg.
Beneath the surface, hundreds of other factors influence demand. Not accounting for each one will undermine your forecasts, leaving you with excess inventory and lost sales. Here are just a few of the subtle variables that shape demand:
Seasonality: Retailers have always lived with the way sales change during the year. For retailers with regional or national footprints, however, seasonality will vary by store, and each product may have their own unique seasonality.
Price elasticity: First-year business students learn that sales go up when prices go down. But promotions, assortments, market conditions, and other contextual factors have a strong influence on price elasticity. Not all products have the same reaction to a change in price.
Promotional uplifts: Forecasting promotional uplifts has as much to do with both consumer perceptions and sales channels as the offering. For example, the perceived value of rebates may not be the same in-store as it is online. Also the impact of promotional uplifts on demand is difficult to forecast in advance.
Supply chain/vendor lead times: When vendors lead times are measured in months, you often have to place your orders before knowing demand. In many cases vendor lead-times have a degree of variability that will affect the availability of your merchandise.
Different store types: Medium to large retailers often have multiple store types, for examples express/convenience, pop-up stores, mall locations, standalone showrooms, and outlet centers. Also retail is no longer just brick-and-mortar. Customers shop differently when they are on your website, using your app, or searching through a third-party market.
Geodemographics: The communities surrounding each of your stores have different tastes, shopping patterns, and interests. These can impact per-store-per-SKU sales quite dramatically even for stores within the same city. Retailers can’t simply offer the same assortment and inventory levels across all stores. Personalized assortment has quickly become the preferred method of allocation of inventory.
Competition: Each product’s competitive environment is just as unique. Today’s retailers need to keep an eye on merchandise they compete on in terms of pricing and promotions in-store and online.
Product demand cannibalization: Promotions, new product launches, and other activities that draw customers’ attention to one SKU will naturally draw their attention, and purchases, away from other SKUs. Product cannibalization is a challenging variable to account for in demand forecasting.
Assortment depth vs. diversity: Your assortment strategy influences demand. Variety reassures customers that you can meet their individual taste while depth gives customers confidence that they can buy what they want when they want. At the same time, you know that 20% of your products will yield 80% of profit, so retailers will not generally want to stock the same inventory level across products.
In addition to these common influences over demand, many others are specific to particular retail categories. Customers’ constantly-shifting color preferences will influence the demand for apparel. Apparel retailers also need to understand optimal size distributions for fashion products. On the other hand, technical specifications like screen size and processor will affect demand for smartphones.
Products without a predictable sales history add more layers of complexity to the demand forecasting process. Sales of games, movies, and other media are often too front-loaded for time-series forecasting tools. In the case of new product forecasts, you not only lack sales histories for the SKUs you bring in, but you also lack the cannibalization histories for adjacent SKUs. The same issue exists with products that sell sporadically, which don’t offer a consistent sales history to identify a pattern from.
Managing this complexity is the ultimate weakness of traditional demand forecasting.
Traditional demand forecasting tools were designed during a more analog time (where access to data was very limited) — and cannot provide the big data analytics that modern retailers require.
When using legacy demand forecasts, retail analysts must build hundreds of demand forecasting constraints into their models and adjust them for each SKU in each store. Even an army of analysts could not do this effectively for thousands of SKUs across hundreds or thousands of stores — no matter how savvy they are with Microsoft Excel of business intelligence tools for data mining and visualization.
Click here to learn how Retalon’s AI software automates demand forecasting for medium to large retailers, automatically accounting for the many variables that impact demand.
The Modern Demand Forecasting Methodology
Traditional forecasting methods fail to handle the complexity and data-intensive nature that today’s retailers require from demand forecasting. By adopting more modern approaches like Retalon’s advanced analytics system, all of the variables that influence demand at the product and store level are automatically accounted for.
Leveraging proprietary Artificial Intelligence and Predictive Analytics, Retalon helps retailers evaluate their historic data. Understanding the true historical demand reveals where retailers lost sales due to understocks and what drove their clearance markdowns on overstocks.
We’ve seen how traditional forecasting tools are over-reliant on past data and do not respond quickly to changing conditions. Retalon uses machine learning to automatically adjust forecasts and improve forecast accuracy for every SKU. Behind the scenes, Retalon blends multiple forecasting methods and algorithms to identify the best demand forecast for any product at any location for any given time.
Retalon’s modern approach to demand forecasting can handle the increasingly integrated, multi-channel nature of today’s retail business.
Customers may buy a product in the store, on the retailer’s website or mobile app, or through third-party e-commerce and social media sites. Those purchases may be fulfilled in-store, through direct shipments, or third-party fulfillment. Retalon Predictive Analytics incorporates signals from each combination of channel and fulfillment to produce accurate demand forecasts.
Traditionally, forecasting new products or product categories relied on guesswork informed by market research and qualitative inputs.
Retalon’s Retail AI systems break down each existing SKU into sets of attributes, each of which gets its own demand measurement. By mapping these attribute metrics to the attributes of the new product, Retalon gives retailers a forecast of true new product demand based on their customers’ actual purchasing behavior.
As retailers look at where their businesses are heading, they see that the old ways of doing things simply won’t work any longer. But even well-managed retailers have been slow to change, daunted by the complexity of forecasting true demand.
Retalon’s Artificial Intelligence and Predictive Analytics system is a much more intelligent and agile approach to demand forecasting than retailers’ traditional methods. This is why we have helped top-performing retailers get ahead of the pack by embracing today’s digital era of retailing.
Getting started with Retail AI is as easy as running a proof of concept pilot on your own data. Even this quick-take will generate insights that may improve future revenue and profits. Contact our team to see what Retalon’s advanced demand forecasting system can do for your business.Retalon’s demand forecast powers our advanced analytics solutions for retailers. See our full solution suite here.