Demand forecasting for new products is a challenge ripe with many pitfalls.
Without historical data to draw upon, how do you predict the best way to allocate new products between stores, set the most profitable pricing, plan for future promotional demand, and balance inventory between DCs, warehouses, and B&Ms (in-store vs. online)?
Guess wrong and you will either (a) run out of inventory, (b) stock too much inventory that will need to be marked down, or (c) stock the right inventory in the wrong place (that will need to be transferred, or worse, marked down). These are all very costly mistakes.
Yet, of the traditional forecasting methods used by retailers, none are very accurate for predicting the performance of new products.
For example, retailers can try to use a standard sales forecasting approach by looking at the past sales of the most similar products. But this is likely not going to be accurate. If you had products similar enough in price, quality, style, function, and brand — you probably wouldn’t be bringing in the new product, would you? New products are brought in because they are different (at least in consumer perception).
Furthermore, traditional methods won’t tell you anything about how a new product will fit into your store footprint, existing product mix, customer demographics, etc. You’ll need teams of analysts and planners crunching numbers and making guesses.
So is it all a guessing game? Should bringing new products into your assortment always be a gamble?
There’s a way, and it’s easier than you may think.
But before we explain how to forecast demand for new products — let’s look at why it’s so risky, and what makes it so difficult.
Why Demand Forecasting for New Products is So Risky
Introducing replacement SKUs or line extensions for existing products shouldn’t be too difficult. Your current sales or demand forecasting process may be enough.
In these situations, your new SKUs will be very similar to existing ones (in price, quality, brand, etc.), and their performance can be forecasted without major changes to forecast accuracy. While there’s some risk, it’s comparable to the risk of forecasting demand for your regular products.
But this only applies to products that that are very similar, or closely related to products you have already been selling.
Introducing a brand-new product that your stores have never sold before is a much riskier proposition. Launching a new product requires making significant commitments to your vendors. Your vendors often lock your initial orders and early replenishment orders to coincide with their own forecast windows. Any co-marketing funds you receive require investments in-store merchandising and promotional campaigns. And in some cases, the future of your vendor relationship will depend on the sales performance of the new product.
And again, without an accurate demand forecast, you risk misallocating inventory and missing your revenue targets while watching turns collapse. Fixing the problem through stock rebalancing or price markdowns will be expensive. Not to mention the lost opportunity of bringing another product instead and losing cash.
That’s why getting off on the right foot is critical.
The larger your business, and the more frequently you introduce new products, the less you can rely on best guesses. How you estimate demand for a new product matters. However, systematic approaches get stymied by the number of variables that influence demand.
Main Difficulties of Forecasting New Product Demand
Large retailers can’t employ enough analysts to understand everything driving product demand. Forecasting demand for existing products is difficult enough to do manually because to get an accurate prediction you need to account for:
- Personalized assortment
- Space constraints
Sure, smaller retailers might be able to make some assumptions and educated guesses for a couple of locations.
But how do you accurately forecast the demand for each store, when you have hundreds of stores and multiple channels across dozens of regions — all with their own demographics, space constraints, local competition, etc.?
And that’s just to predict demand for existing products.
When it comes to forecasting new product demand, there are even more layers of complexity.
For example, when trying to forecast the performance of new products, retailers need to account for the cannibalization effect (both on the new products and by the new products). And this is no easy task. Cannibalization is a very subtle interaction of relative pricing, inventory levels, changes in demand, shelf positioning, and other factors.
And that’s just to predict demand for new evergreen products.
But as retailers know, there are few products that are truly evergreen. Forecasting new seasonal or limited-run products is even harder. For example, in video game retail, new product sales are heavily front-loaded with huge spikes on launch day, with demand quickly dwindling in the weeks that follow. In this type of industry, it’s vital to get the launch inventory just right — because you won’t have another chance to fix it. If you stock too much at launch, you’ll be stuck with inventory for months. If you don’t stock enough, you likely won’t get another chance to recover your lost sales.
Furthermore, introducing new products during promotional campaigns will skew your forecasts. Uplifts are difficult to calculate when campaigns center around the new product. Even more challenging, measuring the impact on new product sales when campaigns center around adjacent categories.
And if that wasn’t difficult enough, your product pricing strategy will have a massive impact on demand (making or breaking your performance). That’s why before you can start answering questions about inventory, you need to make sure your pricing strategy for the new product is well-defined.
What should initial pricing look like? Is launch pricing different from everyday pricing? Is the pricing seasonal or does the vendor have a predictable cadence of price drops? And how do you set promotional pricing to optimize sales, inventory, and profitability?
Are Traditional Methods of Estimating New Product Demand Accurate?
As complicated as forecasting may be, the first forecast has to come from somewhere.
Buyers and analysts try to do their best by finding data that hints at the new product’s future performance.
Again, in some cases, you may have existing products that are “close enough” to draw comparisons. Different-styled apparel meant for the same customer; laptops with different specs but similar prices; or last year’s products from the same holiday season can all inform new product forecasts.
Outside market research from retail specialists can give some insight into new product categories. Consultants can help you understand new kinds of customers, assess a product’s mindshare on social media, or conduct local focus groups.
A pilot project may provide actionable information when a new product’s demand is too uncertain. If you don’t have enough stores to create a representative sample, a limited-run pilot in all stores may be your only choice.
Product pre-orders are another way to gauge demand for products that have never been sold at your stores. But this requires a solid infrastructure for promoting and taking pre-orders, and some data on the relationship between previous pre-orders and in-season sales.
Some retailers use pricing to discover optimal demand. They introduce the new product in low quantities and a higher price point, reducing it over time until it reaches the most profitable run rate. However, this approach runs the risk of cutting too far and is heavily reliant on trial and error. Once you’ve set a lower price in customers’ minds, reversing course is difficult.
These and other traditional approaches to forecasting products suffer from the same failings.
None of them are accurate predictors of demand. None of them predict your business’ unique situation. And to an extent, all of them simply inform gut decision-making.
Reliable data only arrives after the fact which can make new product launches very expensive. This is why new product sales forecasts have traditionally been more of an art than a science.
How to Improve Accuracy While Forecasting New Products
Retail AI platforms like Retalon take demand forecasting for new products to new levels of accuracy. Machine learning algorithms can analyze many more inputs and tease out trends better than any analyst identifying the factors that impact demand for the new product. By using Predictive Analytics, you can produce more accurate by-SKU-by-store demand forecasts even when you have no sales history.
Predictive Analytics automatically generates a forecast based on a new product’s attributes rather than on the product as a whole. For example, how do red shirts short-sleeve shirts and cotton shirts perform? The completed analysis yields intelligently allocated inventories for each store and distribution center. At the same time, Retalon’s AI evaluates the cannibalization of other products and reduces their upcoming orders.
The system automatically accounts for seasonality, assortment depth vs diversity, store times, e-commerce fulfillment demand, promotional uplifts, price elasticity, related products and much more.
Once the new product launches, Retail AI learns more about the product’s true store-level demand faster than traditional analysis. The system automatically responds to the reality on the ground and adjusts future orders, allocation, promotions, and pricing accordingly.
Retalon’s Predictive Analytics platform enables a unified, analytical approach to retailing that integrates pricing with inventory management. The AI not only identifies future demand but also optimizes pricing throughout the product life cycle to align with store-by-store inventory levels to maximize revenue and profit.
The best way to see how your specific business can benefit from this technology is to request a demo on a sample of your own data. Get in touch with our team to see how our Predictive Analytics can reduce the risk and maximize the return of your next new product launch.