The simple guide to SKU-level demand forecastingBy Jon Taylor on February 21, 2022 - 10 Minute Read
Inflation is at the top of every business owner's mind right now, and it's driving up costs in one crucial part of the supply chain: warehouse storage.
According to a report by Warehouse Quote, average warehouse costs are up 12% on baseline costs, but consumer spending is down. A lot of companies stacked their storage facilities with products, but as inflation hit and the economy started to shrink, trouble hit.
That’s the thing about stock levels and cash flow. Unless you have a system that simultaneously acts on historical data, customer demand and industry trends, you can end up making decisions that will cost your business a huge amount of money in overstocking and storage costs.
And that’s exactly why a strategy like SKU-level demand forecasting can be worth its weight in gold. Using a mix of forecasting methods, data and metrics, SKU-level demand forecasting can help businesses effectively manage their supply chain and make great, data-driven decisions about pricing and inventory numbers.
This guide is going to walk you through why SKU-level forecasting is a game changer for companies, and provide some handy hints on how best to use it. We’ll cover…
- The nuts and bolts of SKU forecasting
- SKU-level analysis: what is it?
- The four main benefits of SKU forecasting
- How to (accurately) forecast inventory levels with SKUs
- Best practices for SKU-level demand forecasting
Let’s get started 👇
The nuts and bolts of SKU forecasting
SKU forecasting predicts the demand for specific products in a company’s inventory.
The process analyzes data, such as past sales and consumer trends, to help businesses predict future product demand and keep optimum amounts of stock on hand without overpaying for storage space or tying up cash in excess stock.
Essentially, this information allows companies to meet consumer demand without ever running out of stock but still optimizing storage costs.
There are two main types of SKU forecasting methods: qualitative and quantitative.
👨💻 Qualitative forecasting
Based on expert analysis and opinions, qualitative forecasting is done by humans who use judgment to make predictions if data is scarce. Market research, customer feedback, BETA launches and product surveys can all help experts make a more accurate forecast.
📊 Quantitative forecasting
This method is a bit more cut and dry. Quantitative forecasting relies on statistical analysis and data modeling, so the results you get are more data-driven and less likely to have human bias or errors. Methods like machine learning, artificial intelligence (AI) and series analysis fall into this category of SKU forecasting.
More often than not, the most effective way to approach SKU forecasting is to mix qualitative and quantitative approaches. Yes, humans have a certain degree of error when they are put up against a machine, but modeling can also miss personal aspects of SKU forecasting like new product launches or one-off events.
SKU-level analysis: what is it?
SKU-level analysis involves examining individual products or SKUs in your inventory to determine their performance and potential demand.
It looks at data like stock levels, seasonality, lead times and multiple other factors that can impact demand and product sales. By conducting SKU-level analysis, you can identify which products are selling well, and which ones aren’t performing as well as you’d like. At a basic level, this type of analysis can give businesses the data they need to discontinue product lines and create new ones based on customer behavior and buying trends.
More often than not, the most effective way to approach SKU forecasting is to mix qualitative and quantitative approaches.
The four main benefits of SKU forecasting
For most businesses, especially those operating online or with large amounts of stock in warehouses, SKU forecasting is a key way to optimize inventory levels and cut unnecessary costs.
Here are four (important) reasons you should consider using SKU forecasting 👇
Optimized inventory management
Inventory management and optimizing stock levels remain a balancing act for businesses selling physical products. But warehouse prices have skyrocketed due to skill shortages and building materials in the storage sector.
SKU forecasting can help companies manage inventory levels so that you always have the right amount of stock on hand to meet customer demand, but never have excess stock that you are paying to store.
Improved supply chain communication
SKU forecasting gives you real-time data on your inventory levels that can then be shared with suppliers to ensure products running low are replenished before they run out. Giving suppliers the heads up as soon as a product line needs topping up helps them plan and schedule production ahead of time. This reduces the chances of your business losing sales and keeps customers happy.
More insight into market trends
SKU forecasting can give valuable insights into customer behavior and market trends so your business can make optimized decisions about pricing and product demand. It’s (kind of) like having a crystal ball into what your customers want — based on sales, wider industry trends and competitor products — so your business can produce items that meet that demand.
Better cash flow
Finally, SKU forecasting can help save money by optimizing warehouse space and stop you from overstocking products you don’t need. This money can then be reinvested into your business or used for other strategies like marketing campaigns or product development to spur growth.
So, how do you actually create and use SKU forecasts? 🤔
How to (accurately) forecast inventory levels with SKUs
An accurate SKU forecast depends on three factors: your data, your method and your metrics.
Let’s take a closer look at each of them 👇
1. Consider all the factors in your industry
Sales data, seasonality, lead times.
Every industry is different. Clothing retailers cannot use SKU forecasting in the same way as a manufacturer. The first step of any successful SKU forecasting strategy is to collect and analyze data from your own sales and start looking for patterns.
2. Choose a forecasting method
Once you have some data, there are a few different methods you can use to forecast SKU demand. Each method has its own benefits and it’s important to choose one that suits your industry, inventory goals and tech setup.
Method 1: time series forecasting
Time series forecasting is the simplest method on our list and uses historical sales data to predict future sales. It assumes that past sales patterns and levels will continue in the future, and it also takes into account peaks and troughs like seasonality.
Why it’s a good choice
Time series forecasting is ideal for companies just getting started with SKU forecasts. It can analyze simple patterns, like seasonality and holiday demand, to help businesses predict how much product they should stock and sell. For example, a company that sells winter jackets would want to use time series forecasting to predict demand during different seasons to help optimize stock levels.
The cons of time series forecasting
This method is only accurate if you have a large amount of historical sales data. Its simplicity can also be a downside, as it can’t do more complex analysis like competitor or consumer behavior.
How to use it
- Collect historical sales data for each product SKU
- Invest in a tool (like using a software program to analyze the data and generate a forecast)
- The tool will then use an algorithm, like Autoregressive Integrated Moving Average (ARIMA) model, to comb through your data to predict future demand
Method 2: machine learning forecasting
Machine learning forecasting is a more advanced method of SKU-level demand forecasting that uses algorithms to analyze data and spot patterns humans might miss. This method can be more accurate than time series forecasting because it can identify complex factors, like consumer trends, that impact product demand.
Why it’s a good choice
As machine learning forecasts analyze linear and non-linear relationships, it can handle more complex data than just consumer buying trends and seasonality. For example, a company that sells video games might find that demand for a particular game is influenced not just by the time of year, but also by the release date of a competing game. Machine learning can use this information to create a more accurate forecast about product and consumer demand for the video game company.
How to use it
- Collect large amounts of data on each SKU, including sales data, demographic data, market data and consumer expectations
- You then feed this data into a machine learning algorithm (usually using a SKU-level demand forecasting software), which will analyze the data and generate a forecast
Method 3: regression analysis
Regression analysis pits dependent variables (like sales and inventory) against other changing values such as price, consumer spending and supplier cost to allow for more accurate forecasts.
Why it’s a good choice
Regression analysis is one of the most in-depth forms of SKU forecasting because it can handle multiple independent variables. If your business usually juggles lots of factors, like different suppliers, seasonal changes or supply chain fluctuations, this method may be a smart choice.
How to use it
- Collect data — like sales, manufacturing costs and marketing costs — for each SKU in your inventory
- Then, use a platform like Peak to run a regression analysis and generate a forecast based on the relationship between each variable
Next, you need to decide how to measure SKU forecasts.
Choose your metrics
To really get the most out of your SKU data, you must track metrics and set KPIs to see how accurate the forecasts are. Here are three key metrics to measure when SKU-level demand forecasting:
Metric 1: stock turnover ratio
Stock turnover ratio measures how quick you sell (and replace) products in your inventory.
It divides the cost of goods sold (COGS) by the average inventory you keep to see how effectively you are managing inventory. Let’s say a clothing retailer has $200,000 in starting inventory of t-shirts, and an average inventory of $10,000 of stock on hand. The stock turnover ratio would be calculated as follows:
📊 Stock turnover ratio = COGS ÷ average inventory = $200,000 ÷ $10,000 = 20
The t-shirt seller has a stock turnover ratio of 20, which means the average level of inventory is sold through about 20 times a year, or nearly once a fortnight.
A good rule of thumb is high stock turnover means inventory is well managed and you pay minimum amounts to store products. But a low stock turnover could signal wasted money on storage or overpriced stock that is hard to shift.
Metric 2: gross margin return on investment (GMROI)
Gross margin return on investment (GMROI) measures how much money a company earns from inventory.
This metric is calculated by dividing the gross profit by the average inventory cost, and it’s particularly useful for finding SKUs with high-profit margins. For example, a CPG business sells a canned soda that has $50,000 in gross profit and an average inventory cost of $10,000. The GMROI of this SKU can be calculated easily:
💰 GMROI = Gross profit ÷ average inventory cost = $50,000 ÷ $10,000 = $5
The GMROI of five shows the CPG manufacturer generates $5 for every dollar it invests in the soda.
By tracking the GMROI for each SKU, the business can identify which products are generating the highest returns, understand demand surges and adjust production as needed.
Metric 3: weeks of supply
Finally, it’s always a smart idea to track how many weeks of supply of inventory your company has on the shelves.
A high level of stock on hand could start to ring alarm bells that you are wasting money on storing products you don’t really need. On the other hand, if you have low weeks of supply, you risk being understocked and not being able to meet a surge in customer demand.
Let’s say a retailer has 145 units of a particular SKU in inventory and sells an average of 10 units per week. Here’s a simple calculation to figure it out:
👨💻 Weeks of supply = current inventory ÷ average weekly sales = 145 ÷ 10 = 14.5 weeks
The retailer has just over 14 weeks of supply left for this SKU in its product inventory. The reorder level for this particular SKU will decide when another order from the supplier needs to be placed, but the retailer should be asking “do we really need 14 weeks worth of stock for a single product?”
Tracking key metrics like these can help you understand how accurate any SKU forecasts are and if you are using the right method for your particular industry and needs.
Best practices for SKU-level demand forecasting
1. Manage your stock wisely
Don’t run your shelves dry — but don’t have them overflowing either. Alongside a solid SKU forecast strategy, you should always have safety stock and set reorder points to keep each SKU at an optimized product level. This means you should:
Think about safety stock
Safety stock is another term for extra inventory that’s kept in case you experience a sudden surge in demand or hit problems with your usual supply chains. It’s important to keep different levels of safety stock for each SKU, based on factors such as stock turnover ratios, lead times from suppliers, historical customer demand and storage costs.
Have a reorder point for each SKU
Calculate a reorder point for each SKU in your inventory. This should take into account the cost of each SKU along with safety stock levels, seasonality, supplier lead times and GMROI. Ideally, the reorder point will optimize inventory levels and allow you to meet surge demand without having to hold a ton of stock and pay for storage costs.
Optimize reorders with an economic order quantity (EOQ)
The EOQ for each SKU you stock is the optimal order size that minimizes the total cost of inventory. It’s calculated by looking at costs like ordering, carrying and storing stock. By calculating the EOQ for each SKU, you can order the right amount of inventory at the right time based on forecasted demand and minimize holding excess stock.
Even if you have a system in place, it needs to be monitored and tweaked when your company grows or customer expectations change. This brings us to the next best practice.
2. Monitor everything (and change things up when needed)
Regularly monitor sales and inventory levels for each SKU, and compare them to the reports generated by your chosen SKU-level demand forecasts.
It’s vital to tweak your forecast and reorder points based on any unexpected changes in demand or inventory levels to keep them accurate.
To do this, compare actual sales, surge demand and inventory levels to see how accurate the forecasts are and if you need to tweak any inputs or outcomes. Make sure you keep an eye on every SKU in your product line to ensure you can always meet customer demand without blowing a hole in your bottom line.
3. Use forecasting software
Finally, successful SKU forecasting is only possible with the right tools.
It’s impossible for humans to keep an eye on stock levels, forecast demand changes and suggest optimal reorder quantities and stock numbers. But forecasting software can do just that — and do it in real-time, too.
A platform like Peak can free up cash flow that’s tied up in excess stock and help you maximize availability to meet consumer demand. Thanks to AI, our SKU-level demand forecasting app, Reorder, collects transaction, product, pricing and warehouse data to enable businesses to benefit from cutting-edge AI models and accurate demand forecasts.
Accurate minimum order quantities, seasonality forecasts and warehouse capacity are updated each time a customer buys a product or an SKU is running low. Instead of spending hours staring at spreadsheets, your team can get back to doing what they do best: growing your business 💰
Discover the future of forecasting. Say hello to AI.
Download our demand and supply whitepaper to learn how you can connect the dots across your organization with AI. Optimize inventories, fulfill demand and set dynamic safety stock levels, all the time.