Understanding demand forecasting for eCommerce enables you to predict buyer behavior and keep inventory updated.
One of the most important parts of running an eCommerce business is managing your inventory to meet the demand of your customer base. With too much inventory, you risk increasing warehousing costs and restricting cash flows. With too little you will stockout and won't be able to meet your sales. In order to perfect your inventory levels to meet your actual demand an in depth analysis of your business is required. This analysis is called demand forecasting.
We’re diving deep into how to forecast demand, the different types of demand forecasting, and the benefits and challenges of this process.
Demand forecasting may sound like a simple process of estimating the customer demand for your products and inventory required to meet those demands. But realistically, it is a much more involved process of analyzing historical sales data and market trends to make data-driven projections of actual demand for your eCommerce business.
Without demand forecasting, your eCommerce business can make ill-informed decisions about the products, sales and target markets – all jeopardizing warehousing costs, supply chain kinks, product stockouts, customer satisfaction and overall profitability.
Macro-level planning explains forecasting in the most comprehensive terms. Macro-level forecasting examines the entire market or the macroeconomic structure in which an eCommerce business operates. This method of demand forecasting requires substantial market research for the purposes of meeting strategic goals and objectives.
Dissimilarly, micro-level demand forecasting focuses on the business operations for sales forecasting and planning. Micro-level planning can include, past sales performance of products categories–separated by SKU, profit margins or cost of production. This level of demand planning allows businesses to look at their end and make internal projections on cash flows based on production.
Demand forecasting is focused on the future of the eCommerce business. With short-term forecasting though, businesses are generally thinking about the next 3-12 months. Short term forecasting typically gives insights into the seasonality of a business and how demands for products and services changes through the course of a year. The easiest way to think of short-term demand forecasting is to consider Black Friday shopping and the companies that have huge spikes in business around that time of year. Using short-term demand forecasting can help your business identify and prepare for these spikes.
Long-term demand forecasting looks further into the future. Generally, with long-term planning, you are looking at the next 24-48 months and identifying trends based on previous years' data. With far-reaching planning, you are looking at product trends and the larger business strategy. This might help identify new sales channels to adopt and changes in the supply chain that are needed.
Quantitative forecasting demand forecasting uses data analytics to create mathematical models to gauge future sales and inventory levels. To yield the best quantitative results, there is a need for robust data sets which includes high and low sales months, and seasonal peaks and trends over the last year.
A strong quantitative analysis can be broken down by product line or SKU and would include the following:
Qualitative demand forecasting is an economic model that uses economic and market forces to make broader predictions of demand in your industry. Qualitative forecasting is much easier for new eCommerce businesses as it uses external data sets rather than internal data.
To get started with qualitative forecasting, consider these factors:
Qualitative forecasting can be a major benefit to your business. With inventory management, for example, if you have identified products with shorter life cycles, you will want to focus on increasing the sales of that product before it becomes dead stock. Or, if you find that product prices will likely rise in the next few months, you can adjust your next order to increase your inventory levels– that is if warehousing costs don’t outweigh the price change.
Demand forecasting can generate significant growth in your eCommerce business. By using accurate data and forecasting methods you can easily navigate the changing landscape of eCommerce.
The biggest impact demand forecasting can have on your business is inventory management and supply chain management. With accurate sales projections and forecasts, you will know the future demand for your products, how much inventory to buy and the best time to buy it. This will help your business move smoothly, avoiding stockouts, lowering inventory holding costs and increasing inventory turnover rates.
Additionally, demand forecasting can help you integrate new products into your portfolio. With qualitative forecasting, you can identify which products will perform well and the best time to introduce the products.
Proper forecasting can also help balance the risks of your new eCommerce business. By forecasting demand you will have the knowledge to plan for new competitors, slow-starting products, seasonal impacts and any change in external economic factors.
While demand forecasting can offer huge benefits to your eCommerce business, it comes with its fair share of challenges and setbacks.
One common challenge faced is applying the wrong method of forecasting. If you use the wrong method you could be making meaningless projections that may have negative impacts on your stock levels and sales.
Another challenge you may face is using incomplete data sets. Specifically for quantitative forecasting, you could be using the right demand forecasting method, but with incomplete data, your predictions will be meaningless to your eCommerce business. For example, if you only have sales data for the last month you cannot accurately forecast for the next few months or for the next year. Fortunately, when starting your eCommerce business in the time of big data, incomplete data is less of an issue. With strong research, you can find the data you need to get a feel for the future of your business.
There are many demand forecasting methods and will provide you with valuable metrics to improve your business. Here is a list to help you understand how to better improve your forecasting and planning.
Time-series analysis is a great quantitative method of forecasting. Using historical sales data, you can make projections for future sales based on the growth of sales over time. Generally, time-series analysis is graphed with time on the X-axis and units sold on the Y-axis. This can be a great method for sales and inventory planning but it requires data from many previous years.
The next method of forecasting is trend projection. Where you can assume that any sales trends from the last year will repeat this year. With trend projection, you are assuming that there will be the same level of demand all else held constant. Trend projection is a classic method but may not be an accurate forecasting method as it assumes that trends will continue to the next year.
The barometric forecasting method uses economic indicators to make future projections. The first indicator is the leading indicator, which is used to predict future events. Next is the lagging indicator, which analyzes past events and performance. And lastly, the coincidental indicator, which measures real-time events.
The barometric method is a great way to make predictions when no past data is available, but with all forecasting methods, there are some drawbacks to the barometric method. One drawback is indicators might not be accurate for forecasting specific business needs. Another is that indicators can show the change in the economy but not the magnitude of change. Lastly, this method can only be used for short-term forecasting.
If you have access to mass amounts of data and the time to analyze it all, causal forecasting may be the best method of forecasting. Causal forecasting can make very accurate predictions using data from sales, competitors, marketing and advertising and economic activity. Combining this data you can make accurate forecasting models for the next year.
With a focus on insider knowledge, the collective opinion method leverages the knowledge and experience of your sales team to make predictions on customer demand. Members of the sales team compile reports of individual products and regions which can provide a broad view of demand from the data. You can also use this data to compare it to previous sales to see how certain products perform during high and low sales months, seasonal peaks and overall trends.
It’s important to remember that good forecasting makes use of hard data and the judgment of experienced managers and sales teams. Using the collective opinion approach in combination with another method could yield great predictions to benefit your business.
Customer surveys can provide key information specific to your target market. This method uses individual customer demographics and economic data to understand the needs of your customers. There are many ways to conduct customer surveys which can yield different results for forecasting.
Similar to customer surveys, the expert opinion method makes use of experts in the field to determine future market activity. Using the Delphi technique, you can make predictions for your business based on the consensus of subject matter experts.
Market experiments are another method of survey forecasting where online retailers can understand customer behavior under controlled conditions. You can use A/B testing to determine how promotions and site features and layout appeal to customers. While this can help understand customer behavior on your eCommerce platform it may not translate to all customers in all areas.
Statistics are powerful for improving your eCommerce business. Using a statistical method of forecasting gives you detailed and reliable information to use in your projections.
There are many statistical methods including:
Using any statistical method can help your business make meaningful, accurate and data-driven projections.
Another method that yields great results if you have access to mass amounts of data is exponential smoothing. This method uses past trends and performance to create a weighted average, where recent trends carry more weight than older ones.
The weighted average can be expressed as
New Forecast + (ɑ)*(last period actual sales)+((1-ɑ)*corresponding forecast))
Where ɑ is the weight allocated based on time. ɑ can be calculated from ɑ=2(N+1) where N is the number of periods from the forecast.
With this method, old data is reduced in its impact without negating its importance in the forecast.
Correlation analysis is a method to quantify the relationship between sales and external events. While external events may seem like a broad category it can easily be narrowed down to any event that may have an impact on your business, such as government policy or economic downturns. In order for this method to help you make good forecasts, there needs to be a good and credible relationship between the trend and the event that triggered the trend.
If you lack substantial past data, linear regression can help you out. This method models the relationship between a dependent variable and independent variables. For eCommerce businesses the dependent variable is usually sales, while independent variables can be conversions, email sign-ups or social media follows. The linear regression model allows you to measure the impact of the independent variable on the dependent. Simply put, with linear regression you can measure the impact of social media ad spending on sales compared to the impact an email blast has to better manage your marketing campaigns.
Apteo isn’t your normal eCommerce analytics platform. We enable organizations to understand which users are likely to re-purchase and which ones are likely to churn. This enables companies to focus their marketing and advertising efforts on the best customers. To see Apteo in action, schedule a demo.
Shanif Dhanani is the co-founder & CEO of Apteo. Prior to Apteo, Shanif was a data scientist and software engineer at Twitter, and prior to that he was the lead engineer and head of analytics at TapCommerce, a NYC-based ad tech startup acquired by Twitter. He has a passion for all things data and analytics, loves adventure traveling, and generally loves living in New York City.