Through our experience with D2C brands globally we have been exposed to a variety of data sets and problems brands face in inventory forecasting. This is an age-old problem that is currently being solved using multiple spreadsheets and a vast array of assumptions - both of which do not scale as a business grows.
💡 Inventory Forecasting, or demand planning, is the practice of using past data, trends and known upcoming events to predict needed inventory levels for a future period.
When it comes to inventory forecasting there are two problems a business is trying to solve: running out of stock on key products that leads to a loss in revenue; and over-stocking on products that leads to stale capital.
Below are some of the most common problems we see brands encounter:
1. Incorrect data and/or human error
brands are currently exporting data from multiple tools such as Shopify, Amazon FBA and Google ads and are manually editing this data in order to conduct inventory forecasting. Besides the time it takes brands to compile this data, formula errors are common and decisions are sometimes based on the incorrect data. There are also no processes in place to determine if the data is accurate. Instead the data is shared with the team in the hopes that someone will realize if a value looks incorrect which is not sustainable. The effort required and the possibility of incorrect data leaves too much room for error.
2. Not updating past product data
As brands scale, they release new product categories and update product data. This provides a problem for past sales data, as the new changes must be reflected in continued products. For example, if product X is under the sneakers category we assume, based on past data, products in the sneakers category have a certain demand curve. However, if the product X category is changed to footwear then we would have no historical context of sales for this product which means we would be unable to predict a future sales forecast even though it’s the same product.
3. Don't Aggregate data
Most brands use excel for forecasting by running simple averages and trend lines over 30, 60, and 90-day intervals to determine future demand. Depending on the scale of the business this could be sufficient, but as a business grows it can lead to many costly mistakes.
Aggregate data hides subtle signals such as demand shifts between weekdays and weekends, holidays, discounts and anomalies that can be attributed to specific events including marketing campaigns or large wholesale orders. It is best to model future demand based on granular transactional data
4. Incorrect data and/or human error
A common trend we see across brands is that they don’t ignore sales anomalies, such as an unusually large wholesale order. Taking into account anomalies can inflate sales for a period of time which leads to brands over-estimating demand. When a business receives a large wholesale order, it should not include this in its demand planning but rather keep track of it separately. Usually, anomalies are relatively predictable and can be tracked.
Ad campaigns or email newsletters are a bit harder to predict demand but should be included in demand planning. These types of campaigns usually do have a correlation to an increase in sales that occurs over the length of the campaign (longer than the once-off increase experienced with wholesale orders). Historical data becomes extremely important in order to determine the effects these campaigns might have on future sales.
5. Take into account demand drivers
Inventory forecasting is completely dependent on the levers that drive demand. Levers such as repeat customers, marketing campaigns and discounts therefore, it's important for brands to be able to understand the effects of these levers. Currently, we see brands extrapolating past sales data without taking these effects into account which frequently lead to running out of stock and also overstocking. In order to achieve accurate inventory forecasting brands require a centralized up-to-date view of their data.
By implementing the above practices a business can optimize its inventory with small changes to existing processes. The earlier these practices are implemented the less a business will have to worry about inventory forecasting as the business scales as it will be working from an accurate and scalable foundation. Forecasts are only as good as the data brands are working with and so it's crucial that brands start with making sure their data is well prepared.
