Best Practices for Sales Forecasting

What are the best ways to forecast sales?

This is a question that has haunted online merchants as long as there have been eCommerce sites. To help, we’ve outlined best practices when it comes to forecasting sales for your online store.

8 Best Practices to Consider for Sales Forecasting

1. Use Sales Velocity

Sales velocity is a big factor when considering sales forecasting. But remember that average sales and sales velocity are two different things.

Average sales are the units sold / time period.

Sales velocity is sales / days in stock.

Average sales include both in-stock and out-of-stock days, but sales velocity only looks at in-stock days. This is an important distinction because if you do not exclude those days out-of-stock when there couldn’t be any sales. In that case, you then under-forecast the true demand, and that will lead to more stockouts. Then it becomes a vicious cycle of not having enough inventory to meet customer demand.

Some merchants handle replenishment by setting reorder points. For example, if a certain product gets down to five units, it is time to reorder. The problem with using reorder points is that it’s a static way about your store’s needs. Even if you revisit this reorder point, you are creating ongoing work to update it for your replenishment process.

Sales velocity is always a dynamic metric. If your store is growing quickly, those five units aren’t relevant anymore; you will run out because you are going through more than that quantity. Sales velocity will adjust with the your store growth.

2. Track Stockouts

This can be tricky to do without some sort of automated mechanism or software. If you manually track stockouts and find it to be tricky and time-consuming, prioritize what your highest value items are and keep track of your core items. Tracking your top 10% of performers may be more reasonable way to measure stockouts than by tracking across all SKUs. Once you get to a certain number of SKUs, the operations get so complex that it is difficult to do manually.

Tracking stockouts allows you to use sales velocity rather than average sales when figuring your forecast. If you are using average sales rather than sales velocity to forecast sales, you could run into repeat stockouts.

3. Consider Trends

When looking at trends, don’t just ask yourself, “What’s my sales velocity over this period of time?” What if a certain product is increasing in popularity? Is the demand for it dying off?

Look at what its most recent sales history has been. For example, if you sell a particular dress at the rate of 115 per month over the last year and you continue to order 115 units per month, you’re ignoring other factors. Is the hemline “in” this year? Is it the “it” color of the season? If it is, you may need to order 125 units next month and 135 the month after that.

If you are looking at trends and adjusting sales manually, put particular weight on the last two or three months. This will easily help you know what the most recent demand has been.

There is one caveat: if you are using short time periods to calculate metrics, the calculations will be very sensitive to any sorts of spikes for dips, having an outsized effect on key metrics. However, if you consider the entire year, that effect would be smoothed out with calculations over the longer time span.

The emphasis on recent months’ trends does hinge on the fact you are using a non-seasonal forecast. Read on to learn how to handle seasonal vs. non-seasonal forecasting.

4. Determine the Right Reference Period for Calculating Forecasts

Should you be looking at sales over the last six weeks or the last two years? The forecasting method you are using can make a large difference in determining how long your reference period should be.

Seasonal products are items that sell at different rates during different times of the year. This could include items only available during a certain period, or it could include items available all year. The important distinction is that sales are dependent on the calendar – due to factors such as seasons or holidays.

Seasonal forecasting looks at sales for the same time period during the prior year. For example, if you are looking ahead to which pairs of shorts will sell in May, February/March/April sales are not as relevant. You need to know what happened last May to better predict the forecast.

If you use a seasonal forecast, look back at least one year—or preferably two years—for a reference period. Was there year-over-year growth? You can factor that in to what is happening for the future. For example, if you only use 2018 data for seasonal products to see what will happen in 2019, you will only have that exact same data you have for 2018. The figures could be just the same, or you can guess and have a blanket increase.

However, if you look at two years of data, you can calculate the year over year growth rate. What was the growth rate from 2017 to 2018? If it was 25%, you can then apply that 25% growth rate to the 2019 forecast.

Forecasting is a part art, part science. You reduce what is unknown if you have two years of information. It gives more context to make a better forecast about what is happening in the future. These are data-based decisions rather than guessing.

5. Use Days of Stock, Not Safety Stock

Setting safety stock is not the best way to build in a buffer for your inventory needs because safety stock is a static number. It does not grow as your store grows. Instead, you will want to build in a buffer to your days of stock.

Days of stock are the number of days that your current inventory will cover considering customer demand for those products. For example, if your sales velocity is 2 units per day and you have 50 units on hand, the days of stock is 25 days.

How much inventory will allow you to meet customer demand without being overstocked and having too much cash tied up in inventory sitting on your warehouse shelves? A typical place to start is 30 days of stock. For a product that is selling well, this should be enough time for you to recuperate your investment and spend your profit on more of that product or another area of growing your business.

A smart sales forecast is calculated using the sales velocity and the sales trends in recent months. Once you have your forecasted sales and determined your desired days of stock to have on hand, then you can think about how to build in a buffer to account for supply chain problems. A dynamic way to handle this buffer is to add to your days of stock.

If you’re using safety stock as a buffer, then you’re creating extra work for yourself maintaining a solution that doesn’t match customer demand. Ask yourself how you’re determining your safety stock level. Is it based on customer demand? If so, when was that demand? Are you changing and updating your safety stock level every time you see a change in customer demand? How often does customer demand change—weekly or monthly?

Using days of stock, your inventory buffer is updated dynamically to match changes in customer demand. If you’re selling 2 units per day in January then later selling 5 units per day in March, the number of units to cover 30 days of stock goes from 60 units to 150 units. When you add 5 days of stock as a buffer, then you’ll stock another 25 units. If you had set the safety stock in January at 10 units, then by March that is only a 2-day buffer. It won’t grow as customer demand grows. Using additional days of stock will steadily increase the units needed to cover another 5 days.

6. Set Lead Time and Days of Stock

Lead time is the amount of time it takes to order a product from the vendor and receive it into inventory. It includes manufacturing time, transit time, and shipping time. As previously mentioned, days of stock are the number of days that your current inventory will cover considering customer demand for those products.

Another way to think about how long to set for your days of stock is to consider what would be your ideal stock cover. How long will this stock last? How long should a new order last once it arrives? One other way to think about Days of Stock is to consider it your purchasing frequency. If your stock lasts 30 days, that means you place an order every 30 days.

However, it’s not just enough to know that. For example, considering sales velocity and trends, you forecast you need 100 units over the next 30 days. You must take into consideration continued selling during the lead time. If it takes 14 days to get the product after the purchase order is placed, what is the stock going to be on day 14 when the order actually arrives? If you don’t think about the continued selling during the lead time, you start when that order arrives and you are actually going to have 16 days of stock rather than 30 days. Plan ahead and forecast what the stock is going to be so you don’t start off on the wrong foot. Use the sales velocity and current stock level to estimate future stock levels.

7. Plan for Promotions

Marketing departments plan promotions months in advance. You might have a “Welcome to Summer” sale, generating an estimated 30% sales increase in June. It is important so that at the appropriate time, 14 days (your lead time) before June 1, you don’t just order what you were originally projected to sell in June, but also that extra 30% so that you can cover the promotion. Plan on extra amounts that are going to be needed, whether that is at a product level or category level. Coordinate across all parts of your company so that you can plan ahead and meet customer demand.

8. Calculate ABC Inventory Classes

Consider using an ABC class analysis of recent sales. That looks at each variant or product’s contribution to your revenue over the last 30 days. Alternatively, you could calculate the contribution to your profit instead of your revenue to maximize your return on investment.

To calculate the contribution of each variant to your total revenue, divide the variant revenue by the total revenue.

Variant contribution = variant revenue / total revenue

Then sort variant contributions from highest to lowest.

Finally, create a running total or cumulative sum of the variant contribution.

The variants contributing to 80% of your revenue are considered A class items. The next 15% of variants are B class, and the final 5% are C class items.

Highest priority is keeping A class items in stock when determining replenishment priority because that is what is delivering on the revenue. Put a lower priority on B and C class items if you can tolerate a stockout on those items. For eCommerce store owners, cash flow is always top of mind, so prioritize A class items first. Then determine if C class items are worth the investment. Alternatively, you could potentially eliminate C class items since they contribute the least amount to your overall revenue.

There is a myriad of things to consider when sales forecasting; everything from lead time to ABC analysis comes into play. Using dynamic forecasting and having reliable data are the two building blocks that all forecasting is based on. When you use the data correctly and are nimble enough to make changes, you have set yourself up for success to make the very best decisions possible for your online store.

‍