Understanding Crisp's forecasts

For many folks, generating demand forecasts is done entirely in Excel. The transition from Excel to Crisp can feel like a huge leap! One of the most frequently asked questions we hear is, "what does my forecast represent?" If you are wondering the same thing, then this tutorial is for you! 

In this tutorial, we'll describe how Crisp generates forecasts, what different forecasts mean for you, and how to dig deeper into the components of each forecast.

Calculating your forecasts

Forecast levels

Anatomy of a forecast cell

Forecasts and time resolution

Calculating your forecasts

After you upload your sales order history data, Crisp quickly evaluates your data for factors such as data density, historical coverage, and sales frequency. Crisp then determines the best possible forecasting algorithm it can use in conjunction with your data. Within seconds, Crisp passes your data through the chosen algorithm and calculates the results. Your forecasts are the results of those calculations.

Forecasts can represent many different styles of product packaging: cases, units, pounds, kilos, and more. What your unique forecast represents will depend on the data that you upload into Crisp.

For example, if the sales orders in your data represent the number of cases of a product that were purchased, then Crisp's forecasts will represent the predicted future demand for your product in terms of cases too.

In the screenshot below, the sales order history data uploaded into Crisp reported the amount of the product sold in pounds. The forecasts therefore represent the number of pounds of the product that are expected to be sold in the future.  

In this example, Crisp has forecasted that 3,919 pounds of Food Appetizers (0.33 lbs x 20) will be sold during the week of 3/23-3/29. 

Forecast levels

Crisp provides you with two types of forecasts: product-level and product-by-customer level. By default, your account's products page will display all of your forecasts at the product level. Product-level forecasts predict the overall demand that you will experience across all customers for a single product. This is also known as a top-down forecast.

Crisp also provides you with a product-by-customer forecast. This forecast represents the demand that you are predicted to experience from a specific customer for a specific product. This is also known as a bottom-up forecast.

You can view product-by-customer level forecasts on your Products page. To view these forecasts, click the drilldown button beside the product to display the customers you have selected for forecasting:

In the example above, Crisp has predicted that the customer Lion Shack will likely purchase 228 pounds of Clear Moony (0.25 lbs x 30) during the week of 3/23-3/29. The customer Fresh Lite has empty cells in the forecast columns because it may not been selected for forecasting or Crisp does not yet have data for this product and customer combination. 

You may notice that the product-level or "top line" forecast does not reflect the aggregate of the forecasts for each customer. This is because the product-level forecast reflects the expected demand across all customers, not just a few selected customers. In general, the product-level forecast is going to use the most of your data in order to generate a forecast. More data often results in more accurate forecasts.

Uploading as much data as possible for your biggest customers will increase Crisp's ability to generate accurate forecasts for you at the product-by-customer level. 
 

Anatomy of a forecast cell

Clicking on a forecast cell on the products page will cause the cell inspector to pop open:

The cell inspector provides you with information related to the forecast, such as what demand drivers were identified and how much those drivers contributed to the forecast estimate. The cell inspector also lists any events that might be present and allows you to quickly add events related to this specific forecast.

In the example above, the majority of the forecast was based on the product's overall trend, which Crisp has calculated from the data that was uploaded for this product. There is also a small seasonality effect that increased demand during this period. Finally, someone entered an upcoming event, which Crisp has included as a positive demand driver. 
 

Forecasts and time resolution

Forecasts are displayed depending on the time resolution that you have selected. Adjusting the time resolution on your products page will change your forecasts as well. 

For example, setting the time resolution to this:



Will result in your forecasts being displayed like this:

In this example, on the left of the black dividing line, 2 columns of sales history are displayed along with an average for the last four weeks of sales. The Last Week cell is blank because sales order history data from last week has not yet been uploaded into Crisp. On the right of the dividing line, the next 4 weeks of forecasts are displayed including a column for the next 13 weeks total. 

Changing the time resolution to Daily or Monthly would result in very different sales history and forecast numbers being displayed on your products page. Increasing or decreasing the number of sales history and forecast columns will enable you to expand or narrow down the amount of data that you want to view too. Select the time resolution and number of columns that most closely matches the type of data that you need to be successful! 

You can learn more about adjusting time resolution here.