# Forecast sales using weighted average variance

Updated Aug 26, 2020

Forecasting your sales using a Weighted Average Variance is similar to calculating with Weighted Averages, except this formula actually lets you eliminate random outliers from being included in the calculations. TimeForge offers a very customizable Sales feature, and this lesson is designed to help you navigate through setting up a weighted average variance forecasting configuration painlessly. After configuring Sales Categories, you’ll likely want to consider setting up ShiftBuilder Rules so you can start generating schedules with the ShiftBuilder based on your Sales Projections.

## Understanding weighted average variance.

Forecasting your sales using a Weighted Average Variance can help to eliminate random outliers from being included in the calculations. It's what happens when you're calculating a grade for a class, but the teacher lets you drop your lowest grade, but only if it's 30 points lower than the average of your other grades.

The following is an explanation of each setting under the "Weighted Average Variance" category. ### 1. Look at last year, month, or week? By day or by date? Should the algorithm pay attention to a day from last week, from last month, or from last year? Select whichever option is the best indicator for sales.

Should the algorithm search by date or by day? For example, If it's forecasting for June 2, 2013, should it look at June 2, 2012 (date) or the first Sunday in June in 2012 (day).

### 2. Look for how many days? By day or by week? Enter into this field the number of days that the alogorithm should consider. Additionally, select whether the algorithm should look for these days by day or by week. In the example above, we're considering 8 weeks in the future and 8 weeks prior to today to find the forecasted number. (It would be 8 days if "day" was selected)

### 3. Use forecasts if actuals aren't available? Or skip? This option indicates what TimeForge should  do if there are no Actual Sales for the algorithm to calculate from. Choose to either use the forecasted sales for that date as the basis for the current forecast, or simply skip that day.

### 4. Calculate variance by 1-hour intervals? Since our forecasts are based on hourly sales, this option determines if we run the algorithm based on 1-hour intervals or based on 15 minute intervals (if unchecked, we use 15 minute intervals).

### 5. Use Median Averages? This option determines if we use median averages or raw data. If you check this box, the algorithm will use median averages with the percentage in step 6 to exclude any "random" sales data.

### 6. Enter the allowed range for the median average. How far from the "middle" should we base the forecast on? In the example above, any data above or below 33.3% of the average will be excluded from the forecast calculations.

Still need help?