Forecasting a Month of Income From Your Analytics

Start from a base

Take the last 28 days of estimated revenue as your base. Do not use all-time averages because they flatten seasonal swings. For each of the next four weeks, project using the same weekday mix as the base window. This simple adjustment removes most weekday noise.

Split the base into known stable videos and new uploads. Treat new uploads as zero revenue until they reach the minimum monetized views you set. That keeps your forecast conservative.

Adjust for known shifts

Apply one adjustment at a time. If a traffic source is trending up for three weeks, adjust that source's revenue by a small percent. If RPM fell for three weeks across your top videos, subtract a small percent. Do not apply five adjustments at once. Overfitting the forecast creates false confidence.

Factor in any scheduled uploads by multiplying expected RPM by a conservative estimate of monetized views based on your last similar video. If you have no similar video, do not forecast revenue for it.

Track error weekly

At the end of each week, record actual estimated revenue against forecast for that week. Keep a running error rate over the last four weeks. If error stays above your tolerance for two weeks, lower your adjustments. If error is small, you can be slightly bolder next month.

Do not rewrite the whole forecast every day. Rewrite once per week using the same method. Changing the method hides error and makes it hard to learn.

Key takeaways

• Use the last 28 days with weekday adjustment.
• Forecast new uploads only after a minimum threshold.
• Apply at most one or two small adjustments.
• Track weekly error for four weeks.
• Keep the same method to learn from mistakes.

A conservative forecast you actually follow is better than a perfect forecast you ignore.

Make your first forecast

Project week one to week four now using the last 28 days. Write your adjustments and your error tolerance. Then check against actuals next Friday. Adjust only the size of changes, not the whole process, when error runs high.