Averaging percentages gives every post equal weight. The overall series figure divides the sum of actions by the sum of the selected base, so it depends more on posts with a large base. Both calculations are useful if it is clearly labeled which question each one answers.
A verifiable pair of posts
A hypothetical post A received 20 actions with a reach of 100: ER equals 20%. Post B received 40 actions with a reach of 1 000: ER equals 4%. The ordinary average is (20 + 4) / 2 = 12%. The ratio of sums is 60 / 1 100 × 100 = 5,45%.
The average 12% describes the average percentage of one post out of these two. The value 5,45% describes actions relative to the sum of post reaches. This sum is not the unique reach of the entire series: a single account could have seen both publications.
How to reproduce it in the tool
In the ER calculator, select the series, the reach base, and likes only. Prepare a CSV with the header name,reach,likes and two rows: A,100,20 and B,1000,40. Click calculate and compare both results. The likes field here represents the selected type of actions in the educational example; in a working report, do not mix different actions under one column.
If the number of followers is constant, the average ER relative to them divides the sum of actions by the number of posts multiplied by the number of followers. Dividing the entire sum by followers only once will give a different metric — activity for the period.
Which rows to exclude
First choose a single window: for example, 48 hours after publication. Separately decide whether to include pinned and advertising posts. An empty selected metric or a zero base excludes the row from both totals; the number of posts counted is visible in the result. You must not quietly exclude weak publications just because they lower the percentage.
Save the original CSV and the calculation export side by side. For each subsequent series, use the same rules. Differences in the math can also be compared with the weighted calculation of overall conversion, without mixing web visits and social interactions.
If a separate public announcement in Telegram has been prepared for this work, you can check the current view terms. The ER calculation does not determine the quality of purchased activity. Purchased activity does not confirm organic interest, recommendations, enquiries or sales.