Imagine you had 3,000 installations of your app. At a 5% margin of error and 95% confidence level, you only need data from 341 users to estimate the usage behavior for all 3,000 app users. So, if you just had 341 of your app users logged into the app and these users saw the sponsor splash screen on average 50 times during the event, you can then estimate that the sponsor splash screen received about 150,000 impressions (50 impressions on average multiplied by 3,000 installations). From experience, we know that at least 20% of users log into the app, which in this example would nearly double the actual sample size and thereby reducing the margin of error.
Extrapolation assumes that, if you have a sufficient sample size with a comfortable confidence level and margin of error, your findings will be consistent for most all attendees despite data deriving from only a portion of those attendees. The estimates are based on your sample size (logged in users) that meets a certain level of confidence while assuming a margin of error.
The margin of error determines the likelihood of your data consistency. For example, a confidence level of 95% with a margin of error of 2% indicates that, for every iteration of this analysis, we would expect our results to fall within 93-97% accuracy of the initial iteration (+ or - 2% of 95%).
Your sample size is directly correlated to your confidence level and inversely correlated with your margin of error. In other words, as you increase your confidence level, your sample size will need to increase as well to achieve accurate results.
Calculate your required sample size: https://www.surveymonkey.com/mp/sample-size-calculator/