What is Margin of Error?

Because you've sampled a subset of the population, not the actual entire population, the population's answer could differ from your sample's answer by +/- the margin of error.

Different ~~men~~ people often see the same subject in different lights

For example, let's say your audience is the Australian population of N=25 million+ people but you only sampled n=1,000 of those Aussies 😱 and 50% of your sample say they love tomato sauce on their meat pies. How precisely, and with what degree of confidence, can we trust that 50% of ALL Aussies (not just 50% of our random sample) actually ❤️ tomato sauce on their meatpie? If we plug these numbers into our moe formula:

n = sample size (n=1,000) x
σ = population (N=25,000,000) standard deviation x
z = z-score (95% confidence = 1.96)

We find that the actual answer for the population could be anywhere between 47% and 53% and thus, the precision of our n=1,000 sample vis-à-vis the N=25,000,000+ population, i.e. our margin of error, is +/-3% with a confidence level of 95%, which means 95/100 times, the population's true answer will fall in this range of 47% to 53%, and only 5/100 times the population's true love of tomato sauce on meatpies will be lower than 47% 😬 or higher than 53% 😁

So when doing market research, your sample's answer is never 100% correct, but the higher your sample size, the more precise your sample's answer is, e.g. sample of n=100 gives you +/-10% moe, while sample of n=1,000 gives you +/-3% moe, while n=10,000 gives you moe (or precision) of +/-1%, at 95% confidence level. So 90% precise vs 97% precision vs 99% precision, but almost never 100% precise UNLESS you sampled every last person in Australia (or the entire world, or the entire universe, or your entire business, or all your customers) with your survey (improbable, if not impossible!)

In Glow's AV2 dashboard, when you tick the margin of error in your dashboard settings, you'll be shown a margin of error for each of the sample sizes, i.e. Total (n), that are being used to calculate the percentages in your table so you can see how precise those percentages are. Conversely, if the sample size is NOT being used to calculate percentages in your table, then it WON'T show a margin of error so it's easy for you to read your table correctly.

Examples

If your table's Value = Count, then because no percentages are being calculated, so no margin of error values appear.

If your table's Value = Row %, then because percentages are being calculated as Cell Count / Row Total (n) , so margin of error values appear next to each of your Row's sample sizes.

If your table's Value = Column %, then because percentages are being calculated as Cell Count / Column Total (n) , so margin of error values appear below each of your Column's sample sizes.

If you're looking for significant differences between age groups e.g. "A little passionate" and you're comparing 41-60 (58%) and 61+ (49%) then the diff is 58-49=9%. To assess whether these are actually different you'd need to add both MOEs (5+6% =11%). Since 11% is greater than 9% they are not likely to be significantly different.

If your table's Value = Total %, then because percentages are being calculated as Cell Count / Total Total (n), so margin of error values appear next to / below the Total sample size for that view in the bottom right corner.

If Value = Sum, then because no percentages are being calculated, so no margin of error value appears

If Value = Average, then we'll use a different margin of error formula to calculate how precise our average is (in MoE v3)

Finally, it should be noted, that for MoE v1, we are only calculating the maximum margin of error (for now), NOT the actual margin of error for each percentage, i.e.

the closer the percentage is to 0% or 100% (min moe), the smaller the margin of error.
but the closer the percentages is to 50% (max moe), the higher the moe.

Thus, for MoE v2, we plan to develop actual margin of error next.

We'll then calculate the moe for average values last in MoE v3.