Tool To Calculate Error Bars / Confidence Intervals / Lower Bounds For Marketing Funnels

I wrote about the risk of trying to use the old Central Limit Theorem for marketing funnel conversion in this article. The gist is that you should use the Exact Binomial Test, or EBT. It better matches the binomial distribution / Bernoulli process, which applies in situations like marketing conversion funnels (and other processes with binary outcomes, like medical intervention efficacy).

In marketing, we usually don't care if the actual conversion rate is higher than our estimate or lower bound. So we can use a one-sided test. The "upper bound" of a one-sided test is always 100% - e.g. everyone converts. That would be great! What we are worried is about is too few people convert. So we want to set a lower bound on conversions, within a confidence level like 90%, 95%, or 99%.

If this was helpful, please let me know! And I was thinking about creating another feature where you could get a nicely formatted PDF of the same analysis, with additional fields for date range, source of data, type of observation, and type of conversion. If that would be interesting, please let me know. Email me at yaoshiang@gmail.com.