This is why you don't give the business unreliable metrics. Whoever came up with this requirement at capital one had the sense that email surveillance was capa

This is why you don't give the business unreliable metrics. Whoever came up with this requirement at capital one had the sense that email surveillance was capable of delivering meaningful read metrics. Who let that happen? You may understand the implementation makes the metric comically unreliable, especially for email with so many clients that block images by default. But when is the last time you quantified the margin of error in all the metrics reporting? Never? Great rigor you've got going there, data science. Even if you do clarify how poor the signal is, how long until someone copy-pastes it into Excel, removes that context and shows the graph to someone with more decision making power? And that's the best case scenario. People are, in general, way more statistically illiterate than you suspect. Remember how many people saw Hillary had 85% on fivethirtyeight, and were convinced that it meant she would win? I've had arguments with Amazon employees that: no, they do not know how many people open their emails. Trying to explain what they have is a lower bound that is closer to "how many emails are read in gmail on an iphone" than "how many emails are read"? Nope. The UI says "unread," therefore it is impossible that the user read the email. Stop implementing bullshit metrics in your software.