<b>"+12% uplift" with a 95% interval of [-3%, +27%] is a coin flip dressed as a win.</b>

<b>"+12% uplift" with a 95% interval of [-3%, +27%] is a coin flip dressed as a win.</b>
The point estimate is the least informative number in your readout. The confidence interval (range the true effect plausibly lives in) is the decision. If it straddles zero, you have nothing.

— Report the interval first, the point second.
— Wide interval = underpowered, not "promising."

Read the number, not the story. [95% CI is the verdict]