![]() ![]() Maybe the fiscal year ended around the same time and all of your users got big tax returns. Sales may go up after the embiggening of the button, but you still can’t attribute that to the changes you made to the button. More people click on it, and sales go up, right? In reality, there’s no reason to believe that this is the case. Maybe you make it even bigger and even redder. Because of this conclusion, you want more people to click on the button. It’s very tempting to conclude that clicking on the button made them more likely to buy something. The analytics tell you that users who click the big red button in the middle of the home page buy something 6% more often than users who don’t. The reality could just as easily be that some of your users have an attribute Z that causes them to do A and B more often.Īs a practical example, let’s imagine that you run an ecommerce site. Saying that doing A causes them to do B is adding more information to the story. The information you actually have is very simple - users who do A are more likely to do B. It seems correct on the surface, but in reality you’re making up information. This is not the correct conclusion, and is very easy to miss. I hear it all the time while working, too - users who perform action A do thing B at a rate that’s higher than the baseline, so action A must increase the probability that people do B. Young people who have moved overseas may have higher salaries, but there’s no reason to believe that the moving caused their salaries to be higher. ![]() It claims that moving overseas when you’re young boosts your salary. This bloomberg article is a classic example. Every nerd worth their salt knows this, but falls victim to it all the time. A classic example of this is “correlation does not imply causation”. Christian Scott Christian Scott twitter github email me Big Red ButtonsĪ lot of “good advice” that’s easy to internalise is also easy to ignore. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |