M&Ms 49: Attribution
Hey Team,
People miss attribute stuff all the time.
You especially notice miss-attribution when someone is convinced the wrong things are right on a topic you know well. When this happens, the other person is either full of shit, and you need to discount everything they say or, more than likely, they've miss-attributed.
I like to give people the benefit of the doubt, primarily because I've missed attributed a lot myself. And looking back, I realize how I may have seemed full of crap too.
So why do we miss-attribute? The main reason is that we want to attribute in the first place. Because we cannot accept that we might not know why something succeeded or failed. So we look for answers, find data, and attribute a story.
But the world is complex.
Seven years ago, I worked for an eCommerce startup as an engineer; I worked closely with the marketing department. And they wanted to attribute their ad-spend to the right advertising channels desperately.
The marketing department wanted to know which advertising channel was responsible for a given sale. You might say, with eCommerce, that's easy, wherever the click came from. But according to marketing principles, that would be the wrong answer because people usually see multiple advertisements before they convert. For example, they see the same laundry detergent advertised on Google, then on Facebook, and they finally decide to click and buy from The New York Times advertisement. If we incorrectly only credit The New York Times with that sale, we might overspend precious advertising dollars on a channel that's pretty crappy.
So to address this problem, we paid for a service that did multi-touch attribution. But the challenge is, it is impossible to track people across the web entirely. Unless you were Google or Facebook, whose OS, browser, and pixels are tracking all of us across sites. And more recently, at least on iOS, Apple has crippled that visibility and tracking. But, our partner could not do any of that tracking. So they used a complex model to attribute percentages of credit for the sale to each advertising channel.
Except most of it was bullshit.
How do I know? Because one channel was responsible for over 80% of our sales, most of the other advertising channels didn't do anything. We could have turned the other 20 advertising channels off and still got over 80% of our sales. In fact, that happened a few times.
My friend Chris Wong explains how they attributed things to data at one of his older jobs. They didn't even bother to use the models; they just made up the story and decision first.
"There was the job where management didn't even use the analysis that my group created. Decisions were made and then we would come up with analysis that justified a preconceived decision. I could have done nothing and there would have been the same result. I know it's true because I tested it."
People miss-attribute all the time, either unwittingly or very willingly, to match their story. So, what chance do regular people have if all these data scientists can miss-attribute things with terabytes of ad impression data and fancy models? Not much.
I've found it helpful lately to step back and ask myself, how would I proceed if I knew I couldn't attribute my wins and losses until it was painfully obvious?
Now I ask you, could you be miss-attributing some things in your life?
One Article I read this week:
This long and detailed article by Morgan Housel about the U.S. economy post World War II is excellent.
Given all of the changes going on in the world, I am convinced nations will prioritize security over trade efficiency in the coming decades. This may lead to various tradeoffs, some of which could be a lower global GDP in the short term.
I found this article to be quite good on how we got here.
Two Tweets from this week:
A fascinating tweet thread on how the indigenous Hadza tribe in Africa lives.
There are many great takeaways about how we used to live in the past. We've spent a lot longer living like the Hadza tribe than we spent living the way we do now.
A provocative statement by Naval that I am not sure I agree with anymore.
The world has changed a lot in the last two decades, and more products exist today than before.
To stand out in that sea of things, you need more than just a great product.
In fact there are many example now of great products getting beat by mediocre ones with excellent distribution.
Two Memes from this week:
This meme has been going around in various forms and basically says that because we spend so much on the military, we cannot afford universal healthcare as a nation.
I don't think that's necessarily true, but the meme is still funny in these tense times.
What do you see? Fish, mermaid, seal, donkey.
Thank you for reading. I hope you have a wonderful weekend.
Louie
P.S. you can respond directly to this email and I will do my best to reply. I'd love to hear from you.