I feel bad for data scientists that have gotten stuck on ad targeting.
When your job is to increase engagement on something you've shoved in someone's stream or in someone's search results, then you're effectively pissing against the UX wind. That can't feel good. Your job is antagonistic to your users' goals (because your users aren't your customers).
When your job is to increase engagement on something you've shoved in someone's stream or in someone's search results, then you're effectively pissing against the UX wind. That can't feel good. Your job is antagonistic to your users' goals (because your users aren't your customers).
Furthermore, what happens when a graphic designer comes in, removes the beige background from an ad, and produces more ad revenue than your deep learning models ever could?
What's the end game for better ad targeting? Is it the deepest of all deep learning? Hmmm. Maybe it's just making ads blend in visually with real content and duping your users. That's why I feel bad. Who wants to learn a bunch of math only to realize your job is done better by simply removing a background color?
And this is why I love my current job. My team's goals are lined up directly with the user's. The company I'm at makes its money off of monthly subscriptions, and my job is to make the product better using data so that folks want to join and stay. That's much more rewarding.
Now, am I building high end AI models? Occasionally I get to build an AI model I'm proud of. But predictive modeling for its own sake is not the goal.
No, the goal is a better UX. And that means that I can use data in small ways too. I'll give an example.
Users entering customer support would constantly complain to us about reCAPTCHA. Just look at this screencap. Yikes.
And this is why I love my current job. My team's goals are lined up directly with the user's. The company I'm at makes its money off of monthly subscriptions, and my job is to make the product better using data so that folks want to join and stay. That's much more rewarding.
Now, am I building high end AI models? Occasionally I get to build an AI model I'm proud of. But predictive modeling for its own sake is not the goal.
No, the goal is a better UX. And that means that I can use data in small ways too. I'll give an example.
Users entering customer support would constantly complain to us about reCAPTCHA. Just look at this screencap. Yikes.
I fail these humanity tests every other time. And we tried subbing out reCAPTCHA for other more game-ified humanity tests. But those didn't work out.
We realized that our own internal anti-abuse models could be turned on this problem. Now we're able to validate the humanity of a vast number of customers who come to us -- and we're able to just hide reCAPTCHA altogether. Better living through data science!
This project didn't take long, and it's certainly not going to grab headlines. But we didn't do it to increase a KPI, impress the tech blogs, or justify our graduate degrees. The team did it for our users, and that feels good.
We realized that our own internal anti-abuse models could be turned on this problem. Now we're able to validate the humanity of a vast number of customers who come to us -- and we're able to just hide reCAPTCHA altogether. Better living through data science!
This project didn't take long, and it's certainly not going to grab headlines. But we didn't do it to increase a KPI, impress the tech blogs, or justify our graduate degrees. The team did it for our users, and that feels good.
RSS Feed