My core issue is the unfulfilled promise of the panopticon.
I'm still waiting for the recommenders, personalizers which help me.
Not boost engagement. Not amplify tiny differences. Not catalysts for virality.
I was on the recommender, personalization team for a high end fashion retailer. Joining, I thought "Woohoo! Teach the computer tell me which dress shirt to buy! Pick the right t-shirts! Find tasteful but understated socks! Finally!"
It took me a while to peel back all the layers to reveal the team's secret sauce. Turns out there isn't any. The most performant algorithm was "stuff you've looked at before" (~70%), followed by "what's hot" and "what's new".
While most of our effort was put into all the Big Data Machine Learning Booyah stuff, I'd characterize the attributable "lift" as little better than noise. Terrible ROI. We would have been MUCH BETTER off improving the data quality, search features, and browsing experience.
(I had some other more radical ideas. A whole thesis built around authenticity and actual engagement. Way past StitchFix. Alas, too weird for the brick & mortar types. Imagine explaining TikTok influencers to your great aunt. But I'd be happy to have someone pay me for a brain dump.)
In conclusion, nothing this last decade has shaken my hunch that digital ads are a giant con job. At least outside of political advertising. (My bro has worked in ad tech for 15+ (?) years. Our spirited debate has never stopped.)
I'm still waiting for the recommenders, personalizers which help me.
Not boost engagement. Not amplify tiny differences. Not catalysts for virality.
I was on the recommender, personalization team for a high end fashion retailer. Joining, I thought "Woohoo! Teach the computer tell me which dress shirt to buy! Pick the right t-shirts! Find tasteful but understated socks! Finally!"
It took me a while to peel back all the layers to reveal the team's secret sauce. Turns out there isn't any. The most performant algorithm was "stuff you've looked at before" (~70%), followed by "what's hot" and "what's new".
While most of our effort was put into all the Big Data Machine Learning Booyah stuff, I'd characterize the attributable "lift" as little better than noise. Terrible ROI. We would have been MUCH BETTER off improving the data quality, search features, and browsing experience.
(I had some other more radical ideas. A whole thesis built around authenticity and actual engagement. Way past StitchFix. Alas, too weird for the brick & mortar types. Imagine explaining TikTok influencers to your great aunt. But I'd be happy to have someone pay me for a brain dump.)
In conclusion, nothing this last decade has shaken my hunch that digital ads are a giant con job. At least outside of political advertising. (My bro has worked in ad tech for 15+ (?) years. Our spirited debate has never stopped.)