Ask HN: Does targeted advertising work at all?

Adblock Plus for Chrome: https://chrome.google.com/webstore/detail/adblock-plus-free-...

It takes 2 clicks to install, definitely less than writing on Hacker News.

6 years in adtech, lets say advertisers can distribute a lot of budget and waste it just to get 1 sale and cover for the losses so the numbers make sense. I have seen conversion rates increase 40% with purchase intent (if we know that a user added something to basket).

So what you see is an adtech tax - partially wastage but people will overlook this if they are still profitable on their marketing ROI.

The ad and pitch on the website was typical for a European higher-end brand that has local history but not the name recognition to really qualify internationally as luxury. They told a credible story of the artisanal tradition and so on. I knew already Portugal had a good reputation for leather products. The risk of buying a pair and maybe having to return it didn’t seem high.

I think I was living in London at the time (pre-Brexit). There’s obviously no shortage of nice shoes in England, but I was new there, so finding a physical store I like was a lot more work.

> Do you think Nike would be happy if their vendor just said, “yeah everything drives positive lift”.

It's fairly known in the industry (worked in a range of different ad-tech/marketing companies for around a decade), that more or less this is what makes Nike happy.

> They are employed by neutral third parties who make money regardless of the outcome of the study.

This is a bit naive. Same is true for home appraisers, funny how it always works out that the house appraised for just over what you offered (even in insane markets).

This issue is you might get paid this time for being honest, but you won't get paid again.

> completed by literal data scientists and statisticians.

I'm both of these. Saying that "adtech is bullshit" is absolutely too reductionist but you are wildly too naive about the reality of the industry.

Take A/B testing as an example. I've helped people run A/B tests for many years. I establish rigorous testing setups, make sure people understand confidence in the results, distribution of possible outcomes, etc. I'm honest in my work and so are the people running the tests.

But the truth is that A/B test on customer populations are not controlled experiments. One thing that literally no one in the industry does is go back and review the results of the past years A/B tests. The reason why? People don't really want to know. However, it's obvious these results can't be quite right because teams will run 10 A/B tests in a year, each claiming a 5% improvement in conversion, but clearly you don't see this cumulative impact of a >60% increase in conversion over time.

No one want to know this though, because knowing this helps nobody. Nobody is outright lying, but nobody wants to ask. The one time I worked outside of adtech but saw a problem where model results were almost entirely random noise I got in huge trouble for pointing this out.

Same is true across the board in the adtech world.

Someone else recommended "Subprime Attention Crisis by Tim Hwang" and I very strongly recommend that you read this. It's very well researched and was so inline with my experience that I found it so obvious as to almost be dull.

Yes