Attribution (how do I prove the ad I bought actually influenced a person’s decision to do something) is still not a solved problem. If a company’s furniture business increased revenue 10% in a quarter, what percent of that was due to marketing. Also which campaign attributed the most to that increase? Because companies often run multiple campaigns through different channels and vendors.
If I can prove spending 50M on TV ads resulted in 150M in additional sales, that company would probably spend 150M on the next campaign to try and generate 450M.
The problem is what kind of data it takes to prove attribution. If I could say ID 123 saw an impression on Jan 31, made an Internet search in that vertical on Feb 1, and traveled to a location with that product on Feb 2. That would be pretty fucking convincing, but it also involves knowing ID 123 person’s activities in extreme granularity.
That’s a use case for selling furniture, but what else could you accomplish? Honey pot people seeking abortions in places where it use illegal? Identify people who maybe politically subversive. Scams on scams.
Pretty much anyone with a budget and a goal can access a commercial surveillance network.
I used to work in ad-tech and this is exactly right. We had a lot of problems that caused a considerable amount of noise on the results:
Cookie trackers getting added to the main blocklist for ad blockers basically rendered them useless
IP tracking was basically done by matching IPv4 addresses which are not permanent. Apple launched a feature which made every person who had ever paid for an iCloud subscription resolve to the same IP in any given country.
People generate so much data that it’s actually hard to try to tell a meaningful “story” out of it. Like GP is saying 90% of a marketing budget is wasted but it’s hard to tell which 90% that is.
I also think “attribution” is fundamentally flawed. Like the Coca Cola ads they show at Christmas are the most successful ad campaign ever but no one goes on their website after watching one of them.
Attribution (how do I prove the ad I bought actually influenced a person’s decision to do something) is still not a solved problem. If a company’s furniture business increased revenue 10% in a quarter, what percent of that was due to marketing. Also which campaign attributed the most to that increase? Because companies often run multiple campaigns through different channels and vendors.
If I can prove spending 50M on TV ads resulted in 150M in additional sales, that company would probably spend 150M on the next campaign to try and generate 450M.
The problem is what kind of data it takes to prove attribution. If I could say ID 123 saw an impression on Jan 31, made an Internet search in that vertical on Feb 1, and traveled to a location with that product on Feb 2. That would be pretty fucking convincing, but it also involves knowing ID 123 person’s activities in extreme granularity.
That’s a use case for selling furniture, but what else could you accomplish? Honey pot people seeking abortions in places where it use illegal? Identify people who maybe politically subversive. Scams on scams.
Pretty much anyone with a budget and a goal can access a commercial surveillance network.
I used to work in ad-tech and this is exactly right. We had a lot of problems that caused a considerable amount of noise on the results:
I also think “attribution” is fundamentally flawed. Like the Coca Cola ads they show at Christmas are the most successful ad campaign ever but no one goes on their website after watching one of them.