3 ms·

At some point they'll just resort to calling the metric "person type" and the gender and age will be baked into it. They'll then use that "person type" metric t

At some point they'll just resort to calling the metric "person type" and the gender and age will be baked into it. They'll then use that "person type" metric to determine what you want to see, hear and consume.

andy99 2y ago
That's existed forever, it's customer segmentation. It's the underlying reason people want this info, age etc is just a proxy. From an advertising perspective nobody cares how old you actually are, they want to know if you're likely to respond to an ad.
- rchaud 2y ago
  Customer segmentation is based on user-submitted info like name, zip code and product SKU. This is using typed communications by anonymous users to make a guess about those attributes.
- johnnyanmac 2y ago
  >nobody cares how old you actually are, they want to know if you're likely to respond to an ad. sadly, people will pay a lot to know the age of someone precisely because they just wanna sell you stuff. Which is the problem in and of itself. I'm glad governments are slowly starting to stamp down on this
jerf 2y ago
Discord says I'm a 21ecaf26-4e27-477d-a6c0-45c2c69a7645 but I identify as a 076db833-35be-4fa8-856e-fe2fa4128cee.
treflop 2y ago
People already do this on their own and advertisers already use it. The Venn diagram of people who subscribe to 'The Economist' minimally overlaps with subscribers to the 'Razorcake' zine and very few companies would want to have an ad in both.
- HPsquared 2y ago
  Cosine similarity is low.
jareklupinski 2y ago
> the metric "person type" "demographic"
- soared 2y ago
  “persona” “segment” https://docs.oracle.com/en/cloud/saas/data-cloud/data-cloud-help-center/Resources/Images/audience_newflow/audnf_audbuilder_key.gif https://docs.oracle.com/en/cloud/saas/data-cloud/data-cloud-...
- Spivak 2y ago
  Yes but demographic has the unfortunate problem of needing to be connected to real world observable traits. Clustering people into as many groups as naturally fall out can capture way more. The downside is that you don't know what the clusters actually mean precisely but you figure that out broadly by running tests.
  - adammarples 2y ago
    There's no specific requirement for this, it's just a word. In this case it's being connected to the real world traits of language use though.
fire_lake 2y ago
An unsupervised approach makes way more sense in some ways but advertisers often like to pick demographics manually so the categories need to be understandable by a human.
- mrbungie 2y ago
  If the advertisers still use A/B testing and WANT to learn from said tests. Sure. If the advertisers use multi-armed bandits you wouldn't care less about the segments themselves.

andy99 2y ago

That's existed forever, it's customer segmentation. It's the underlying reason people want this info, age etc is just a proxy. From an advertising perspective nobody cares how old you actually are, they want to know if you're likely to respond to an ad.

rchaud 2y ago
Customer segmentation is based on user-submitted info like name, zip code and product SKU. This is using typed communications by anonymous users to make a guess about those attributes.
johnnyanmac 2y ago
>nobody cares how old you actually are, they want to know if you're likely to respond to an ad. sadly, people will pay a lot to know the age of someone precisely because they just wanna sell you stuff. Which is the problem in and of itself. I'm glad governments are slowly starting to stamp down on this

jerf 2y ago

Discord says I'm a 21ecaf26-4e27-477d-a6c0-45c2c69a7645 but I identify as a 076db833-35be-4fa8-856e-fe2fa4128cee.

treflop 2y ago

People already do this on their own and advertisers already use it. The Venn diagram of people who subscribe to 'The Economist' minimally overlaps with subscribers to the 'Razorcake' zine and very few companies would want to have an ad in both.

HPsquared 2y ago
Cosine similarity is low.

jareklupinski 2y ago

> the metric "person type" "demographic"

soared 2y ago
“persona” “segment” https://docs.oracle.com/en/cloud/saas/data-cloud/data-cloud-help-center/Resources/Images/audience_newflow/audnf_audbuilder_key.gif https://docs.oracle.com/en/cloud/saas/data-cloud/data-cloud-...
Spivak 2y ago
Yes but demographic has the unfortunate problem of needing to be connected to real world observable traits. Clustering people into as many groups as naturally fall out can capture way more. The downside is that you don't know what the clusters actually mean precisely but you figure that out broadly by running tests.
- adammarples 2y ago
  There's no specific requirement for this, it's just a word. In this case it's being connected to the real world traits of language use though.

fire_lake 2y ago

An unsupervised approach makes way more sense in some ways but advertisers often like to pick demographics manually so the categories need to be understandable by a human.

mrbungie 2y ago
If the advertisers still use A/B testing and WANT to learn from said tests. Sure. If the advertisers use multi-armed bandits you wouldn't care less about the segments themselves.