Quote:
Originally Posted by simplicitus
So, Cambridge Analytica can surely tell you that a 25 y/o unmarried or "it's complicated" woman in NE Ohio who likes basset hounds and gin rummy only has a 48% chance to vote (assuming they can get real names and cross-reference them to voting rolls) and a 35% chance to support Trump (assuming they can get that % via some other demographic data, like exit polls, which aren't accurate at such a granular level. They can't get who you voted for from the state or any other source).
So, great, maybe you know where the non-voters are who are likely to support your candidate, but actually getting them to vote or persuading others to support your candidate is the same old problem that campaign managers/workers/supporters have faced for the last 100-200 years.
There is nothing that I can tell that gives them a significant edge other than puffery and confusing clients and journalists. This is just standard silicon valley-type hype that hopes to explain the "conspiratorial" election of Trump.
I agree that CA's ability to sway the election was much lower than some are making it out to be (in part because they didn't actually do some of the stuff they said they could) but here's how this would actually work:
- Take the voter file and cross-reference against a few thousand household-level data providers
- Look for things like registered party, propensity to vote, which primaries they vote in, etc
- Build correlations around the other data and use that to build FB interest and demographic targeting that can be focused down to the 5-digit zip
- Do a ****ton of testing around various kinds of messaging and basically live in the Facebook Insights tool that lets you break performance down by all kinds of variables
The other interesting thing would be how they used email lists to build custom audiences and lookalike audiences for marketing. I assume (and hope) the campaigns have all kinds of info on each email address, what emails they open, what they do on the site, etc, and can come up with strategies for advertising different messages to them on other platforms.