Read on Twitter
If you're in #marketing or #marketresearch, I think it's crucial to understand election polling vs. results. Why?
A thread.
This year, polls in some states nailed it; e.g. WA, where variance was +/-1%. But in other states, polls had >2X their margin of error (e.g. FL).
/1
A thread.
This year, polls in some states nailed it; e.g. WA, where variance was +/-1%. But in other states, polls had >2X their margin of error (e.g. FL).
/1
You need to make decisions about where & how to invest for the best ROI. Opinion + behavior data, at scale, is crucial to your success.
Just like it is to political campaigns! As this NYT graphic shows, campaigns choose where to invest based on poll data. /2
Just like it is to political campaigns! As this NYT graphic shows, campaigns choose where to invest based on poll data. /2
This US election showed us some wise moves, and some not-so-wise ones. E.G.
- Biden probably should have put more in GA and less in FL
- Trump probably should have put more in AZ and less in IA
Bad data = bias = mistakes = lower ROI
/3
- Biden probably should have put more in GA and less in FL
- Trump probably should have put more in AZ and less in IA
Bad data = bias = mistakes = lower ROI
/3
Same thing holds true in market research. If you rely on surveys as a substitute for behavioral data, you're gonna have a bad time.
Surveys are pretty solid for collecting info like location, opinions on a familiar topic, salaries (e.g. this one from @Chenected).
/4
Surveys are pretty solid for collecting info like location, opinions on a familiar topic, salaries (e.g. this one from @Chenected).
/4
But if you survey chemical engineers w/ questions like:
"What topics do you regularly discuss, search-for, and engage-with online?"
or "What publications & people do you follow in your field?"
The answers you get are unlikely to be helpful, accurate, or complete. /5
"What topics do you regularly discuss, search-for, and engage-with online?"
or "What publications & people do you follow in your field?"
The answers you get are unlikely to be helpful, accurate, or complete. /5
Similarly, even if you ask the right questions, you'll get bad data if you don't correctly balance the surveyed audience.
In the 2020 election, many analysts attribute poor polling results to this mistake. Incorrect presumptions about demographic patterns set 'em back. /6
In the 2020 election, many analysts attribute poor polling results to this mistake. Incorrect presumptions about demographic patterns set 'em back. /6
If you want behavioral data at scale, don't ask; gather.
But it's CRITICAL that when you gather, you understand the biases of the tech you're using and the audiences they can analyze.
E.G. Even my startup, https://sparktoro.com , which has cool data, has clear bias issues. /7
But it's CRITICAL that when you gather, you understand the biases of the tech you're using and the audiences they can analyze.
E.G. Even my startup, https://sparktoro.com , which has cool data, has clear bias issues. /7
Just as a platform like SurveyMonkey can't account for audiences that didn't take a particular survey, SparkToro can't account for folks who don't use social media platforms, have private accounts, don't talk about/identify themselves in their accounts, etc. /8
No matter what system you use, it's your responsibility to understand:
- The audience it covers
- The methodology it uses
- The data is gathers and misses
- Skew, sampling, and accuracy
More detail on all this here: https://sparktoro.com/blog/election-polls-vs-results-can-help-explain-how-sparktoros-data-works/
- The audience it covers
- The methodology it uses
- The data is gathers and misses
- Skew, sampling, and accuracy
More detail on all this here: https://sparktoro.com/blog/election-polls-vs-results-can-help-explain-how-sparktoros-data-works/
