I wish these sorts of insta-surveys would disclose more about how the polling was done. The cheapest way to do polling is via a pop-up notifcation on a heavily trafficked digital property, with online button-clicking to follow.

But then you aren't really getting data on all Americans. You're getting data on people who happened to be hanging out on cnn.com, smokinggun.com or whatever. And that's going to skew quite severely to people who have above average digital activity.

Same problem if they're using a ready-built panel of people who like to take online polls.

They might be surveying differently, and spending more money. But given that the publishing organization, decisiondata.org, is in the business of guiding people on their ISP choices, I don't think this is likely to be a demographically balanced phone survey, let alone Census-grade field work.

Sure, lots of us do rely on home internet to do a lot of work, especially now. But any effort to quantify this really ought to avoid big-time sampling problems.