Chartbeat creates a new Referrer Pickup when a story first receives traffic from a new referrer.
As you might expect, the most popular stories on your site might see many unique traffic referrers as these stories are picked up by other outlets and shared across social media. To avoid triggering too many Referrer Pickups, we filter each story's list of Pickups to only include referrers that are predictive of significant traffic.
But aren't all referrers predictive of traffic going up?
It turns out that the answer is no. Some referrers tend to pick up a story early in its life and drive many additional readers. Others (certain search engines, for example) tend to pick up a story later when readership is already starting to decline. Our pickups aim to differentiate between the two.
To predict whether a given referrer is likely to drive an increase in traffic after a pickup, we build a model every day for every referrer to each of your unique sites. This model measures traffic on each story, before and after the pickup, to see whether a pickup is typically followed by an increase in traffic.
Referrer Pickups differ from Spike Alerts. Spike Alerts measure whether a story's traffic has already started to increase in a dramatic way. Referrer Pickups try to form a very early alert that a story's traffic may increase.