Desktop dark social mystery solved?
Update: Yahoo homepage traffic is now categorized properly on Chartbeat, and is more impactful than people thought.
Earlier this week, I wrote about an audience anomaly on ChicagoTribune.com that I was terming desktop dark social, tho I doubted it was social despite the classification by Adobe Analytics and Chartbeat. Turns out, it likely isn’t.
To recap, we were seeing sustained audience spikes typically between 1,000 and 3,000 realtime concurrents on Chartbeat, and as high as 7,000 concurrents. The traffic was considered social, classified as “Email, apps and IM” — consistent with dark social. But, instead of mobile, the browser types were mostly desktop Chrome 48.0 and Internet Explorer 11 — totally inconsistent with dark social.
Given the browser type, anything mobile or social seemed like an unlikely source. But according to Adobe Analytics, most of our referring domains were stable and consistent.
This top five graph highlights some growth typed/bookmarked, but not much else. Google, Facebook, Yahoo and Twitter were driving the most traffic, with Facebook likely secretly counting for much of typed/bookmarked. So I started to go back into November and look for changes, referrers we weren’t seeing now. Adobe Analytics didn’t tell show much, but oddly Chartbeat did.
The listing “hsrd.yahoo.com” is the referral source for a story not from Yahoo search, but the main page of headlines. We weren’t seeing that much anymore and there was a reason, Nick Herzeca from Chartbeat confirmed.
“Yahoo recently changed their HTTP handling to HTTPS, which conceals the referral data and would cause traffic from Yahoo to be bucketed into the dark social category.”
So, could this be simple Yahoo story curation hidden by HTTPs? The browser types make sense. So, in fact, does the size and duration of the audience.
So we needed to catch a few examples in the wild. And it didn’t take long during the week of Feb. 22, 2016. The first example was a small spike of desktop dark social on Monday that drove around 800 concurrents.
Was it on Yahoo? Yes, the 23rd headline down.
One day later, we saw a little larger spike of DDS. About 1,500 concurrents. High velocity and sustained.
And again, it was on Yahoo’s main page, although for me (not logged in) 51 items down.
It feels a bit low on the page to drive that type of traffic, but again, it fits the pattern. And another odd audience bump the next day; same result.
And this again appears to correlate with Yahoo homepage play, 23rd down again (somewhat odd).
So, case solved? Likely so, agrees Nick.
“It’s not surprising Yahoo is still appearing as a referrer in your analytics services because they are not passing through HTTPs across the board. For example, we have seen traffic from Yahoo for hard links (that would appear in your Chartbeat referral sources as hsrd.yahoo.com) pass through HTTPs more frequently than traffic from Yahoo Search (which would appear in your Chartbeat referral sources as r.search.yahoo.com). We have seen almost all traffic from hsrd.yahoo.com from desktop and mobile devices excluded from our named referral sources with tablets being the one device where we continue to see the some referral traffic come in from hsrd.yahoo.com that we can identify. Furthermore, the type of referral traffic that we tend to see from hsrd.yahoo.com (large spikes with high velocity) match the pattern you described from desktop dark social.”
But, it’s important to note that in Adobe Analytics, Yahoo search and Yahoo news headlines all seem to fall under “Yahoo.com” in the standard reports.
Even as Chartbeat was tracking hsrd.yahoo.com, Adobe Analytics doesn’t appear to have done so, thus separating search and news.
So what now?
Without Yahoo adopting UTM codes identifying the referral source as Yahoo, it’s all going to look like dark social, only desktop.
But the good news is that Yahoo links are not Google News fragile, nor as high maintenance. Once you have the audience, you’re good. That bad news is that if you get consistent Yahoo headline referral traffic, it may inflate your search or social success, depending on your metrics platform. And junk audience data is toxic to strategy — especially when it’s already so murky.