Back in July a colleague shared a blog post she’d discovered about how Twitter drives four time as much traffic as you think by Jonathan on awe.sm
The blog explains how referrers are a poor way of attributing traffic from social sharing, in that much of our social traffic actually shows up under ‘Direct’ (Google Analytics) or ‘Typed/Bookmarked’ (SiteCatalyst), rather than under t.co, facebook.com etc or ‘social sites’, as the analytics programme is unable to attribute the original referrer.
This is due to a number of reasons:
- When a user clicks on a link in any kind of non-browser client (this could be Outlook or some of the mobile/tablet apps you use for social media), no referrer information is provided for that visit (if you didn’t come from a web page then you must have typed it into your browser) – so it’s attributed to ‘Typed/Bookmarked’ and you’ve just lost some of your social traffic. This also applies where people are moving from a secure site (https) to a non-secure one.
- This means that a lot of social traffic is invisible to our existing analytics programmes and as a result, we’re not getting a true measurement of the value it’s providing us in terms of traffic and conversions. This unattributable social traffic is being called ‘Dark Social’.
So how can we start to get a handle on this situation? Well, a few months after the initial blog post, we stumbled across another one – ‘Dark Social: We have the hole history of the web wrong’ by Alexis C. Madrigal on The Atlantic.
Alexis explores how to take this ‘Typed/Bookmarked’ traffic and filter out the true typed/bookmarked pages, leaving us with the unattributed social media traffic.
I’ve been experimenting with creating a segment within SiteCatalyst which applies to all visits where the referrer equals Typed/Bookmarked, then excluded all other referrer types, all tracking codes, all newsletter URLs and website sections (these are more likely to be typed into the browser or bookmarked).
What we’re left with is unattributed URLs, which don’t have a referrer, and are too long to be reasonably typed in by hand. This is our ‘Dark Social’. So far the results look positive, I need to show the segment around to some more skilled SiteCat experts than I to make sure it’s pulling the right data, but assuming I’m not far off, it’s looking like 25% of our visits and unique visitors could be attributed to Dark Social – a massive increase on what we’re reporting today.
Have you had any successes with tracking Dark Social?