Facebook Graph Search launched in January, and I was fortunate to have early access. Upon gaining access, I began heavily testing Facebook’s new social search capabilities. That research led to my first post covering my initial thoughts about Graph Search, including its impact on local search, privacy, reputation management, etc. You should check out that post out after reading through this one.
Graph Search and Not Provided
One (unfortunate) point I covered in my post was the apparent use of “not provided” in Graph Search. If you’re not familiar with “not provided”, last year Google started encrypting searches for users that were logged into a Google account. As a result, the keywords that users searched for would not be passed in the referrer when they visited your website. Essentially, webmasters started seeing “not provided” show up in their analytics reporting versus the organic keywords that led to the site. Needless to say, this was a huge problem and the “not provided” numbers have been growing ever since. For some websites, “not provided” accounts for over 60% of all organic searches.
Now back to Facebook Graph Search and how it handles searches. When Graph Search launched, I quickly realized that Graph Search falls back to Bing results when it can’t answer your query. You can read my post about Graph Search to learn about the autocomplete funnel that you get forced down. So, if you aren’t searching for photos, people, places, or interests, then Facebook will display the “web results” for the query (which is basically the Bing results).
During my testing, I quickly checked the referrer when clicking from the fallback results to websites and noticed that Facebook was not passing the keyword in the referrer! It was basically Facebook’s version of “not provided”. When you checked Google Analytics, the visit looked like a typical referring source (a “Social” visit from Facebook). You couldn’t see the keyword that triggered the visit. Needless to say, I thought Facebook was making a big mistake. If Facebook passed the keywords along, business owners, advertisers, SEO’s, etc. would love them for doing so… We could all start to track the impact of Graph Search (at least any searches that fall back to Bing). I guess we were out of luck… or were we?
SES NY Session Research and the “Referrer” Moment
SES NY is this upcoming week, and my session covers Facebook Graph Search (based on the research I mentioned earlier). While I was finalizing my presentation over the past week, I noticed something very interesting. I started a slide on “not provided”, and then double-checked some searches leading from Facebook to external websites. What I found surprised me!
I noticed that the keyword was now being passed along in the referrer! Below is an example of a new referrer. You can clearly see the q={keyword} being passed now.
That’s awesome, but now we need to track the keywords. Read on.
Google Analytics and Graph Search
As you can guess, I jumped into Google Analytics to see how this was being picked up. Since Facebook isn’t an official search engine in GA, it was still showing up as a referring site (without the keyword showing up). But, since the q= querystring parameter was being passed in the referrer, I knew I could surface those keywords via advanced filters. So, I quickly set up a new profile and added a filter that would capture graph searches from Facebook. And it works. I explain how to quickly set this up below.
How to Track Graph Searches in GA
I’m not going to explain the inner workings of advanced filters in this post (it could easily be a series of posts). Instead, I’ll show you how to quickly set up an advanced filter that will capture graph search keywords, and then report them in Google Analytics. Note, you can tackle this several ways, and this is just one solution. Feel free to tailor your own GA setup as you see fit.
1. Launch Google Analytics and Access Your Website Profiles
You can access your profiles by clicking the “Admin” button in the upper right-hand corner of the GA interface. The first tab should list all profiles for the website at hand (for the “property” in GA).
2. Create a New Profile
Click the “New Profile” button and give it a name. Then choose your reporting time zone. Then click the “Create Profile” button.
3. Add an Advanced Filter
Once you create your profile, you should click the “Filters” tab so you can add an advanced filter.
4. Adding a New Filter
Click the “New Filter” button, and then make sure the settings match what is list below. We will name the filter, create a custom filter, select advanced, and then enter patterns to match in order to capture graph search keywords (and then report them in GA).
5. Save the Filter
Click “Save” at the bottom of the form and you should see your new filter listed for your profile.
6. Check Your Reporting
Now you need to wait for graph searches to come in. Note, this filter will just pick up graph searches that fall back to Bing and that lead to your site. If you have access to Graph Search, then you can quickly test it out by searching Facebook for a query that your site ranks for in Bing organic. Then click through the Facebook Graph Search results and check your reporting later on (the standard GA reporting isn’t real-time, so the reports will lag slightly).
To find the new data, you can click the “Sources” Tab, and then click “All Traffic”. Then find “facebook.com” and click through. Then you can add a secondary dimension for “User Defined Value” which will contain your graph search keywords. When setting the secondary dimension in your reporting, click “Content” and then select “User Defined Value”. Note, the new advanced filter will capture all q={keyword} combinations. That’s why you need to drill into “facebook.com” in the sources to isolate Graph Search results.
Your reporting will look something like this:
Once you start picking up graph searches, you can tie those searches to site activity, performance, conversion, etc. In addition, we can start to see how many graph searches are being conducted that drive users off of Facebook.com (meaning those searches can’t be answered yet by Facebook Graph Search).
Summary – Capture Graph Search Keywords
Again, I’m glad to see that Facebook is passing keywords along in the referrer! For a while, I thought we were going to be stuck with another “not provided” situation. Although Facebook Graph Search has only been rolled out to a few hundred thousand users, it’s only a matter of time before the 1 billion+ active user base gains access. And that will be across mobile devices too. My recommendation is to set up this advanced filter today and start tracking graph searches. You never know what you are going to find. :)
Last, but not least, if you are attending SES NY this week, and are interested in learning more about Facebook Graph Search, you should definitely check out my session. It’s at 2:15 on Thursday, March 28th. I’ll be covering a range of important findings, based on my research.
GG