I know it’s a horribly dead horse-beaten topic now, but (not provided) is a bigger problem than ever for search engine marketers.
If we obsess over it, we lose sight of more important parts of a campaign.
If we don’t address it, the client thinks organic traffic on their bread and butter keywords is plummeting.
Damned if we do, damned if we don’t, right?
And there are several workarounds to guess-timate the keyword data you might be losing, but those are imperfect at best.
There’s a better way.
SEOs are always jealous that PPC folks don’t lose any data to (not provided), but since the two mediums are so symbiotic, why not grok their data on this?
Not only would it be a more educated guess as to what the user was searching for initially, but more importantly, you can glean the intent of these searchers.
Harness Paid Data
Here’s what I mean:
Go to the Top Conversion Paths report in Google Analytics.
Some housekeeping at the top of the report — Make sure only your primary goal is selected: an ecommerce transaction or a lead generated.
Then set the path length to 2 instead of 2 or more. That’ll be most of where the action is, and it’ll make your data export from the report a lot cleaner.
Then, set your date range wide. I like to do year-to-date, but at least go 90 days.
Then click on Other under Primary Dimension and start typing “key” and select Keyword Path.
Once that loads, click Edit next to the Advanced Filter box. Make sure the path includes (not provided) and then exclude your brand name(s) using a Regular Expression like this:
Looking at the first few rows of the data, it’ll be awful. Unavailable just means a non-keyword-driven channel like email, direct or referral:
But when you dig deeper, you’ll find a treasure trove of situations where (not provided) was the first touch or last touch in a string of visits that resulted in a conversion, where there was also a provided keyword! (Blurred out here, but trust me – they’re real!)
Export this out to Excel and you can start doing all kinds of things with it! Separate First Click and Last Click using Text to Columns.
Take some of this information back and at least start making some smarter inferences about the kind of data you’re missing.