Firefox & Safari add to the "not provided" issue!

Browsers Boost Google's Encrypted Search

Currently Google is blocking the keywords used for organic search for anyone logged into Google (e.g Google+, Gmail...) and hence using Google encrypted search. This means that the proportion of keywords defined as "not provided" is on the increase.  And the the effect?  As we can see below this site's "not provided" traffic has increased by a factor of x18 from 0.58% to 9.2% within 7 month and is still trending up.

Browser Search and "not provided"

So what gives with browser based search? Well it's not good news for data quality:

• Chrome's synchronization logs people into Google by default across all their browsers = not provided

• In Firefox 14 and above, the Google search box defaults to secure = not provided
• Using Safari on iOS6 not only do we see no keywords but we don't even get referer information!

See full article by Adriel Michaud

The upshot of this is once the new browsers begin to deliver "not provided" traffic (or switching "referrer" traffic to "direct") this trending will not only go up but will accelerate the process. So what should we do?

First, it must be noted that this is a deliberate act by Google, Firefox and Apple. It is policy rather than a technical anomaly.

Secondly, this affects all tools and is an universal error.

Finally, at what point do you need to worry about this section of your traffic and your lack of identifiable keywords?

Well if your "not provided" traffic is not above, let's say, 20% then the question you really need to ask is "Does 20% of my keyword audience act any differently to the 80% I can see?" Does the "hidden" audience differ in the way they search? Probably not. So you can probably ignore the "not provided" and analyse the sample of keywords you can see .

When do I worry it's affecting my analysis?

The line in the sand needs to be drawn by you. Any analysis is based on a good sample of data so you need to ensure that's the case. There are many factors that you need to take into consideration (see Statistical Sampling Size information):

• The % of the overall  traffic that contains "not provided" (low good - high bad)
• The number of keywords (low good - high bad)
• Overall traffic numbers (low bad - high good)

For example is a website attracts 1000 visitors a month, the "not provided" traffic is 50% and there are 250 keywords then... boy do I have analysis issues!

On the other hand if I have 1 million visitors a month, my "not provided is 80% and I have 100 keywords I would be confident in my analysis.

It's down to all the factors above. You need to make a judgement!

We have heard of some practical tips that you may want to try.

• Look at landing pages content and filter by keyword "(not provided)". This will give the landing pages from unspecified keywords which give an indication of keywords that may have been used. Eg if users arrive on "/blog/google-tag-manager-not-quite-silver-bullet" chances are they searched for "lynchpin tag manager blog" or a variation of those terms.
• Similar to above but using filters. Write a filter that replaces (not provided) with the Request URI, in other words replaces the (not provided) with the landing page. Thanks to Dan Barker for that tip.

It will be interesting to see how SEO and SEM agencies cope with the lack of identifiable keywords!