|Understanding Google Webmaster Tools Search Queries CTR|
Having seen CTR reported in Search Queries much lower than what I would have expected, I have done some tests in order to understand how Google calculates it.
I looked at a branded search in a small period of time where I know the site is ranking as #1 but where CTR was being reported as only 26%. For branded searches I would normally expect this percentage to be close to 100%.
So for this specific search, filtered on the country of Austria, Google WMT reports the following:
Impressions 89, Clicks 23, CTR 26%, Avg. pos #1
Drilling down by clicking on this specific search query and amalgamating information on page that ranked and a position in search results, I get the following info:
#1 German Home Page, Impressions 23, Clicks 12, CTR 52%
#2 German About Us Page, Impressions 21, Clicks 9, CTR 39%
#3 German Product type list page, Impressions 21, Clicks 2, CTR 9%
#4 Site Home Page (English), Impressions 23, Clicks 0, CTR 0%
#5 German a particular product page, Impressions 1, Clicks 0, CTR 0%
Adding up all impressions reported above results in 89 impressions and adding up all clicks results in 23 clicks.
However, being this a brand name search, and knowing that the home page will always rank for it, it was easy to figure out that there were only 23 searches for this query, which resulted with 23 clicks.
So out of 23 searches, every one resulted in click through to a site. The 100% CTR would be more appropriate number to report.
The additional 66 impressions were result of multiple pages being shown in SERPs for these 23 searches and they result in WMT understating CTR for queries where more than one page ranks for the same query.
Unfortunately, WMT does not show "Searches" column that would say how many times a query was searched for in a selected period, and then calculating CTR based on number of searches for a query instead of on number of impressions.
If you have only one page ranking for a specific query, WMT CTR would be more correct in terms of how many visitors that searched for that query, clicked through to your site.
Great Info -- Thanks!
That is kind of interesting and might explain why you see much host crowding and it might also explain why so many ecommerce sites took a hit and can't recover.
|The additional 66 impressions were result of multiple pages being shown in SERPs for these 23 searches |
That helps explain why the reported numbers for "searches" and "pages" are so wildly different. There are still some missing pieces, though. I've got places where there's a one-on-one correspondence: only one page could possibly bring up results for this query. And they still show different numbers.
:: detour to gwt ::
gwt defaults to my personal site (just two little-used directories, plus a bunch of non-indexed stuff). According to the "queries" table, nobody has come to this site from google in the past month. Literally nobody, zero, straight down the list. It is then a little mystifying to switch to the "pages" tab ... and suddenly find solid numbers. (Which I already knew, from logs.) There's even one given as 100%.
Sandra, did anyone ever figure out what the difference is between "Impressions" (bigger number) and "Displaying" (smaller number) at the top of the Searches table? I don't mean "what's the difference", of course; that's obvious. I mean, what's the criterion? It can't be raw numbers. These are tiny sites, so the list goes right down to 1 and 0. Pages shows the (same) bigger number for both "Impressions" and "Displaying", but I seem to remember it isn't always an exact match.
|Sandra, did anyone ever figure out what the difference is between "Impressions" (bigger number) and "Displaying" (smaller number) at the top of the Searches table? I don't mean "what's the difference", of course; that's obvious. I mean, what's the criterion? |
I always thought that "Impressions" is the total number of results that were served on SERPs. So 20 searches, 15 of which showed one entry in SERPs, 3 of them showed 2 entries on page 1 and for 2 of searches there were 2 results on page 1 and one further result on page 2 where 2 times visitors did paginate to page 2 would result in 15 x 1 + 3 x 2 + 2 x 2 + 2 x 1 = 27 "Impressions"
My understanding of "Displaying" is how many Google displays in Search Queries in WMT. Basically, for whatever reason, they withhold some data, even when the site is small enough (not much queries/traffic) to show all of it.
So if you were to add all impressions that are reported in Search Queries part of WMT broken down by keyword, you would get to "Displaying" number.
Or better - below the chart you have two buttons.
- If you click on the first [Download this table] and then add up "Impressions" column, you get "Displaying" total shown above the chart.
- If you click on the second button [Download chart data], this gives you impressions per day and when you download this data and add up Impressions column, you get "Impressions" total shown above the chart.
It basically means that some data is withheld. From a few spot checks that I have done, it seems that withheld data are long tail queries rather than listed queries having incorrect (deflated) total, but I really did a very crude spot check to conclude this - it would need a more detailed analysis.
I have a site where "Displaying" is 4500 impressions and another site where "Displaying" is 100 (out of 150 impressions) so why not show all - it cannot be the space as the bigger site has "Display" 4500 out of "Impressions" 12000.
With regards to Top Pages tab, from what I can see, this shows impression of this page regardless of a query.
But if return to "Top queries" tab and then click on one of queries listed, you get more detailed breakdown, showing page with impressions for this query and showing position in search results for that query. If you have only one page ever that ranks for this query, then you could see if your numbers add up there.
Yes, that was the mystery: What's the criterion? It definitely isn't lack of space, considering all the 1's and 0's they have room for in the Pages tab :(
Click on a page in the Pages tab to get a breakdown of queries that lead to the page. The list doesn't add up to the total given in this tab-- but the numbers for any one query do match the numbers given in the Queries tab.
:: wait, stop, rewind ::
:: pause to kick self upside head ::
We did figure this out before. It's the https searches. They're included in the total-- "Impressions" on the Search side, everything on the "Pages" side-- but we're not shown what the actual searches were.
It then becomes interesting if there's a difference between the two percentages: Impressions total/displayed, and Clicks total/displayed. It tells you that people who log in to the search engine are either more interested or less interested in your content than the non-logged-in people. In my case, the figures are about 21% for impressions vs. 18% for clicks. Does this mean that people who use secure search find my material marginally more interesting (82/79) than people who don't? Or that people who have taken the trouble to log in are more committed to pursuing the search? Or simply that logged-in users habitually click on more results in a single page, so everyone gets a higher ctr?
Lucy, I do not think it is the lack of space either. Maybe they want to give us an idea of the keywords driving traffic, but not the whole info in case it would be somehow abused.
|It's the https searches. They're included in the total-- "Impressions" on the Search side, everything on the "Pages" side-- but we're not shown what the actual searches were. |
I am confused here. I thought that most of the searches are https and the ones that are not still pass search query to "ordinary analytics programs". I can certainly see this in Statcounter - most searches are https, but almost all that are http are passing search query. Few are http searches with encrypted searches and I am wondering whether this is logged in user but whose searches were not done on https.
I would think that if missing are https searches, that the number of queries and impressions reported would be much much smaller. Or have I misunderstood what you are saying and you do not mean all https? I get https whether or not I am logged in Google, but on some browsers of certain versions I am still http when not logged into Google.