the problem is that you don't know what they are doing during those five minutes, or whatever it is. They could have walked away from the computer and be off drinking a cup of tea, or left the browser open whilst they're writing an email in another window, or something.

Yea I frequently have ten or more tabs open in any given browser (sometimes a LOT more) which would make that a pretty noisy signal. I also don't think GA can always tell when a user has left entirely.

Now engagement - I'd be surprised if they weren't measuring that. But that would probably be more than just pageviews or time on site - it would be clicking on things, and moving through the site, and maybe filling out form fields. Scrolling. That would make more sense to me.

Time on site is a very dirty metric. It is hard to connect user satisfaction with time on site. There are too many reasons why the time can be short or long. The best thing about time on site as a metric is that almost no one is bothering to manipulate it.

Average time on site doesn't tell Google much about a site's value.

Consider: On our site, at least 20 per cent of the pageviews come from visitors who view 20+ pages in a session (and most of our pages are text-heavy). The site is obviously useful to visitors who like to read and want in-depth information, but that doesn't mean it's useful to everyone.

Also, if Google is doing its job right, visitors shouldn't always have to spend much time on a page to get the answer they're looking for. If Joe Searcher simply wants to know the local bus fare in Widgetville or the patron saint of hypochondriacs, there's no reason why he should be spending a lot of time on the site that provides a quick answer. For all Google knows, Joe may be in a hurry to catch a bus or to say a prayer to the appropriate saint.

When I look at my sites' stats, "time on site" and page both seem to be very important to G as far as the SERPS are concerned.

It is of course all relative - comparing time on site / page for a simple question / answer type of site with a more detailed informational site is of going to be pointless. But comparing those stats with similar sites / pages does reveal a lot about how useful the viewer rates the information, especially as far as detailed informational sites are concerned.

Yes there are always some viewers who randomlly leave pages open when not viewing them but the majority don't and their effect should be evenly spread. You also need to make the comparison where the number of page views is significant. That must be in the thousands to make the stats worthwhile.

It also seems to take months before the stats reflect the position in the SERPs, for my sites anyway.

But comparing those stats with similar sites / pages does reveal a lot about how useful the viewer rates the information, especially as far as detailed informational sites are concerned.

That raises the question of what "similar site" or "similar pages" might mean. This forum and Search Engine Land are both about search, and both have user-generated content (forum posts here, Disqus comments at Search Engine Land), but would Google consider them to be similar enough for a time-on-site comparison? I'd guess not, because user intent comes into play (people mostly go to SEO to read news, while they mostly come here for discussions), but then again, who knows?

And let's not forget that time on site and quality or usability aren't necessarily the same thing. If the searcher has to dig for the answer to his query once he's been sent to a page by Google, his time on site may be substantial, but he won't have a good user experience.

Imagine a website forcing you to watch a 30 second video before it finishes loading the page that has the actual content you want. Is that a good user experience? There are many factors that can impact time on site without having a strong correlation to user satisfaction.

Time on site is one of the most important signals as far as I'm concerned, only surpassed by brand searches. With just those 2 signals I can tell how valuable a site is when I compare it to all other sites for the same queries/industries (tons of calculations you can do with these metrics when you have them for the entire web like Google does).

I'm sure Google values time on site highly. (just my opinion of course)

I'm curious if you are sure that Google values time on site highly, do you think Google wants to see a high time on site (following the logic that users linger on good sites) or does Google want low time on site (following the logic that sites with good usability satisfy users in shorter time)?

It depends entirely on the query and niche, everything is relative. Searching for the weather would probably yield an entirely different behavior than searching for a flight to Miami.

One also has to factor what happens when the user comes back to the search results. Do they keep clicking on more results? If I searched for the weather and clicked on 2 results and stayed for 10 seconds per site, then clicked on another site and stayed for 5 minutes, if I'm Google I think that might be a pretty good indication of what site is the best for that query. Over the course of billions of queries Google can get pretty good at evaluating pages and domains.

It all comes down to providing the very best user experience for that particular query. If you do that people will share, people will come back, and every once in awhile you will get a link. It all adds up to better rankings.

Dog - That's a bit what I was trying to say. They have so many user metrics that it's almost like they're standing behind the user and viewing exactly what they're doing. How can chrome and analytics combined with such important things as time on site not be highly valued into the algo (I personally have yet to see it like your industry). I still find fluff content listed on the top because of links, in an industry where links are NOT given out freely but penalized harshly when paid for and caught.

@ddogg Sorry but I am still confused. You say Google values time on site highly. Then you say that time on site is relative and depends on query and niche when I asked if it should be short or long duration. This doesn't make sense to me since Google is a computer and needs to be coded. Time on site seems that it can be interpreted to mean anything depending on the circumstances so I question how it could used as benchmark especially since we can make the case that both shorter time on site is good (demonstrates good usability) and that longer time on site is good (demonstrates user engagement) at the same time.

You use the example of someone searching for the weather. If someone goes to Google and searches for "Google HQ weather" do they just want the current temperature (so therefore Google should reward fast time on site) or do they want the extended forecast (so now we should reward longer time on site because they are reading an in depth extended forecast). Let's say that the user actually types in "Google HQ extended weather forecast". We now know that the user is definitely looking for a more detailed answer and thus more time on site should be the clear answer. But what if the weather website has many ads & videos and bad usability which delivers a long time on site experience but it does not reflect a satisfied user. Time on site seems like a dirty metric. If I was Google I would probably look more at % of users that bounce back to the serps. That seems like a more accurate way to gauge user satisfaction.

Since you are sure that Google is using time on site, what testing did you do to confirm this?

My theoretical time on a site can be far longer than I spend looking at it. Typically if I see several potentially relevant results to a search I open them all in separate tabs. I might close the last one without bothering to read it but it still has the longest "time on site". Also I often don't bother closing tabs.

Quite simply absolutely not, I always have a minimum of three different browsers open with many tabs and for an example one of them is open 12-14 hours per day solely for WebmasterWorld even though I may not be reading or commenting.

So if you open something every day for 12-14 hours that's not indicative that it's a very good site? Whether you're commenting or not?

I understand there are many 'buts' and 'what if's' but isn't that why they build algorithims? 30 second ad, can't they measure that (youtube and other signals). Not being on the page, can they not measure that too (chrome)? Page too hard to find answer, most people are so lazy they just hit back when they see tons of text. Measured. They own the most popular internet browser - and most people are logged into gmail. I just don't see how they can't properly measure it.

Open tabs etc. are not relevant for time on site.

Most user analytics systems use the Page Visibility API as implemented in about all modern browers to check if a site is just opened in a tab (passive) or actively visited, read, scrolled, etc.

and most people are logged into gmail.

I have absolutely no statistics on this but I very much doubt that most business users are logged-in constantly to gmail, yes I do know that some companies have gmail business accounts however I would have thought that the majority of businesses use their own email?

I don't know, don't shoot me, however I am certainly not logged-in to gmail, I do not use it for anything.

the majority of businesses use their own email?

Majority still, probably yes, but an awful lot use business Gmail. Heck, even the University of Michigan switched over all its mail to Google mail a few years ago (despite still using the umich.edu TLD) and I bet a lot more have as well.

but an awful lot use business Gmail.

I am not doubting that in US/CA, why it's beyond me, but are there any precise figures? I've seen loads of figures bandied about however nothing I would bet my house on.

FWIW not even 10% on my email database use a non-business email address and I "reckon" that 25% of my global trade are subscribed to me.

When someone is emailing to me from a supposedly bona fide company considering huge purchases and they do not have their own company email address, well, for the sake of a few Dollars am I supposed to take it as anywhere near genuine? You want to buy $n millions of stuff yet cannot afford $10 for you own domain name and email?

Anyone doubting this...phishing, ok? I've been living with this since the 90s!

Honestly, it's an interesting question, how the heck anyone would get anywhere near the reality is another thing.

Gee...I've gone off-topic again, how did that happen?