by Hristo Hristov, Last Updated: 5 Nov 2018
I have long suspected dwell time of being a Google ranking factor, but there has been no official confirmation from Google.
While I have favored longer form content for SEO many years before the SEO industry caught up with its effectiveness, I was doing it for other reasons (higher keyword frequency, more keyword variants for long tail searches...).
My skepticism about dwell time vanished as soon as I dug my teeth into the USPTO patent database and found numerous Google patents on using duration of user visits as a ranking factor.
I found patents that ranked specific pages for specific search queries and also patents that calculated a query-independent domain score based on "click length". In other words, dwell time can affect the rankings of a whole website or just specific pages for specific queries.
obtaining a plurality of measurements of durations of user visits to resources included in a particular site; and determining a site quality score for the particular site based at least in part on the plurality of measurements, wherein determining the site quality score for the particular site comprises computing a statistical measure from the plurality of measurements.
Higher quality sites tend to keep the visitor attention for longer on average across all landing pages. Accordingly, Google takes visit durations for a given website and period of time and then computes a site quality score based on some form of average function of these durations (average dwell time).
The ranking score is site-wide and query-independent, which means it affects all pages and search queries of a website.
Obtaining Duration of User Visits
I like this patent because it gives a very clear definition of what duration of user visit means. What is dwell time?
a measurement of the duration of time that elapsed between the time that the user clicked on the search result and the time that the user navigated back to the search results web page
The patent also mentions alternative means of obtaining the data:
For example, the visit data can include data obtained from user devices, e.g., from application programs, e.g., web browsers, that run on user devices and display web pages, add-ons to the application programs, e.g., web browser toolbars, or both. The visit data can also include, for example, data obtained from network monitoring systems, e.g., routers, proxy servers, firewalls, or other hardware or software systems that are associated with user devices and monitor communications by the user devices over the network 112.
Discarding Some Measurements
If the "dwell time" of a visit is below a minimum threshold, it is discarded. According to the patent, visits that are too short can't be representative because "a user is unable to ascertain whether a resource contains high-quality or relevant content without viewing the resource for more than the threshold time".
In other words, quick bounces are discarded from the calculations. Those visitors who come, take a quick look and bounce will not hinder your domain quality score. Their dwell time is below the minimum threshold.
However, slower bounces are the type of visits that are problematic and that's when a visitor comes, gets interested and starts reading/scanning (passing the minimum time threshold), only to find your page does not serve his/her needs and goes away quickly afterwards.
Eliminating short visits from the calculations also accounts for people who quickly open multiple tabs from the SERPs.
The patent also talks about eliminating durations from unnatural user behavior (click bots).
Visit Durations are Capped
The normal searcher behavior is search, go to some results, come back to search, check out results...
Eventually users don't come back to the search results page which results in a very long visit duration. To counter it, Google adjusts measurements that exceed a predetermined maximum value to be equal to the maximum value.
To set proper minimum and maximum dwell time thresholds for a page, Google takes into account the type of page (media - video, audio, images, text) and length (text length, video length).
Longer pages and pages with media get higher minimum and maximum thresholds.
Furthermore, dwell time from a visit to a long page/page with a long video is a stronger signal than dwell time from a page with a few images and a little bit of text and accordingly these measurements may be weighted more into the whole mix of visit measurements of a site.
To sweeten the deal even more, Google may adjust the whole site score up if a certain percentage of the pages of a website are of a certain type (media/long texts). The reverse would also be true: sites with pages that are predominantly short will not only naturally get shorter visit durations but may also get their quality score further adjusted down(ouch).
Who wins the user engagement wins in ranking
The 2nd interesting patent I'd like to mention is: Determining resource quality based on resource competition.
In this patent Google describes a process where for each page it measures wins and losses in user sessions.
When a user performs a search, then selects 2 or more documents from the results, the page that got the longest click (dwell time) gets a win and the other selected pages get losses. Pages that were not selected (clicked) get neither a win nor a loss.
Pages that get more wins and fewer losses get promoted.
Pages that predominantly lose get demoted down.
You need to make sure that your pages have the longest user engagement for the important queries.
The process described in this patent kills 2 birds with one stone.
1. In results with lots of quality documents, it promotes the more engaging documents up.
2. When an overoptimized page appears for terms that it shouldn't rank, it generates many losses and gets demoted down.
Based on these wins/losses, Google may adjust the rankings of pages for specific queries or for whole domains.
There are more patents that deal with dwell-time but I found these two the most interesting.
Some take home points: