When a user makes use of a search engine, typing a query, more or less complex, uses the terms that seem to him to be the most relevant and functional to have access to the desired results.
Sometimes the answer is included in the first results of the SERP returned, maybe later a short period of Pogo Sticking; other times, however, either because the words used have multiple meanings (ambiguity of the query), or because the question is unusual (Google continuously receives new queries, every day 15% of total requests), or because of the genericity of the request, the user will not be satisfied by the first solutions presented.
At this point, theoretically, the dissatisfied and willing user will continue to query the search engine, refining the query from time to time until he finds a rewarding result.
Google, in recent years, has been very interested in user-refinding sessions, and this can be seen from the various patents filed on the subject.
As we know, patents are useful to understand which are Google’s directions of study – and investment – but they do not necessarily correspond to what is used; they are therefore an indication of “how Google might reason”, but not how it actually reasons.
The first algorithms on the subject date back to 2007, specifically:
- The demotion of repetitive search results
- Session-based dynamic search snippets
The demotion of repetitive search results
Function: During navigation with queries related to the same topic and that fall for the search engine in the same session Google downgrade (ranking) the results (both sites and pages) that the user had not found interesting for the first query.
The parameters to identify the same session: queries concerning the same topic made within a certain “reasonable” time (5 minutes not a month).
Logics: It is plausible that the downgrade only affects the resources that the user had consulted compared to all those shown, but the patent could also intervene independently of the clicks on the resource in the SERP. The downgrade will have to respect a relevance threshold, so some results will be lowered in ranking only if there are valid alternatives.
Session-based dynamic search snippets
The patent was granted for the first time in the same period as the “Demotion of repetitive search results” (it is based on the same session concept) but then it was revisited in 2011.
Function: Customisation of Snippets in SERPs; in this way a page (which was not consulted) presented for the initial query to the user, can be re-proposed to the browser in the same Search Session (if relevant in the session-refinding) with a modification of the Snippet (highlighting the relevant part of the page) just to encourage the click on the resource.
Last updates: Contextual search term evaluation
A more recent patent, again focusing on the same issue, was granted in 2014.
Function: the concept behind this new patent is to give more weight to the terms used in the queries of the first search requests and less to the terms added in the following steps (obviously related to the same session).
Google is constantly experimenting with multiple algorithms on the same topic, testing different methodologies, often even conflicting, and then choosing the best solution.
Monitoring and analyzing Search Sessions will provide the search engine with data about them:
- Query suggestions
- Taxonomic relationships between generic and specific queries
- Query Expansions
- Co-occurrence between search terms (if many users perform similar searches this will indicate that there is a strong relationship)
- “Related searches”
- The relevance and accuracy of advertising will improve
- The data collected will be used in the calibration of the algorithms (Hummingbird).
The content of a page is and will increasingly be a relevant factor (Snippet customization).
People are Strange
Michael Martinez: I actually find myself using Bing more often when I need to discover new information
By calibrating the operation of the search engine more and more on the majority of users, is it possible that other search engines can help us to intercept the “minor” user niches?