User search behavior value: Analysis of the value of vertical search engine data

Vertical search engines are search windows provided on websites/APPs, allowing users to directly access target content by searching for keywords . A brief explanation has been given in the previous article, and its three application stages have been explained from the perspective of readability.

This article details the data value that users bring to companies in the process of applying vertical search engines. It can even be said that the value of user search behavior is ignored by many companies.

“Hot searches” are not exclusive to Weibo, and companies should “assess the situation” accordingly

Let’s first talk about the value of “hot search terms” (abbreviated as “hot searches”).

Everyone is familiar with Weibo's hot searches, and anyone who uses Weibo will probably click on the current "hot search terms" to view related content. This is the guiding charm of popular search results.

For every website, whether it is e-commerce , information or video, it is normal for users to directly reach their needs through search. The words that are searched most frequently by users within a period of time constitute the “hot search words”. For e-commerce websites, "hot searches" are products; for information websites, "hot searches" are media content; for video websites, "hot searches" are videos. Although the content attributes behind "hot searches" vary from website to website, they can all reflect the hot demands of website users.

At this time, the company can arrange the content according to the current "hot searches", increase the quantity of a certain product in advance to avoid "out of stock", increase the frequency of content exposure, or change the display location of the content to improve user search efficiency.

At present, "hot search" is one of the effective ways to solve the cold start phase of personalized recommendations, which will not be elaborated in this article.

“Fastest rising/falling” search terms reflect underlying trends

The "fastest rising/falling" search terms are the terms whose user search times have risen/fallen the most over a period of time. Compared with "hot search" words, "fastest rising/falling" search words can better reflect the trend and direction of search, and can also better guide companies to pre-set operational actions.

When faced with the "fastest rising" search terms, companies can quickly make operational arrangements for the related surrounding content to increase exposure and improve clicks; while for the "fastest declining" search terms, companies can reduce their exposure or even cut the Gordian knot at a certain time. For information companies, the statistics of these two indicators are particularly important, because content involving "timeliness" is not a trivial matter.

"Search results without results" should also be taken seriously, and "make up for the loss" to retain users

If "hot searches" are search data that have operational value to enterprises, then the role of "search results without results" (abbreviated as "resultless words") can be comparable to them.

Think about the motivation behind a user’s search. It must be to find the target object from the search results and the user believes that the website may have such content. If there are no corresponding search results after searching, the user experience can be imagined. Unless it is a monopoly enterprise, the speed at which users "switch to" competitors may be the time it takes to search for competitors on Baidu .

Therefore, companies should pay great attention to the "fruitless words" after users search, because it indirectly reflects the users' unmet needs. At this point, the company can check for deficiencies and fill in the gaps in the website content to ensure that users feel satisfied when they search again. "It's never too late to mend."

Is the existence of “low click-through rate ” search results useless?

"Low click-through rate" search results are similar to "restaurants with poor business", which are those restaurants that are not clicked in the search results presented to users. The reasons for the low click-through rate of search results may be:

1. The quality is really not high, and it can’t solve the user’s needs at a glance;
2. The position is really bad, just like in Baidu search results, we may only look at the first five or ten results.

If it is reason 1, the company needs to adjust the search algorithm and move the display position of search results with low click-through rates to the back, so as to improve user satisfaction with the search results. Taking a reading website as an example, after a user searches for "rebirth", among the top ten search results, a book called "Rebirth of ×××" has a very low click-through rate. In this case, the position of this book can be adjusted, or even other operations for this book can be abandoned, because there is no traffic value, let alone paid conversion .

Do users prefer "sales ranking" or "price ranking"?

This data is apparently more suitable for e-commerce companies because the products on e-commerce websites have both price and sales volume. If it is an information website, it may be "sorted by time"; if it is a video website, it may be "sorted by number of views" or "sorted by time". But they all belong to user-filtered behavioral data.

Here we take e-commerce websites as an example.

Users filter by "sales volume ranking", which means that users are more concerned about the sales volume of products and consider the public's choice more when purchasing products. Users filter by "price sort", which means that users are more sensitive to product prices. Enterprises can analyze user screening behavior data and reasonably display the default sorting of search results, thereby reducing the number of user operations and improving user experience.

Summarize

The data value of user search behavior is mainly developed from three aspects:

1. Hot search terms;
2. Fastest rise/fall;
3. Search for no results for the term;
4. Low click-through rate search results;
5. User screening behavior.

However, different industries may have different analysis dimensions for user search data. Due to limited space in this article, we will not share in depth.