How easy is it to pick up a mobile phone?
Imhonet opened the Mobile Phones advisory service, which introduced two significant innovations: collaborative filtering and a complex equalizer filter .
After examining the filter selection of mobile phones from other sites, we realized that there are still many consumer requests that are not solved in the best way. With a lack of incoming information from the buyer, he is at best offered to buy the most popular . At the same time, the popularity of goods depends not on its qualities, but on marketing, branding, etc.
This problem is partially solved by the direct collection of assessments, but even here we stumble upon the fact that it is not clear how to separate different people who evaluate one product. For example, I do not care about the assessment of the phone by a person who has completely different needs than me, and vice versa, the opinion of people who are similar to me in needs (like-minded people) is very important.
')
In the mobile phone section, they made a system that allows groups of people to easily unite and interact in the process of the usual process of evaluating and selecting phones.
How it works?
1. People come to the section, register, evaluate familiar phones.
2. The same people are looking for something for themselves or friends.
Subtotal: we get a primary idea of ​​what a person comes from when evaluating a phone.
3. Comes another person who is looking for a phone. He specifies his preferences in the filter.
4. The system selects for it people with very similar search queries (filter settings) and looks at which of the phones corresponding to the formal criteria they rated highly.
The result: a person sees the recommended phones from people with similar needs, which are selected from the whole mass of phones that meet formal criteria.
In order to gain more profiles of different people (and not just industry experts), we tried to translate the main technical parameters into human language and add the “desirable” parameter. So it is easier to express your attitude when you generally understand why some function is needed, but you are not sure how important it is to you.
As a result, we were able to focus the issue for a variety of requests, for which excessively many products had previously been highlighted.
Example:
A person wants to get a good touch phone up to 5 thousand rubles, while it is desirable that it could take good pictures and show it to friends, and GPS will not hurt. Such phones for sale at least 10 pieces. This complicates the choice, especially given that the original request was not rigid.
The recommended list selects 1 phone, which was highly appreciated by people with the same needs, while it meets all the desired conditions, except GPS (there were no such models).
As a result, the request is satisfied, and the time to choose a proven model is reduced many times.
Development plans
Now the recommendation mechanism in products in version 0.9.
We are at the stage of statistics accumulation and in the future we plan to make possible a separate manual configuration of our profile and sort the recommended list according to a forecast estimate. So we will increase the focus of recommendations on the initial needs of a particular group of users and move further and further away from generalized opinions.
After examining the filter selection of mobile phones from other sites, we realized that there are still many consumer requests that are not solved in the best way. With a lack of incoming information from the buyer, he is at best offered to buy the most popular . At the same time, the popularity of goods depends not on its qualities, but on marketing, branding, etc.
This problem is partially solved by the direct collection of assessments, but even here we stumble upon the fact that it is not clear how to separate different people who evaluate one product. For example, I do not care about the assessment of the phone by a person who has completely different needs than me, and vice versa, the opinion of people who are similar to me in needs (like-minded people) is very important.
')
In the mobile phone section, they made a system that allows groups of people to easily unite and interact in the process of the usual process of evaluating and selecting phones.
How it works?
1. People come to the section, register, evaluate familiar phones.
2. The same people are looking for something for themselves or friends.
Subtotal: we get a primary idea of ​​what a person comes from when evaluating a phone.
3. Comes another person who is looking for a phone. He specifies his preferences in the filter.
4. The system selects for it people with very similar search queries (filter settings) and looks at which of the phones corresponding to the formal criteria they rated highly.
The result: a person sees the recommended phones from people with similar needs, which are selected from the whole mass of phones that meet formal criteria.
In order to gain more profiles of different people (and not just industry experts), we tried to translate the main technical parameters into human language and add the “desirable” parameter. So it is easier to express your attitude when you generally understand why some function is needed, but you are not sure how important it is to you.
As a result, we were able to focus the issue for a variety of requests, for which excessively many products had previously been highlighted.
Example:
A person wants to get a good touch phone up to 5 thousand rubles, while it is desirable that it could take good pictures and show it to friends, and GPS will not hurt. Such phones for sale at least 10 pieces. This complicates the choice, especially given that the original request was not rigid.
The recommended list selects 1 phone, which was highly appreciated by people with the same needs, while it meets all the desired conditions, except GPS (there were no such models).
As a result, the request is satisfied, and the time to choose a proven model is reduced many times.
Development plans
Now the recommendation mechanism in products in version 0.9.
We are at the stage of statistics accumulation and in the future we plan to make possible a separate manual configuration of our profile and sort the recommended list according to a forecast estimate. So we will increase the focus of recommendations on the initial needs of a particular group of users and move further and further away from generalized opinions.
Source: https://habr.com/ru/post/97764/
All Articles