Ranking of Search Results Based on Customer Intent
The disclosure describes technology for ranking search results based on customer intent. A set of matching product entries matching the one or more search keywords is determined from among product entries stored in a product database reflecting products purchasable via one or more online marketplaces. For each product entry, scores are computed using different combinations of coefficients and an amount of revenue-per-visit, a number of orders, and a ratio of page visits to product views associated with the product entry. The set of matching product entries are ranked based on a rank associated with each matching product entry in the set of matching product entries. The rank is computed based on a comparison between a plurality of scores associated with the matching product entry.
1 . A method, executable by a computing system including one or more processors and one or more memories, for ranking and sorting search results generated by a search engine, the method comprising:
receiving, using the one or more processors, a search query including one or more search keywords;
responsive to receiving the search query, determining, using the one or more processors, from among product entries stored in a product database reflecting products purchasable via one or more online marketplaces, a set of matching product entries matching the one or more search keywords;
computing, using the one or more processors, a plurality of scores for each product entry using different combinations of coefficients and an amount of revenue-per-visit, a number of orders, and a ratio of page visits to product views associated with the product entry;
scaling, using the one or more processors, one or more of the amount of revenue-per-visit, the number of orders, and the ratio of page visits to product views associated with each product entry relative to a maximum amount of revenue-per-visit, a maximum number of orders, and maximum ratio of page visits to product views available in the set of product entries, respectively;
responsive to determining the set of matching product entries matching the one or more search keywords, ranking, using the one or more processors, the set of matching product entries based on a rank associated with each matching product entry in the set of matching product entries, the rank computed based on a comparison between the plurality of scores associated with the matching product entry; and
transmitting, using the one or more processors, the ranked set of matching product entries for presentation.
2 . The method of claim 1 , wherein computing the scores for each product entry further includes, for at least one score, comparing a ratio of orders and a number of cart additions to a predetermined threshold, and, if the ratio of the orders and the number of cart additions satisfies the predetermined threshold, calculating a score value for the at least one score based on the amount of revenue-per-visit, the number of orders, and the ratio of page visits to product views associated with the product entry.
3 . The method of claim 2 , wherein the predetermined threshold is one or more standard deviations of a scaled value of the ratio of the orders and the number of cart additions or a scaled value of the ratio of page visits to product views.
4 . The method of claim 1 , wherein computing the scores includes calculating a score value for two or more different cases using the equation:
{
(
1
1
+
-
RPV
)
+
(
C
1
*
Log
(
Orders
)
)
+
(
C
2
*
Log
(
Visits
Product
Views
)
)
}
,
the different cases being based on unique combinations of C2 and C2.
5 . The method of claim 1 , further comprising:
computing the rank of each product entry based on the comparison between the plurality of scores associated with the product entry.
6 . A method, executable by a computing system including one or more processors and one or more memories, for ranking and sorting search results, comprising:
receiving, using the one or more processors, a search query including one or more search keywords;
responsive to receiving the search query, determining, using the one or more processors, from among product entries stored in a product database reflecting products purchasable via one or more online marketplaces, a set of matching product entries matching one or more keywords;
computing, using the one or more processors, a rank associated with each matching product entry in the set of matching product entries using a different combination of coefficients and an amount of revenue-per-visit, a number of orders, and a ratio of page visits to product views associated with the matching product entry; and
ranking, using the one or more processors, the set of matching product entries based on a rank associated with each matching product entry in the set of matching product entries.
7 . The method of claim 6 , wherein computing the rank further includes:
computing, using the one or more processors, a plurality of scores for each matching product entry in the set of matching product entries using different combinations of coefficients and the amount of revenue-per-visit, the number of orders, and the ratio of page visits to product views associated with the matching product entry, wherein the rank for each matching product entry is computed based on a comparison between the plurality of scores associated with the matching product entry.
8 . The method of claim 7 , wherein at least one of the plurality of scores for at least one product is computed comparing a ratio of orders and a number of cart additions to a predetermined threshold, and, if the ratio of the orders and the number of cart additions satisfies the predetermined threshold, calculating a score value for the at least one score based on the amount of revenue-per-visit, the number of orders, and the ratio of page visits to product views associated with the product entry.
9 . The method of claim 8 , wherein the predetermined threshold is one or more standard deviations of a scaled value of the ratio of orders and the number of cart additions or a scaled value of the ratio of page visits to product views.
10 . The method of claim 7 , wherein computing the plurality of scores includes calculating a score value for each of two or more different cases using the equation:
{
(
1
1
+
-
RPV
)
+
(
C
1
*
Log
(
Orders
)
)
+
(
C
2
*
Log
(
Visits
Product
Views
)
)
}
,
the different cases being based on unique combinations of C1 and C2.
11 . The method of claim 6 , further comprising:
computing, using the one or more processors, the rank associated with each matching product entry in the set of matching product entries based on the comparison between a plurality of scores associated with the product entry.
12 . The method of claim 6 , wherein one or more of the amount of revenue-per-visit, the number of orders, and the ratio of page visits to product views associated with each product entry are scaled relative to a maximum amount of revenue-per-visit, maximum number of orders, and a maximum ratio of page visits to product views available in the set of product entries.
13 . A system comprising:
one or more processors;
one or more memories storing instructions that, when executed by the one or more processors, cause the system to perform operations including:
receiving a search query including one or more search keywords;
responsive to receiving the search query, determining from among product entries stored in a product database reflecting products purchasable via one or more online marketplaces, a set of matching product entries matching one or more keywords;
computing a rank associated with each matching product entry in the set of matching product entries using a different combination of coefficients and an amount of revenue-per-visit, a number of orders, and a ratio of page visits to product views associated with the matching product entry; and
ranking the set of matching product entries based on a rank associated with each matching product entry in the set of matching product entries.
14 . The system of claim 13 , wherein computing the rank further includes:
computing a plurality of scores for each matching product entry in the set of matching product entries using different combinations of coefficients and the amount of revenue-per-visit, the number of orders, and the ratio of page visits to product views associated with the matching product entry, wherein the rank for each matching product entry is computed based on a comparison between the plurality of scores associated with the matching product entry.
15 . The system of claim 14 , wherein at least one of the plurality of scores for at least one product is computed comparing a ratio of orders and a number of cart additions to a predetermined threshold, and, if the ratio of the orders and the number of cart additions satisfies the predetermined threshold, calculating a score value for the at least one score based on the amount of revenue-per-visit, the number of orders, and the ratio of page visits to product views associated with the product entry.
16 . The system of claim 15 , wherein the predetermined threshold is one or more standard deviations of a scaled value of the ratio of orders and the number of cart additions or a scaled value of the ratio of page visits to product views.
17 . The system of claim 14 , wherein computing the plurality of scores includes calculating a score value for each of two or more different cases using the equation:
{
(
1
1
+
-
RPV
)
+
(
C
1
*
Log
(
Orders
)
)
+
(
C
2
*
Log
(
Visits
Product
Views
)
)
}
,
the different cases being based on unique combinations of C1 and C2.
18 . The system of claim 13 , further comprising:
computing the rank associated with each matching product entry in the set of matching product entries based on the comparison between a plurality of scores associated with the product entry.
19 . The system of claim 13 , wherein one or more of the amount of revenue-per-visit, the number of orders, and the ratio of page visits to product views associated with each product entry are scaled relative to a maximum amount of revenue-per-visit, maximum number of orders, and a maximum ratio of page visits to product views available in the set of product entries.