OpenSearch collects metrics for user applications across four dimensions: core metrics, traffic metrics, user analysis metrics, and query analysis metrics. This topic describes these metrics.
Core metrics
|
Metric name |
Definition |
Calculation formula |
Dependency |
Significance |
|
Search PV |
Page View (PV): The number of successful search requests. Pagination requests are also counted as search requests. |
Each successful search request is counted as 1. |
Measures the total number of successful searches on a given day. |
|
|
Search UV |
User View (UV): The number of unique users who have viewed a search results page. |
Each unique user is counted only once, regardless of the number of visits. |
This metric requires the user_id parameter to be set in search requests sent through the SDK or API. |
Measures the number of unique users who performed a search on a given day. |
|
Pagination rate |
The percentage of pagination requests relative to the total search PV. |
Pagination request PV / Search PV |
The ratio of pagination requests to total search page views measures the performance of search retrieval and sorting. |
|
|
No-result rate |
The percentage of searches that return zero items. |
Search PV with 0 items returned / Total search PV |
Measures the effectiveness of search retrieval and content coverage. |
|
|
Search PV per user |
The average number of successful searches per user. |
Search PV / Search UV |
This metric requires the user_id parameter to be set in search requests sent through the SDK or API. |
Indicates the average number of searches per user. A high value can indicate that users are interested in the search feature. It can also mean that users cannot find what they need with a small number of searches. |
|
Next-day retention rate |
The percentage of users who perform a search and then perform another search the next day. |
UV(t-1) ∩ UV(t) / UV(t-1) |
This metric requires the user_id parameter to be set in search requests sent through the SDK or API. |
Reflects user engagement with the search feature. |
Traffic metrics
|
Metric name |
Definition |
Calculation formula |
Dependency |
Significance |
|
Search PV |
Page View (PV): The number of successful search requests. Pagination requests are also counted as search requests. |
Each successful search request is counted as 1. |
Measures the total number of successful searches on a given day. |
|
|
Search UV |
User View (UV): The number of unique users who have viewed a search results page. |
Each unique user is counted only once, regardless of the number of visits. |
When you use an SDK or API to search, set the |
Measures the number of unique users who performed a search on a given day. |
|
Search PV per user |
The average number of successful searches per user. |
Search PV / Search UV |
This metric requires the user_id parameter to be set in search requests sent through the SDK or API. |
Indicates the average number of searches per user. A high value can indicate that users are interested in the search feature. It can also mean that users cannot find what they need with a small number of searches. |
|
Number of queries |
The number of search queries, deduplicated by user. |
The total number of queries from users, deduplicated per user. For example, if two different users search for the same query, the count is 2. |
This metric requires the user_id parameter to be set in search requests sent through the SDK or API. The query content is determined as follows: If the raw_query parameter is set, its value is used. Otherwise, the value from the default index is used as the query. If a default index is not configured, the query is empty. |
Represents the total number of user search queries and reduces the impact of factors such as pagination and repeated searches. |
|
Number of queries per user |
The average number of search queries per user, deduplicated per user. |
Number of queries / Search UV |
This metric requires the user_id parameter to be set in search requests sent through the SDK or API. |
Represents the average number of search queries per user and reduces the impact of factors such as pagination and repeated searches. |
|
Number of unique queries |
The number of unique search queries from all users. |
The total number of search queries, deduplicated across all users. If multiple users search for the same query, the count is 1. |
The query content is determined as follows: If the raw_query parameter is set, its value is used. Otherwise, the value from the default index is used as the query. If a default index is not configured, the query is empty. |
Reflects the diversity of search queries. |
|
Pagination rate |
The percentage of pagination requests relative to the total search PV. |
Pagination request PV / Search PV |
The ratio of pagination requests to total search page views measures the effectiveness of search retrieval and sorting. |
|
|
No-result rate |
The percentage of searches that return zero items. |
Search PV with 0 items returned / Total search PV |
Measures the effectiveness of search retrieval and content coverage. |
|
|
Low-result rate |
The percentage of searches that return five or fewer items. |
Search PV with ≤ 5 items returned / Total search PV |
Measures the effectiveness of search retrieval and content coverage. |
|
|
Number of request errors |
The number of search requests that fail due to issues such as throttling. |
Each failed request is counted as 1. |
Indicates whether the application needs to be scaled out or adjusted. |
|
|
Request error rate |
The ratio of search requests that fail due to issues such as throttling. |
Number of request errors / (Search PV + Number of request errors) |
Indicates whether the application needs to be scaled out or adjusted. |
User analysis metrics
|
Metric name |
Definition |
Calculation formula |
Dependency |
Significance |
|
Next-day retention rate |
The percentage of users who perform a search and then perform another search the next day. |
UV(t-1) ∩ UV(t) / UV(t-1) |
Depends on setting the user_id parameter in search requests sent through the SDK or API. |
Reflects user engagement with the search feature. |
Query analysis metrics
|
Category |
Metric name |
Definition |
Calculation formula |
Dependency |
Significance |
|
High-Frequency Words |
Query |
The 500 most frequent search queries. |
Ranks queries by search frequency and returns the top 500. |
The query content is determined as follows: If the raw_query parameter is set, its value is used. Otherwise, the value from the default index is used as the query. If a default index is not configured, the query is empty. |
Identifies the queries that users are most interested in. |
|
Search PV |
The search PV for the query. |
Each search is counted as 1. |
When you use an SDK or API to search, you can set the `raw_query` parameter in the search request or omit it. |
Top queries |
|
|
Average exposed items |
The average number of items exposed for the query. |
Total items returned for the query / Search PV |
This depends on whether the `raw_query` parameter is set in the search request when using an SDK or API. |
Indicates the number of results shown for the most popular queries. This helps measure the effectiveness of search retrieval and content coverage. |
|
|
Low-result queries |
Query |
The most frequent queries that return five or fewer items. |
Ranks queries that return five or fewer items by frequency and returns the top 100. |
The query content is determined as follows: If the raw_query parameter is set, its value is used. Otherwise, the value from the default index is used as the query. If a default index is not configured, the query is empty. |
Identifies frequent queries that return few results, which helps measure the effectiveness of search retrieval and content coverage. |
|
Search PV |
The search PV for the query. |
Each search is counted as 1. |
When you search using an SDK or API, the `raw_query` parameter in the search request is optional. |
Indicates the search volume for low-result queries, which helps measure the severity of the issue. |
|
|
Average exposed items |
The average number of items exposed for the query. |
Total items returned for the query / Search PV |
When you use an SDK or API to search, you can either set the raw_query parameter in the search request or leave it unset. |
The exposure of search results for the most popular queries measures the effectiveness of search recall and the richness of the content. |
|
|
No-result queries |
Query |
Queries that return no items |
Ranks queries that return zero items by frequency and returns the top 100. |
The query content is determined as follows: If the raw_query parameter is set, its value is used. Otherwise, the value from the default index is used as the query. If a default index is not configured, the query is empty. |
Identifies frequent queries that return no results, which helps measure the effectiveness of search retrieval and content coverage. |
|
Search PV |
The search PV for the query. |
Each search is counted as 1. |
When you perform a search using an SDK or API, you may not need to set the `raw_query` parameter in the search request. |
Indicates the search volume for no-result queries, which helps measure the severity of the issue. |
|
|
Fastest-rising queries |
Query |
Queries with the highest growth rate in search PV compared to the previous day. Queries are sorted in descending order of their growth rate. A minimum search PV threshold of max(10, 0.1% of total search PV) is applied to the previous day's data. |
Ranks queries by the increase in search frequency compared to the previous day and returns the top 100. |
The query content is determined as follows: If the raw_query parameter is set, its value is used. Otherwise, the value from the default index is used as the query. If a default index is not configured, the query is empty. |
Identifies recent hot spots and emerging queries. |
|
Previous day's search PV |
The search PV for the query on the previous day. |
Each search is counted as 1. |
When you use an SDK or API to search, you may need to set the `raw_query` parameter in the search request. |
The number of searches for recent hot spots and emerging queries on the previous day. |
|
|
Current day's search PV |
The search PV for the query. |
Each search is counted as 1. |
When you use an SDK or API to send a search request, the `raw_query` parameter is optional. |
The number of searches for recent hot spots and emerging queries on the current day. |
|
|
Growth rate |
The growth rate of the query on the current day. |
(Current day's search PV - Previous day's search PV) / Previous day's search PV |
The growth rate of a hot spot query on the current day. |
||
|
Average exposed items |
The average number of items exposed for the query. |
Total items returned for the query / Search PV |
When you search using an SDK or API, you do not always need to set the raw_query parameter in the search request. |
The exposure of search results for recent hot spots and emerging queries. This measures the effectiveness of search retrieval and content coverage. |
|
|
Fastest-declining queries |
Query |
Queries with the highest decline rate in search PV compared to the previous day. Queries are sorted in descending order of their decline rate. A minimum search PV threshold of max(10, 0.1% of total search PV) is applied to the current day's data. |
Ranks queries by the decrease in search frequency compared to the previous day and returns the top 100. |
The query content is determined as follows: If the raw_query parameter is set, its value is used. Otherwise, the value from the default index is used as the query. If a default index is not configured, the query is empty. |
Identifies queries with a rapidly declining search frequency. |
|
Previous day's search PV |
The search PV for the query on the previous day. |
Each search is counted as 1. |
This depends on whether the `raw_query` parameter is set in the search request when using an SDK or an API. |
The number of searches on the previous day for queries with a rapidly declining search frequency. |
|
|
Current day's search PV |
The search PV for the query. |
Each search is counted as 1. |
Depends on the raw_query parameter being used to identify the query string. |
The number of searches on the current day for queries with a rapidly declining search frequency. |
|
|
Decline rate |
The decline rate of the query on the current day. |
(Previous day's search PV - Current day's search PV) / Previous day's search PV |
|||
|
Average exposed items |
The average number of items exposed for the query. |
Total items returned for the query / Search PV |
When you use an SDK or an API to search, you can set the `raw_query` parameter in the search request or omit it. |
The exposure of search results for queries with a rapidly declining search frequency. This measures the effectiveness of search retrieval and content coverage. |
Instructions
-
By default, the single-day data view shows the metrics for the most recent day. You can use the date control to select a different date.
-
The data comparison view displays the changes in each metric over the last seven days as a line chart by default. You can use the date control to select a different time span. You can click the icon for a metric in the chart to show or hide it.
-
In the navigation bar, click Statistical Reports > Operations Reports. Then, select an application, scenario, metric type, and time range to view the statistical charts.