You're viewing version 3.5 of the OpenSearch documentation. This version is no longer maintained. For the latest version, see the current documentation. For information about OpenSearch version maintenance, see Release Schedule and Maintenance Policy.
Experiments
An experiment is a controlled test designed to assess the effectiveness, relevance, or performance of a search engine or its algorithms. These experiments are typically conducted in order to evaluate how well a search system delivers useful results for specific queries.
Search Relevance Workbench offers multiple types of experiments. For more information, see Available search result quality experiments.
Managing experiments
You can retrieve or delete experiments using the following APIs.
View an experiment
You can retrieve an experiment using the experiment ID.
Endpoint
GET _plugins/_search_relevance/experiments/<experiment_id>
Path parameters
The following table lists the available path parameters.
| Parameter | Data type | Description |
|---|---|---|
experiment_id | String | The ID of the experiment to retrieve. |
Example request
GET _plugins/_search_relevance/experiments/b54f791a-3b02-49cb-a06c-46ab650b2ade
Example response
Response
{
"took": 1,
"timed_out": false,
"_shards": {
"total": 1,
"successful": 1,
"skipped": 0,
"failed": 0
},
"hits": {
"total": {
"value": 1,
"relation": "eq"
},
"max_score": 1,
"hits": [
{
"_index": ".plugins-search-relevance-experiment",
"_id": "47cc3861-c37b-43cc-99c4-c1e7c90b4674",
"_score": 1,
"_source": {
"id": "47cc3861-c37b-43cc-99c4-c1e7c90b4674",
"timestamp": "2026-01-28T18:16:44.548Z",
"type": "PAIRWISE_COMPARISON",
"status": "COMPLETED",
"querySetId": "f4c35381-407c-45c7-89ec-094b8a4cd5b1",
"searchConfigurationList": [
"85f31a87-2833-4d4a-89d2-cc83248f410e",
"050b8c98-ba63-4c75-89cb-75f379b0d66e"
],
"judgmentList": [],
"size": 10,
"isScheduled": false,
"scheduledExperimentJobId": null,
"results": [
{
"snapshots": [
{
"searchConfigurationId": "85f31a87-2833-4d4a-89d2-cc83248f410e",
"docIds": [
"B07K1H1G3M",
"B07THVCJK3",
"B07FPP6TB5",
"B07Q7Z9DJ3",
"B07PYPCX21",
"B07WLRKCNW",
"B01HE1IVNA",
"B07ZKDVHFB",
"B071D41YC3",
"B07B6L2QCF"
]
},
{
"searchConfigurationId": "050b8c98-ba63-4c75-89cb-75f379b0d66e",
"docIds": [
"B07ZKDVHFB",
"B07FPP6TB5",
"B07THVCJK3",
"B071D41YC3",
"B07JD5RT4D",
"B079QHML21",
"B07ZZVX1F2",
"B07YNLBS7R",
"B01N1SSOUC",
"B07WLRKCNW"
]
}
],
"metrics": [
{
"metric": "jaccard",
"value": 0.33
},
{
"metric": "rbo50",
"value": 0.14
},
{
"metric": "rbo90",
"value": 0.32
},
{
"metric": "frequencyWeighted",
"value": 0.5
}
],
"query_text": "tv"
},
{
"snapshots": [
{
"searchConfigurationId": "85f31a87-2833-4d4a-89d2-cc83248f410e",
"docIds": [
"B07THVCJK3",
"B07FPP6TB5",
"B07PP4882Q",
"B07K1H1G3M",
"B091KB3W63",
"B07JN28KP3",
"B07Q7SGS6Z",
"B07SJZ9X6J",
"B07VNG9ZLM",
"B00DTOAWZ2"
]
},
{
"searchConfigurationId": "050b8c98-ba63-4c75-89cb-75f379b0d66e",
"docIds": [
"B07THVCJK3",
"B091KB3W63",
"B07PP4882Q",
"B07RFFJ7YL",
"B08HGQ7H8F",
"B083BNQYBP",
"B07ZKDVHFB",
"B07N1CMGQQ",
"B07FPP6TB5",
"B071D41YC3"
]
}
],
"metrics": [
{
"metric": "jaccard",
"value": 0.25
},
{
"metric": "rbo50",
"value": 0.77
},
{
"metric": "rbo90",
"value": 0.58
},
{
"metric": "frequencyWeighted",
"value": 0.4
}
],
"query_text": "led tv"
}
]
}
}
]
}
}
Delete an experiment
You can delete an experiment using the experiment ID.
Endpoint
DELETE _plugins/_search_relevance/experiments/<experiment_id>
Example request
DELETE _plugins/_search_relevance/experiments/47cc3861-c37b-43cc-99c4
Example response
{
"_index": ".plugins-search-relevance-experiment",
"_id": "47cc3861-c37b-43cc-99c4-c1e7c90b4674",
"_version": 3,
"result": "deleted",
"forced_refresh": true,
"_shards": {
"total": 1,
"successful": 1,
"failed": 0
},
"_seq_no": 6,
"_primary_term": 1
}
Search for an experiment
You can search for available experiments using query domain-specific language (DSL). By default, the results data is not returned in the response. To include the results data, specify the _source field in the query.
Endpoints
GET _plugins/_search_relevance/experiments/_search
POST _plugins/_search_relevance/experiments/_search
Example request
Search for experiments that use a specific query set in order to measure search relevance performance:
GET _plugins/_search_relevance/experiments/_search
{
"query": {
"term": { "querySetId": "f4c35381-407c-45c7-89ec-094b8a4cd5b1" }
}
}
Example response
{
"took": 2,
"timed_out": false,
"_shards": {
"total": 1,
"successful": 1,
"skipped": 0,
"failed": 0
},
"hits": {
"total": {
"value": 3,
"relation": "eq"
},
"max_score": 1,
"hits": [
{
"_index": ".plugins-search-relevance-experiment",
"_id": "47cc3861-c37b-43cc-99c4-c1e7c90b4674",
"_score": 1,
"_source": {
"judgmentList": [],
"size": 10,
"scheduledExperimentJobId": null,
"searchConfigurationList": [
"85f31a87-2833-4d4a-89d2-cc83248f410e",
"050b8c98-ba63-4c75-89cb-75f379b0d66e"
],
"isScheduled": false,
"id": "47cc3861-c37b-43cc-99c4-c1e7c90b4674",
"type": "PAIRWISE_COMPARISON",
"timestamp": "2026-01-28T18:16:44.548Z",
"status": "COMPLETED",
"querySetId": "f4c35381-407c-45c7-89ec-094b8a4cd5b1"
}
},
{
"_index": ".plugins-search-relevance-experiment",
"_id": "90e4a31b-3b4a-4b2e-b492-0399358961cc",
"_score": 1,
"_source": {
"judgmentList": [
"505d00cf-2fce-422b-bb97-2e3a95ce9446"
],
"size": 8,
"scheduledExperimentJobId": null,
"searchConfigurationList": [
"85f31a87-2833-4d4a-89d2-cc83248f410e"
],
"isScheduled": false,
"id": "90e4a31b-3b4a-4b2e-b492-0399358961cc",
"type": "POINTWISE_EVALUATION",
"timestamp": "2026-01-28T18:16:45.696Z",
"status": "COMPLETED",
"querySetId": "f4c35381-407c-45c7-89ec-094b8a4cd5b1"
}
},
{
"_index": ".plugins-search-relevance-experiment",
"_id": "71608f58-4827-4cd3-b6a6-61a527975a23",
"_score": 1,
"_source": {
"judgmentList": [
"505d00cf-2fce-422b-bb97-2e3a95ce9446"
],
"size": 10,
"scheduledExperimentJobId": null,
"searchConfigurationList": [
"97f5b450-9571-469d-9dbb-347f94d164ba"
],
"isScheduled": false,
"id": "71608f58-4827-4cd3-b6a6-61a527975a23",
"type": "HYBRID_OPTIMIZER",
"timestamp": "2026-01-28T18:17:00.811Z",
"status": "COMPLETED",
"querySetId": "f4c35381-407c-45c7-89ec-094b8a4cd5b1"
}
}
]
}
}