What Is Elasticsearch Boost?

Q: How do I boost a field in a simple query?

Use the caret notation directly in your fields list. Open Kibana Dev Tools (8.12+), then run: GET /blogs/_search { "query": { "multi_match": { "query": "elasticsearch", "fields": ["title^2", "body"] } } } Now every match in the title field scores twice as high as anything in body .

Quick Fix:

Need to boost a field in Elasticsearch 8.12+? Just add ^N to your query. Example: "title": "elasticsearch^2" makes title matches twice as relevant. For more control, use boost: 2.0 in a function_score query.

What’s Happening

Think of Elasticsearch “boost” as a multiplier you slap onto a field or term to make it jump higher in search results. A field with ^3 suddenly counts three times as much as the baseline. Boosts don’t filter anything—they just nudge the ranking scores up or down. This magic happens inside Elasticsearch’s function_score query, which lets you tweak scores on the fly without touching your index.

How do I boost a field in a simple query?

Use the caret notation directly in your fields list.

Open Kibana Dev Tools (8.12+), then run:

GET /blogs/_search
{
  "query": {
    "multi_match": {
      "query": "elasticsearch",
      "fields": ["title^2", "body"]
    }
  }
}

Now every match in the title field scores twice as high as anything in body.

How do I boost via function_score?

Send a function_score request and let it do the heavy lifting.

Here’s a quick example:

GET /blogs/_search
{
  "query": {
    "match_all": {}
  },
  "functions": [{
    "field_value_factor": {
      "field": "views",
      "factor": 0.5,
      "modifier": "log1p"
    }
  }],
  "boost": 1.2,
  "boost_mode": "multiply"
}

The views field now contributes a factor of 0.5 × log(views+1) to the final score. boost_mode: multiply simply multiplies that factor against the base score.

How do I boost a term instead of a field?

Drop the caret notation right in the term clause.

Try this:

GET /products/_search
{
  "query": {
    "bool": {
      "should": [
        { "match": { "description": "wireless^3" }},
        { "term": { "category": { "value": "electronics", "boost": 2.0 }}}
      ]
    }
  }
}

That wireless^3 triples the score contribution from that term, while the category term uses an explicit boost: 2.0.

How do I verify the boost actually worked?

Turn on explain mode and inspect the _explanation block.

Add "explain": true to your query:

GET /blogs/_search
{
  "explain": true,
  "query": {
    "multi_match": {
      "query": "elasticsearch",
      "fields": ["title^2", "body"]
    }
  }
}

Peek at the _explanation block in each hit—you’ll see the exact multiplier Elasticsearch applied.

Why didn’t my boost have any effect?

Check three common culprits first.

Field data types
Boosting only works on fields that are actually indexed. If your field is keyword or has "index": false, boosts get ignored. Reindex with "index": true and try again.

Large result sets need rescoring

For big hit lists, run a fast first query (say, a match on title), then apply a second rescore query with boosted terms:

GET /blogs/_search
{
  "query": {
    "match": { "title": "elasticsearch" }
  },
  "rescore": {
    "window_size": 50,
    "query": {
      "rescore_query": {
        "match": { "body": "elasticsearch^1.5" }
      }
    }
  }
}

Debug with the Profile API

Turn on profiling to see the scoring breakdown:

GET /blogs/_search
{
  "profile": true,
  "query": {
    "multi_match": {
      "query": "elasticsearch",
      "fields": ["title^2", "body"]
    }
  }
}

Look in shards[0].hits[0].profile for warnings like “no field data.”

How can I keep boosts from breaking in the future?

Document a clear boost policy and refresh your mappings regularly.

Set consistent boost rules across the cluster and store them in a _boost_policy index or a runbook. Reindex every quarter to purge deprecated fields and refresh mappings. Watch score distributions in Kibana Lens—if the average score suddenly tanks, you’ve probably got a mapping or boost regression. For extra speed, pre-sort your index with index.sort.field using your primary boost field; that improves query speed and cache locality.