How do must, should and filter work in a bool query?

Question

IT Interview Practice · Accepted Answer

Use `must` for required matching that contributes to relevance, `filter` for required yes/no conditions, and `should` for preferences unless you explicitly make them required.

## Search published shoes and prefer one brand

Assume `title` is text, while `status` and `brand` are keyword. Send this body to the product search endpoint:

```json
{
  "query": {
    "bool": {
      "must": [{"match":{"title":"shoes"}}],
      "filter": [{"term":{"status":"published"}}],
      "should": [{"term":{"brand":"example"}}]
    }
  }
}
```

- A published shoe from another brand can appear: it satisfies the required clauses.
- A published shoe from `example` also matches the preference, which can raise its score.
- A draft shoe is excluded regardless of brand because it fails the filter.

The preference contributes to ranking; it is not a guarantee that this brand always ranks first under every scoring situation.

## When should must become required

If at least one preference must match, set `minimum_should_match: 1` inside `bool`. Otherwise, with `must` or `filter` present, `should` defaults to optional. A bool query containing should clauses but neither must nor filter defaults to requiring one. Filters do not contribute to the score; cache eligibility is separate from whether a particular execution uses a cache.

[Official documentation](https://www.elastic.co/docs/reference/query-languages/query-dsl/query-dsl-bool-query).

Search published shoes and prefer one brand

When should must become required

Related questions