Improving search performance in Elasticsearch is often more effective through data modeling rather than merely tuning settings, as altering document modeling can dramatically enhance performance. For instance, using an n-gram tokenizer instead of wildcard queries can speed up searches by indexing every substring, transforming slow wildcard queries into faster term queries. When users search real-estate listings, aggregations can be optimized by pre-computing data, directly indexing price and surface ranges to replace range and histogram aggregations with terms aggregations. Further optimization is achieved by merging multiple terms aggregations into a single one, reducing the number of aggregations from six to two while maintaining the same information. Although this approach might limit flexibility, as changing ranges would require reindexing, it can complement the traditional method by retaining original data fields and allowing default set ranges for most requests, while using regular range aggregations for more detailed queries.