Elasticsearch is a crucial component of Intercom's infrastructure, supporting various core features like Inbox and API. The company faced an availability risk due to uneven distribution of traffic/work among nodes in its Elasticsearch clusters, leading to load imbalance issues. To address this problem, Intercom migrated some of its clusters to fewer but bigger data nodes, which helped distribute shards more evenly and reduce the load on individual nodes. However, this approach had limitations, including the need for larger nodes, which could be expensive to add incrementally. A longer-term solution involves increasing the number of smaller shards relative to the number and size of data nodes, which can help achieve better load distribution without compromising performance. Intercom achieved significant cost savings and performance gains by implementing these changes, reducing costs by up to 40% in some cases and improving median request latency by 100%. The company's goal is now to make it easier for engineers to interact with Elasticsearch as they introduce new products and features, aiming to provide a seamless experience without worrying about the underlying infrastructure.