Coveo vs Elasticsearch: Licensing and Total Cost

This comparison gets written a lot, and almost every version of it quotes a dollar figure for Coveo. Those figures come from review sites and procurement databases, never from Coveo, which publishes no pricing page numbers whatsoever. Meanwhile the Elasticsearch side usually gets summarized as "open source, so free", which stops being true the moment the deployment is real.

What follows is the cost structure of each, read from primary sources, and the three things that actually move the total.

What each one charges for

Coveo meters query volume and, for its workplace product, seats. Elastic meters compute, storage and tokens. That difference matters more than any single rate, because it decides which direction your bill grows in as the company grows.

The Coveo detail that changes the math at 100 people

Coveo allocates index capacity differently on its two tiers, and for the intranet product the smaller tier is the bigger index until you pass roughly 200 users. Coveo for Intranet on Professional includes a fixed 500,000 items. On Enterprise it is 2,500 items per user, up to 60 million.

Run that out. At 100 users, Enterprise allocates 250,000 items and Professional allocates 500,000. At 200 users the two meet. Only above that does Enterprise become the larger index. A 120-person company that upgrades to Enterprise purely for capacity is buying a smaller index than it already had, and the upgrade is worth it for other reasons: the full connector directory, on-premises connectivity, 100 query pipelines instead of 10, 50 machine learning models instead of 10, a six-month rather than three-month machine learning data window, and 99.99% uptime.

The connector exclusions deserve their own line, because they catch buyers repeatedly. Coveo documents plainly that the SAP, Salesforce, ServiceNow and Shopify native connectors are excluded from the entitlements and must be purchased separately. Those are precisely the systems a large enterprise most wants searched. If your evaluation assumes Salesforce is included because a Coveo tier says "see connector directory", check the contract line rather than the tier chart.

Worth conceding clearly: Coveo includes document-level security on both tiers. Several competitors gate permission-aware search behind their top plan, and Coveo does not. It is also a public company, listed as CVO in Toronto, reported FY2026 revenue of US$148.34 million, and is named a Leader in the 2026 Gartner Magic Quadrant. This is a mature platform, not a bet.

Why the Elasticsearch side is not free

Elastic publishes real rates and they look modest. Its own worked production example, 20 GB of searchable data with 5 percent ingest utilization and 33 percent search utilization, comes to $190 a month on Elastic Serverless Search. The problem is not the rate. It is that the rate covers infrastructure and nothing else.

Until 2025 Elastic sold the rest as a product. App Search handled in-product search with a tunable relevance layer, and Workplace Search handled company tools with hosted connectors and document-level permissions. Both were retired. Elastic states that Enterprise Search "will no longer be available in Elastic version 9.0 and beyond", and that new customers who subscribed after the 9.0 release do not have access to them at all. Existing 8.x deployments hit End of Maintenance on January 15, 2027 and End of Support on July 15, 2027. The full date table and Elastic's published rates are on our Elastic Enterprise Search pricing page.

So a 2026 comparison of Coveo against Elasticsearch is not a comparison of two products. It is a comparison of buying a platform against building one on an engine. The build side of that ledger includes:

• Connector hosting. Self-managed connectors are services somebody deploys, monitors, patches and keeps authenticated against every source system as tokens rotate and APIs change.
• Relevance. What Coveo delivers as query pipelines and machine learning models becomes query DSL that a named engineer owns.
• Document-level permissions. The line item that consumes schedules. Every source has its own model, and each has to be mapped into index filters and re-checked as group membership changes.
• An interface. Elasticsearch ships an API. Employees need a search experience.

None of this is a knock on Elasticsearch, which is a superb engine running a very large share of the world's production search. It is a scope observation. Teams that budget the migration as infrastructure and discover halfway through that they have committed to operating an internal search product are the common failure mode here.

Which one costs less at a mid-sized company?

If search volume is low and engineering capacity is genuinely spare, Elasticsearch is cheaper and it is not close. If search is used daily across departments and nobody has a spare engineer for eighteen months, Coveo is cheaper once the build is priced honestly. The crossover is usually decided by headcount you can commit, not by the rate card.

A fair way to run it is to price the Elastic side with a loaded engineering rate for the build and a standing fraction of an engineer for operation, then compare that against a Coveo quote. Do not compare Coveo's quote against $190 a month. That comparison has appeared in a lot of published articles and it has cost people their credibility in front of a CFO who asked the obvious follow-up question.

Be careful with any figure you inherit. The $200,000 to $500,000 annual Coveo contracts quoted around the web come from procurement databases and review sites, not from Coveo, and the same is true of the $79 and $99 monthly Elastic plans in various roundups. Label them as third-party estimates when you put them in a deck, because somebody will ask for the source.

The third option most evaluations skip

Both of these products date from an era when internal search meant a ranked list of links. What people now expect from a work question is an answer, assembled across systems, with citations they can open and check. Coveo has moved in that direction and Elasticsearch supports it if you build it, but if answers over company knowledge are the actual requirement, the shortlist should include tools built for that from the start rather than only these two. We keep an honest breakdown on enterprise search software, and the adjacent cost comparison on Coveo vs Lucidworks.

It is also worth scoping the corpus before pricing any of it. A lot of what teams assume belongs in a search index does not. High-volume transactional email, for instance, is usually wanted as structured fields in a spreadsheet rather than as searchable documents, and the cheaper fix there is to pull those fields out of inbound email automatically rather than to index every message and pay to store it. Cutting a corpus in half before you sign changes a Coveo pricing unit count and an Elastic storage line at the same time.

How InSearch prices against both

InSearch is a per-seat product, published on the plans page, which makes it the simplest of the three to budget. Google Drive, Slack, Notion, Confluence, Gmail, Jira and Salesforce connect once, centrally, and each source's own permissions are enforced per person at query time, so nobody ever sees a snippet from a document they could not already open. Answers come back cited to the source document. There is no separate connector purchase for the systems most companies actually run, and no query meter that penalizes the teams who use it most.

The honest boundary: if your requirement is customer-facing site search or commerce relevance tuned per segment, Coveo is built for that and we are not. If it is log analytics, observability or search embedded inside a product you ship, Elasticsearch is the right foundation. If it is employees finding answers across internal systems without anybody hosting connectors, that is the job unified search exists to do, and the mechanism that makes it safe is on permission-aware AI search.