Share the article
CarJager, France's marketplace for classic and vintage cars, migrated its site search from Algolia to Meilisearch so its lean product team could scale search more easily and maintain steady search performance during traffic peaks.
About CarJager
CarJager is a French marketplace specializing in the buying and selling of classic and vintage cars, offering a secure, curated platform for collectors and enthusiasts. Headquartered between Montpellier and Aix-en-Provence, the company runs a catalog of around 2,000 listings and 2,500+ editorial blog articles serving a passionate community of vintage car lovers across Europe.
Challenge: an incumbent that didn't scale with the catalog
CarJager had been running its search on Algolia for years, but as the catalog grew and contract renewal approached, the setup stopped making sense on several fronts. A few problems came to a head at the same time.
A hard sorting limit that broke when it capped
Algolia handles sorting through replicas, a separate index per sort order, and each replica is capped at 1,000 entities for navigation. For Tanguy Thomas, CarJager's Lead Developer, that was a recurring source of friction: with ~2,000 active listings, the team had been bumping into the ceiling for a while, and there was no clean way to work around it without restructuring the index.
No room for DevOps
The team also looked at running open-source alternatives like Typesense in-house, but quickly ruled it out. With no dedicated DevOps team, they didn't want to take on infrastructure, updates, or traffic-peak management themselves. The solution had to be SaaS, predictable, provide multi-language support, and be easy to maintain.
A pricing model out of step with usage
On top of the technical limits, the economics had drifted. With ~1.3M+ searches per month against a relatively small index, Algolia's pricing felt disproportionate to what CarJager was actually getting in return, and renewal was the natural moment to act on it.
Why Meilisearch
While evaluating the possible alternatives, a few things tipped the decision for the CarJager team:
An API that felt familiar. Coming from Algolia, the mental model transferred almost directly. Filters, facets, sorting, typo tolerance, and synonyms: everything they relied on was there, and the API surface was clean enough to integrate without a heavy lift.
"It's really straightforward to set up. The API is very simple. Honestly, in terms of implementation, it was almost child's play." - Tanguy Thomas, Lead Developer at CarJager
No artificial sorting limits. Meilisearch's approach to sorting and faceting doesn't require replicas, which removed the 1,000-entity ceiling that had been a recurring annoyance on Algolia.
Managed cloud with automatic scaling. Meilisearch's autoscaling model automatically bumps resources when traffic spikes and only loops the team in for moves to a larger instance class. That fits CarJager's "no dedicated DevOps" reality. The team ran stress tests early on and confirmed the platform held up under their expected peaks.
A cost structure that worked at their size. Pricing for a managed instance was significantly more aligned with CarJager's usage profile than Algolia's per-search and per-record model, enough to make the migration worth the effort even before counting the other wins.
The migration: a phased, low-risk rollout
CarJager designed the migration to minimize risk, running Algolia and Meilisearch in parallel, moving content over in waves, and only switching Algolia off once the new setup had proven itself.
- Phase 1: blog search. Lowest-traffic, lowest-stakes index. Used to validate end-to-end behavior, facet configuration, and dashboard workflows.
- Phase 2: car listings. The core marketplace experience, with ~2,000 vehicle listings filtered by brand, model, price, and sale type (auction, assisted sale, classified).
Throughout the evaluation and migration process, the Solution Engineering team from Meilisearch helped cover the technical questions such as: sorting behavior, autoscaling triggers, index design, and pricing impact of scale-ups.
Results: live, stable, and running comfortably
What changed for the team
- Developer experience: a familiar API surface meant migration was a matter of days of focused work, not a drawn-out re-architecture project.
- Sorting: no more 1,000-entity replica ceiling; sorting works natively across the full listings catalog.
- Operations: autoscaling handles traffic peaks without manual intervention. The product team stays focused on product, not infrastructure.
- Cost: significant reduction in their search bill vs. Algolia. The team right-sized their instance after launch, downgrading from L to M once real-world traffic patterns confirmed the M tier was enough, taking advantage of the cost flexibility.
What's next
With the core marketplace migration behind them, CarJager team has flagged a few directions for further investment:
- Semantic search, using AI-powered relevance to better surface cars by description, era, or use case rather than exact-match keywords.
"For a small team with no DevOps, Meilisearch hit the right balance: an Algolia-familiar API, none of the sorting limits, and pricing we can justify." - Tanguy Thomas, Lead Developer at CarJager
About Meilisearch
Meilisearch is a developer-first search engine that delivers fast, typo-tolerant, AI-ready search out of the box. Available as open source and as a managed service, it powers product, content, and document search for thousands of teams worldwide, from small product teams to large enterprises.
