Headless WooCommerce. Instant.

Median response time across four CoCart v2 product endpoints — paginated listings with different sort orders, product categories, and a single product. CoCart’s rich response format (prices, images, variants, related products) makes payloads larger than standard WP REST API responses, but the drop-in still cuts response time by ~24×.

50 requests per endpoint per mode. All times in milliseconds.

Frosty Cache caches CoCart’s public product catalog endpoints and bypasses session-specific ones. Cart routes use a cart_key or CoCart-API-Cart-Key header to identify anonymous sessions — neither of these trip the standard auth checks, so Frosty Cache’s CoCart module handles them explicitly.

WP REST Cache does not cache cocart/v2 endpoints by default — its allowlist only covers wp/v2, confirmed live: requests to cocart/v2/products against a WP REST Cache install ran with no speedup at all across repeated requests. CoCart does not register itself with the plugin. Manual configuration through the WP REST Cache admin screen is required before any CoCart response is cached, and even then responses are stored as WordPress transients — WordPress still boots on every cache hit. Live-tested on its wp/v2 routes (the same underlying transient mechanism), that’s ~832ms median in production, vs Frosty Cache’s drop-in at ~751ms — see the full live comparison.

The drop-in reads a JSON file and exits — its speed is bounded by file I/O, not by WordPress. CoCart v2 product payloads are 10–20× larger than a standard post response (prices, image variants, related products, add-to-cart metadata) so there is simply more data to read off disk and send over the wire. The ~23ms gap vs the WordPress core benchmark is the cost of that extra payload, not the caching layer. These numbers use the full unfiltered response. Headless storefronts using ?_fields=id,name,slug,permalink,prices,images,categories to request only the fields they render cut the payload from ~52KB to ~23KB — live-tested on the same production host, the filtered request now caches correctly and drops the HIT median from ~997ms to ~856ms, roughly a 14% improvement on top of the drop-in’s existing saving.

20 requests per endpoint against a live CoCart v2 install on shared hosting. MISS baseline uses skip_cache=1 to bypass the drop-in. HIT is a normal request served by advanced-cache.php. Numbers include full network round-trip — real internet latency to a shared host, not a local loopback — which compresses the percentage gain but not the underlying savings: CoCart still skips WordPress, WooCommerce, and its own bootstrap entirely on a hit.

Median saving across the four endpoints: ~47% off TTFB — the largest live saving of the three benchmarked API surfaces on this host.

Same four product endpoints, same drop-in, same machine and same live host — the only variable is which API served the response. CoCart’s richer payload (related products, add-to-cart metadata, per-currency formatting) costs a few extra milliseconds on every mode, but the cache path itself is identical.

All times in milliseconds, median. Local = Local by Flywheel loopback (no network overhead). Live = production run against a shared-hosting demo site (full network round-trip included).

See how Frosty Cache performs on standard WordPress REST API endpoints (~15ms / ~47×) and WooCommerce Store API FSE block endpoints (~38ms / ~23×).