WooCommerce Cart Cache: Test Two Sessions Before You Go Live

Imagine you’ve turned on page caching for a WooCommerce shop. A product page loads quickly, and your first test cart looks fine. That’s only one browser session. The revealing test is whether a second shopper can fill a different cart while both use the same public URLs.

Exclude the store’s Cart, Checkout and My Account pages from shared page caching, and verify the exclusions with two independent sessions. Keep public catalog caching if it passes its own checks. If a test session shows the other session’s cart or account details, pause the rollout and investigate the route before allowing shared caching there. WooCommerce’s caching guidance identifies pages and cookies that need special handling.

Make the two carts impossible to confuse

Use staging with the same relevant cache layers and rules as production, dummy products and test accounts. Open two separate browser profiles. Two private windows in the same browser can share a private session, so use profiles or separate browsers. Use the same canonical hostname and the store’s configured page URLs in both sessions. A query-string cache buster can hide a problem with the URL shoppers actually use.

• Session A: Add the dummy red mug. Open the cart and checkout, then leave the session open.
• Session B: Start with an empty cart and visit the same cart URL before adding the dummy blue plate. Open checkout and confirm that its order summary shows only the plate.
• Return to A: Reload the same URLs. Its cart and summary should still show the mug. With two dummy accounts, repeat the separation check on My Account. Stop before submitting payment or sending an order.

As a control, load a public product page without a cart or a login. Compare what the storefront actually returns. A cache hit for a public description can be fine. A cart total, order summary or account name that crosses sessions is a failure even if the page looked fast.

Use headers to find the layer, then trust what the shopper sees

Open the browser’s Network panel and look at the document response. If the cart or checkout updates after the page loads, also inspect the relevant cart or checkout request in the network log. Block-based checkout can use the WooCommerce Store API. Record the URL, status, cache directives and any vendor cache indicator. Don’t include cookies, authorization headers or customer details in a support ticket. WooCommerce documents Store API checks for checkout troubleshooting.

Under RFC 9111, an unqualified private response can be stored by a private cache but must be excluded from shared caches. no-store instructs caches not to store the response. An Age header reports a cache’s estimate of the response’s age. Its absence doesn’t prove that an origin page-cache plugin or another layer bypassed the request. A vendor-specific hit marker is a clue about the layer that emitted it. The two-cart behavior is the actual check.

Change one cache rule at a time

Start with the configured store paths and WooCommerce’s session-cookie guidance. A WordPress plugin, a host-level cache and a CDN can each apply their own rules. Check each layer that can store the response, including the relevant dynamic endpoints. Keep assets like images and stylesheets separate from page and session responses. After correcting the affected rules, invalidate the affected cached responses where your setup allows and rerun the same test with clean sessions.

If both carts stay separate but checkout still fails, follow the failing request rather than assuming caching caused it. A redirect, blocked request or JavaScript error needs a different fix. If you see signs of cross-session data on production, ask the cache owner to bypass and purge the affected routes and follow your incident process. Avoid trying to reproduce it with real customer data. Share redacted request details and timestamps with the host or cache provider. For a wider launch rehearsal, use the WordPress migration checklist.

This guide was prepared with AI assistance using the linked documentation. The carts are hypothetical and the checks were not run on a live store.