[alecritson]

The checkout renders its own document (lunar-checkout::app), so a store had no way to add Tag Manager to it, and the only checkout event was purchase (confirmation). A store's Checkout or Buy Now request ends in Inertia::location, so the store can't reliably send begin_checkout either.

Changes

• Head hook: lunar-checkout::partials.head, an empty include at the end of <head>. A store overrides it from resources/views/vendor/lunar-checkout/partials/head.blade.php.
• GA4 step events pushed to window.dataLayer, with currency, value and items:
  • begin_checkout on first load of the checkout, once per session per tab (tracked in sessionStorage). The store must not send its own; KJP removed its one in kjpspares.co.uk c44afc6d.
  • add_shipping_info with shipping_tier, when a delivery method or collection is chosen. If the server chose it, it is sent at pay instead.
  • add_payment_info with payment_type (the payment method handle, e.g. card), when the customer pays.
  • Each is sent once per tier or method, so a retry isn't counted twice.
• Consent stays with the host: events go out only when window.dataLayer exists. The host creates it in the head hook once the shopper has consented.
• Tests: npm test (node:test, no new dependencies) covers the event helper and runs in the existing vite build job. There are two feature tests for the head hook.
• README section and a rebuilt resources/dist.

Test plan

• vendor/bin/pest --testsuite checkout (317 passed)
• npm test
• phpstan
• KJP (KJS-209): in GTM preview, a checkout run shows begin_checkout, add_shipping_info, add_payment_info, purchase. With cookies rejected, nothing goes to googletagmanager.com.

Refs KJS-209

🤖 Generated with Claude Code

[alecritson]

Added in 1d9bf9c: the checkout now sends begin_checkout itself, on first load, once per session per tab (tracked in sessionStorage). A store's Checkout or Buy Now request ends in Inertia::location, so the store can't send it reliably (kjpspares.co.uk#155 review). Covered by npm test.

[TheAaronLawrence]

I've left 2 comments: begin_checkout now also fired here; may duplicate storefront's event; payment_type uses the display label, which may be localized or vary.

Additional findings

• payment_type uses the display label, which may be localized or vary (resources/js/composables/useCheckout.js:349) · Low · Quick check

  trackPayment uses method.label ?? method.handle as payment_type. Labels are display text that can change with locale or configuration, which fragments GA4 reports and makes the dedupe key unstable. Prefer the stable handle, or send the handle as the value and keep the label out of it.

[TheAaronLawrence]

begin_checkout now also fired here; may duplicate storefront's event
Medium · Quick check

The PR description and KJS-209 say begin_checkout is already sent by the storefront (KJS-97, basket and Buy Now). This PR adds trackCheckoutStart, which fires begin_checkout again on checkout load. The README even says the store should not send its own. If the storefront still sends it, GA4 will count two begin_checkout events per checkout and distort funnel numbers. The change is also missing from the PR description's change list.

Suggested fix: either remove it from the package, or confirm that KJS-97's event is removed in KJP and document this in the PR. The sessionStorage dedupe only protects against reloads within this app. It doesn't protect against the storefront's event.

[alecritson]

No double count: KJP already stopped sending its own begin_checkout in neondigital/kjpspares.co.uk@c44afc6d (useAddToCart.test.ts checks it stays out). The checkout owns the event now. I've added it to the PR description's change list.

[alecritson]

payment_type: fixed in 24e66a0. It now sends the payment method handle (card, on-account) instead of the label. README updated and resources/dist rebuilt. npm test passes.