Event examples and ecommerce

Data model · Tracking · Beta testing

Generate from the saved model

Open Collection → Event examples to generate anonymous and enhanced request JSON from the active saved model. Use Copy anonymous JSON or Copy enhanced JSON, or download the complete bundle with its consent/type notes. Unsaved model edits do not affect these examples. Clipboard access normally requires HTTPS; if the browser refuses access, the page selects the text for manual copy.

The same read-only generator is available with the dashboard disabled:

php bin/console app:analytics:examples > event-examples.json
php bin/console app:analytics:examples --mode=anonymous
php bin/console app:analytics:examples --mode=enhanced
php bin/console app:analytics:examples --example=ecommerce > ecommerce-examples.json

The command writes JSON to stdout and errors to stderr. The generator reads no events or website registrations, sends no requests, and changes no settings. Normal Symfony/Doctrine startup still requires infrastructure configuration; provide the documented DATABASE_URL and server version for your database. Examples use fixed synthetic values, never descriptions or observed values. Invalid model or collection configuration prevents saved-model generation.

The versioned bundle includes schema_version, synthetic, properties, notes, and examples.anonymous.payload / examples.enhanced.payload. Send only a payload as the request body. Replace the public website-token placeholder and use an Origin registered for that website. Do not use an administrator or organization sharing token. Copying consentState: "granted" does not obtain consent; wire enhanced collection to an actual affirmative choice.

Anonymous examples contain modeled properties explicitly allowed without consent and page_sequence: 2 when the separate page-depth setting is enabled. The same generated count appears in enhanced and ecommerce examples. Enhanced examples illustrate permitted modeled properties; unlisted ordinary scalar properties still require enhanced consent. Reserved marker and legacy reporting-only keys are identified but omitted from customData; disabled page depth is also omitted. The server supplies timestamps and optional geography. Goals have their own allowlist and are omitted. Kill switches and path exclusions still apply.

When the strict collection profile is active, the complete bundle begins with examples.strict.payload, containing only websiteToken, eventName and pagePath, and a note that the server keeps nothing else. The anonymous and enhanced examples remain for reference; the page shows a matching notice. --mode=anonymous and --mode=enhanced are unchanged.

A flat ecommerce starting point

The UI's Ecommerce recipe and --example=ecommerce provide a proposed model and synthetic purchase payloads. Ecommerce purchase JSON opens first, with copy and download controls; Anonymous collection JSON shows the request without consent-dependent properties. Recommended ecommerce model (YAML) contains the matching property configuration. This is a recommendation, not your saved model; generation never installs it. Review and merge the proposed definitions into your existing model, then regenerate saved-model examples and any new reporting columns. All proposed ecommerce properties require enhanced consent by default.

Keep event properties flat. For a purchase, use a total, currency, item count, and broad product category. Avoid customer/order/cart identifiers, email addresses, delivery information, payment details, and arrays of purchased items. Do not place identifying data in event names or page paths. A fixed purchase event name does not automatically record the separately configured purchase goal.

The complete ecommerce purchase JSON is a synthetic request body for POST /api/receive, separate from the model YAML. Replace the public website-token placeholder and connect enhanced tracking to explicit analytics consent. The visitor/session identifiers below are fixed examples, not identifiers to reuse for actual visitors. Copying "granted" does not obtain consent. Anonymous requests omit these identifiers and all proposed ecommerce properties by default.

{
  "websiteToken": "REPLACE_WITH_PUBLIC_WEBSITE_TOKEN",
  "eventName": "purchase",
  "pagePath": "/checkout/complete",
  "referrerChannel": "direct",
  "deviceClass": "desktop",
  "viewportBucket": "large",
  "internalTraffic": false,
  "consentState": "granted",
  "visitorId": "synthetic-visitor",
  "sessionId": "synthetic-session",
  "screenWidth": 1440,
  "customData": {
    "currency": "USD",
    "total_minor": 4999,
    "tax_minor": 400,
    "shipping_minor": 500,
    "item_count": 2,
    "discount_rate": 0.1,
    "product_category": "accessories",
    "checkout_step": "complete"
  }
}

Here total_minor is the final charged total, including tax and shipping; the component fields are a breakdown, not additional amounts to add again. For USD, 4999 represents $49.99. Use the minor unit appropriate to each currency, and group calculations by currency. Never sum unlike currencies. Integers preserve money amounts without binary floating-point rounding within the supported safe-integer range. Use double/float for approximate ratios or measurements such as discount_rate, not exact monetary amounts.

Define types using YAML type or the model editor's JSON value type. JSON strings are quoted, while numbers and booleans are not. The SDK and server omit mismatched typed values; they do not convert "4999" into 4999, or 49.99 into 49. See the type rules.

Emit a purchase once when checkout actually completes; page refreshes and retries can create duplicate events. This recipe does not add order-based deduplication. For item-level analysis, consider a separate fixed event per broad item category only when there is a concrete need and a reviewed privacy policy. Do not create tables per event type or duplicate order totals in each item event.

Numeric views for external calculations

Keep a property's existing column when a BI connection expects text. Add a distinct numeric_column for arithmetic instead of silently changing its type. Integer projections reject fractions and out-of-range values; float/double projections preserve fractions with approximate double precision. Numeric strings in historical rows remain NULL in numeric projections.

These projections belong to private, unsuppressed analytics_custom_* views. Use an appropriately authorized reporting connection for enhanced ecommerce analysis. Routine anonymous BI users must remain limited to approved bi_anonymous_* views, with their existing completed buckets and suppression. Do not combine raw or typed custom projections with those views to recover withheld cells. Visualization and report delivery remain in external tools.