WebEngage

Before You Start

• Identify your JustAI org slug (usually your company name in all lower case - feel free to ask us).

Set up a JustAI test in WebEngage

Create a Content API pointing to JustAI

1. In WebEngage, go to: Settings → Data Platform → Custom Content APIs → Add new API
   • Name: JustAI
   • Endpoint URL: https://worker.justwords.ai/api/generate/<org_slug>?template_id=<template_id>
   • Headers:
     • X-Api-Key: JW_API_KEY
     • Content-Type: application/json
2. Enable parameters you’ll pass to the API (for personalization and variant generation):
   • user_id
   • email
   • gender
   • cohort
   • lifecycle_stage
   • etc.
3. • Set request type → POST
   • Response type → JSON

Example usage inside WebEngage message editor

Within the WebEngage HTML editor (or your “Dynamic Content” block):

{% set jw = callContentApi('JustAI', {
  "params": {
    "email": user.email|default(''),
    "user_id": user.id|default(''),
    "gender": user.gender|default(''),
    "cohort": user.cohort|default('')
  }
}) %}


<!-- Insert the HTML returned by JustAI -->
{{ jw.copy.vars.body|safe }}


<!-- Optional debug/analytics marker -->
<!-- JW_COPY_ID: {{ jw.copy.id|default('') }} -->

✅ Notes:

• You can also use {{ jw.copy.vars.subject }} or {{ jw.copy.vars.preheader }} in your subject line/preheader fields.
• If your workspace uses contentapi() or a different helper, adapt the syntax.
• Always include safe defaults for missing attributes.

Build the experiment in Journeys / Campaigns

1. In Journeys, add an A/B Split Node.
2. Path A (Control): your standard email.
3. Path B (Treatment): clone the email and replace the body (or subject) with the JustAI Jinja block above.
4. Launch the campaign to compare performance between control and AI-generated copy.

Send metrics back to JustAI (Control & Treatment)

Option A — WebEngage Data Pipelines / Webhooks (Recommended, near-real-time)

1. Ask WebEngage to enable Data Pipelines (Event Webhooks).
   • Configure a new webhook destination: https://worker.justwords.ai/api/webhook/webengage/<org_slug>
   • Headers:
     • Authorization: Bearer JW_API_KEY
     • Content-Type: application/json
2. Select event types:
   • Email Sent
   • Email Opened
   • Email Clicked
   • Email Unsubscribed
   • Email Bounced
3. Include identifiers in payload:
   • campaign_id
   • variation_id (A/B branch)
   • user_id
   • email
   • event_time
   • optionally jw_copy_id (if you inject it into the email body as a hidden comment or UTM param)

Example payload → JustAI event:

{
  "platform": "webengage",
  "channel": "email",
  "event": "email_click",
  "campaign_id": "{{campaign_id}}",
  "variation_id": "{{variation_id}}",
  "user_id": "{{user_id}}",
  "email": "{{email}}",
  "timestamp": "{{event_time}}",
  "metadata": {
    "journey_id": "{{journey_id}}",
    "link_id": "{{link_id}}",
    "jw_copy_id": "{{jw_copy_id}}"
  }
}

🧠 How JustAI uses it:

These webhook events are joined (by user_id/email, campaign_id, and timestamp) with JustAI’ serve logs from the Content API (which include copy_id). This enables copy-level lift tracking (open/click/conversion).

Option B — Batch Exports to S3 / Warehouse

If your WebEngage account doesn’t have Pipelines:

1. Configure daily exports to S3 (or Redshift/BigQuery).
2. Include events: send, open, click, unsubscribe.
3. Add identifiers: campaign_id, variation_id, jw_copy_id, and user attributes.
4. Point JustAI to the bucket prefix (e.g., s3://jw-ingest/webengage/<org_slug>/events/).

JustAI will ingest daily to update per-copy performance metrics.

Example Personalization Tests

First-name personalization (safe fallback)

{% set first = user.first_name %}
{% if first %}
  Hi {{ first }},
{% else %}
  Hi there,
{% endif %}

Always provide defaults. Use conditional logic to avoid blank greetings or malformed text.

Segment-based variants

• Ensure user attributes like gender, plan_type, or cohort are tracked in WebEngage user profiles.
• Pass them as parameters to the JustAI API so the engine can return gender- or cohort-specific copy.
• Optionally, layer an A/B split within each segment to test per-segment winners.