Create a Sync from Databricks to Batch Profile Events

Create a Sync from Databricks to Batch Profile Events

Before you start

To create a Databricks → Batch Profile Events sync, you'll need:

• Access to the Batch dashboard
• A Databricks table or view containing one row per event, exposed through a SQL Warehouse
• A Databricks personal access token
• Your Server URL, HTTP Path, Database, and Table details
• A table (or view) that follows the Cloud Sync events input format (see below)

---

1) Prepare your Databricks table or view

Cloud Sync expects your Databricks source (table or view) to include:

1. A profile identifier (to know which profile the event belongs to)
2. An event name (the type of event being recorded)
3. A cursor field (to know which rows are new since the last run)
4. Any number of event attribute columns (sent to Batch as event properties)

---

1.1 One row per event

Your source must contain one row per event. Unlike profile attribute syncs, multiple rows can share the same custom_id — each row creates a separate event on the corresponding profile.

---

1.2 Required columns

Your table (or view) must include the following columns:

Column	Required	Description
custom_id	✅	The identifier of the profile the event belongs to
event_name	✅	The name of the event (e.g. add_to_cart, purchase)
last_updated_at	✅	Cursor used for incremental sync

Important: last_updated_at must be set at the time the event row is inserted and must not change afterwards. It is used only to determine which rows to fetch on the next sync run, not as the event timestamp.

---

1.3 Optional columns

Column	Description
event_time	The timestamp of the event, in RFC 3339 UTC format (e.g. 2026-04-17T04:25:00Z). If omitted, Batch uses the time of reception.

---

1.4 Event attribute naming rules (Databricks-compatible)

Databricks column names cannot contain characters like $, (, or ). Cloud Sync relies on prefixes in column names to represent typed event attributes.

Supported prefixes

Prefix	Meaning	Example
date__	Date attribute	date__purchased_at
url__	URL attribute	url__item_url

Any column without a prefix is sent to Batch as a standard string, number, or boolean attribute.

---

1.5 Handling objects

Object event attributes must be stringified as a JSON object before being passed to Cloud Sync. The connector parses the string and sends it in the correct format to the Profile API.

'{"color":"red","size":"M"}'

Example:

SELECT
  user_id                            AS custom_id,
  'purchase'                         AS event_name,
  inserted_at                        AS last_updated_at,
  to_json(product_details_struct)    AS product_details  -- stringified object
FROM raw.purchases;

---

1.6 Example schema (e-commerce)

CREATE TABLE events_data.batch_events (
    custom_id          STRING,
    event_name         STRING,
    last_updated_at    TIMESTAMP,

    -- Optional event fields
    event_time         STRING,            -- RFC 3339 UTC, e.g. '2026-04-17T04:25:00Z'

    -- Standard event attributes
    item               STRING,
    quantity           INT,
    price              DOUBLE,

    -- Typed event attributes
    url__item_url      STRING,            -- URL attribute
    date__purchased_at STRING,            -- date attribute

    -- Object event attribute (stringified)
    product_details    STRING             -- e.g. '{"brand":"Nike","size":"42"}'
);

How this maps in Batch:

• custom_id identifies the profile the event is attached to
• event_name sets the event type
• event_time sets the event timestamp (falls back to reception time if omitted)
• item, quantity, price become standard event attributes
• url__item_url is interpreted as a URL attribute
• date__purchased_at is interpreted as a date attribute
• product_details is interpreted as an object attribute

---

1.7 Using a View

If your raw event table doesn't match the expected naming or format, create a Databricks View that converts your schema into the correct conventions.

CREATE OR REPLACE VIEW events_data.batch_events_view AS
SELECT
  CAST(e.user_id AS STRING)                          AS custom_id,
  e.event_type                                       AS event_name,
  e.inserted_at                                      AS last_updated_at,

  date_format(e.occurred_at, "yyyy-MM-dd'T'HH:mm:ss'Z'") AS event_time,

  e.item_name                                        AS item,
  e.qty                                               AS quantity,
  e.unit_price                                        AS price,

  e.item_url                                         AS url__item_url,
  date_format(e.purchase_date, 'yyyy-MM-dd')         AS date__purchased_at,

  to_json(struct(e.brand, e.size))                   AS product_details
FROM raw.events e;

This approach lets you:

• rename fields with the correct prefixes (date__, url__)
• set a reliable last_updated_at based on insertion time
• format event_time to RFC 3339 UTC
• stringify object columns correctly

---

1.8 Handling nulls

If an optional column value is NULL, that attribute is simply omitted from the event. It does not affect other attributes on the same event or on the profile.

---

2) Create the Sync in the Batch dashboard

Cloud Sync is configured from the dashboard via a dedicated Sync module.

1. Open the Batch dashboard
2. Go to Data → Cloud Sync
3. Click Create Sync
4. Select Databricks as the source

---

2.1 Configure your Databricks connection

Enter:

Field	Description
Token	A Databricks personal access token with the sql API scope
Server URL	Your workspace instance name (e.g. dbc-xxxxxxxx-xxxx.cloud.databricks.com)
HTTP Path	The HTTP path of the SQL Warehouse to query (e.g. sql/1.0/warehouses/xxxxxxxxxxxx)
Database	The catalog/database containing your source table/view (e.g. workspace)
Schema	(optional) The schema containing your source table/view (defaults to default)
Table	The table or view name

Batch validates the connection before continuing.

---

2.2 Select the destination

In the Destination dropdown, select Batch > Profile events.

---

2.3 Configure event mapping

Cloud Sync applies a simple mapping model:

• custom_id → identifies which Batch profile to attach the event to
• event_name → sets the event type
• last_updated_at → used only for incremental sync logic
• all other columns → mapped to event attributes

---

3) How incremental sync works

Cloud Sync uses incremental processing, which means it does not re-import your full dataset at every run. Instead, it fetches only the rows that are new since the last successful sync.

---

3.1 The last_updated_at cursor

Batch stores the last successful cursor value internally.

At each run, Batch fetches only rows where:

• last_updated_at is greater than the last stored cursor

This makes sync runs faster, more scalable, and more cost-efficient.

Important: For events, last_updated_at should reflect when the row was inserted into the table, not when the event occurred (event_time). Do not backfill or modify last_updated_at after insertion.

---

3.2 Inserts and deletes

Incremental syncs capture:

• ✅ new event rows (inserts)

They do not capture:

• ❌ deletions — events sent to Batch are immutable; they cannot be removed via Cloud Sync

---

3.3 Best practices for reliable incremental syncs

To avoid missing events:

• Set last_updated_at at insertion time and never modify it afterwards
• Use a View if you need to rename columns or apply type conversions
• Partition or Z-order on last_updated_at for large tables

---

4) Test and enable your Sync

Before enabling the schedule:

1. Run a test sync
2. Verify:

   • Events are created on the correct profiles
   • event_time is set correctly (or defaults to reception time as expected)
   • date__ and url__ fields are interpreted correctly
   • Object attributes are correctly stringified

Once enabled, Batch automatically handles batching and retries.