Event + Journey Deduping

Some brands have multiple event sources (GA4, Snowplow, Heap, Elevar, Blotout, etc.). Those sources can record the same touchpoint within seconds of each other, creating the risk of double-counting. For multi-source customers, SourceMedium uses deduplication rules to merge streams into a canonical view that preserves richness while preventing over-counting.

High-level approach

For multi-source data, we consider touchpoints duplicates when core attributes match within a small time window. The dedupe fingerprint typically includes:

Timestamp (rounded to seconds, in UTC)
Standardized event name
UTM source / medium / campaign (with NULL normalized)

Fields like utm_content, utm_term, and click IDs are preserved as enrichment but don’t always participate in the fingerprint (so you can still do creative-level analysis without inflating counts).

Where this shows up

These patterns are most relevant when you query MTA / journey tables, like:

your_project.sm_experimental.obt_purchase_journeys_with_mta_models

Recommended analysis patterns

Verify you’re not double-counting purchases

SELECT
  COUNT(DISTINCT purchase_order_id) AS distinct_orders,
  COUNTIF(sm_event_name = 'purchase') AS purchase_event_rows
FROM `your_project.sm_experimental.obt_purchase_journeys_with_mta_models`
WHERE event_local_datetime >= DATETIME_SUB(CURRENT_DATETIME(), INTERVAL 30 DAY);

Compare touchpoint volumes by source system

SELECT
  COALESCE(NULLIF(LOWER(TRIM(source_system)), ''), '(unknown)') AS source_system,
  COUNT(*) AS touchpoints
FROM `your_project.sm_experimental.obt_purchase_journeys_with_mta_models`
WHERE event_local_datetime >= DATETIME_SUB(CURRENT_DATETIME(), INTERVAL 30 DAY)
GROUP BY 1
ORDER BY touchpoints DESC;

Single-source customers should remain unchanged; multi-source dedupe is only applied when multiple sources exist for the same order/journey.

UTM Normalization Attribution

⌘I

Core Concepts

Attribution

Order Segmentation

Customer Record

Event + Journey Deduping

High-level approach

Where this shows up

Recommended analysis patterns

Verify you’re not double-counting purchases

Compare touchpoint volumes by source system

​High-level approach

​Where this shows up

​Recommended analysis patterns

​Verify you’re not double-counting purchases

​Compare touchpoint volumes by source system

High-level approach

Where this shows up

Recommended analysis patterns

Verify you’re not double-counting purchases

Compare touchpoint volumes by source system