Case 10: November 20th Sales Investigation

Case 10: Real-World Attribution Analysis: November 20th Sales Investigation

The Question

“What happened on November 20th?”

The Goal

Understand which traffic sources drove sales that day

Time to Answer

~1 minute

Step 1: Pull Sales Event Data

Action

Query sales events for November 20th from Moon

What you get (spreadsheet/database export)

User ID	Affid	Source	Campaign Type	Campaign ID	Parameters
019a974d0bb877f19905e4e10c3d5892	-	google ads	g_campaignid	22740491004	{“_gl”:“1*19k±…“}
019aa24361a57f03928122b309e186f6	1000	fb	campaign_id	120236311971890504	{“_gl”:“1*17a±…“}
019aa212fd9773b4b6babfb8d9cdface	1000	fb	campaign_id	120233281060520682	{“_gl”:“1*1uz±…“}
019a9fc799fa7192b394c5fc08e2d417	-	test	-	-	{“content_ids…}
019aa0d1753f7376bd630c614ee55c00	-	test	-	-	{“affid”:"",“co…}
019a9f55a76a7fd988c6401a79f751a2	1001	google_ads	ad_id	781	{“ad_id”:“781…}
019aa0bc82b375a3b1f02bf4983328a9	-	test	-	-	{“_gl”:“1*1e4…}
019a9edc73fb7d6a8b921122171e0bcf	1001	google_ads	-	-	{“_gl”:“1*16a…}
019aa0fffce8710fa99a2c0a9affe962	-	test	-	-	{“affid”:"",“co…}
019aa1f79f9277088d50e7c54eb10eba	-	mega-funnel	-	-	{“affid”:"",“co…}
bceb8c64219a4188bb78bf144f20363c	-	email_crm	direct-crm-ticket	-	{“_gl”:“1*ve7c…}

Time elapsed: 1 minute to export this data

Step 2: Analyze the Attribution

Immediate insights from the data

Traffic Source Breakdown:

Facebook (affid 1000): 2 sales - campaigns 120236311971890504, 120233281060520682
Google Ads (affid 1001): 2 sales - properly tracked with affid
Google Ads (no affid): 1 sale - missing affid but has Google campaign parameters (g_campaignid: 22740491004)
Test traffic: 4 sales - obvious from specific IDs and “test” source label
Email/CRM: 1 sale - from direct-crm-ticket (email_crm source)
Mega-funnel: 1 sale - from alternative funnel
SMS: Hidden in the data (discovered by parameter analysis)

Step 3: Deep Dive - Parameter Investigation

The “Google ads” sale without affid (row 1)

Source shows: google ads
Affid: empty
But parameters show: {"_gl":"1*19k±...}

Look deeper into parameters → Find SMS prospect tracking codes

Actual source: SMS retargeting (not Google Ads)

User originally came from Google Ads (stored in their profile)
Was retargeted via SMS
Clicked SMS link and purchased
Parameters reveal the SMS tracking codes
Original attribution: Google Ads campaign

This is why parameters are the “attribution Bible” → They reveal the true story

The Parameter Deep Dive

Every sale has parameters - JSON object with tracking data

Facebook sale parameters:

{
  "_gl": "1*17a±...",
  "utm_source": "fb",
  "utm_medium": "cpc",
  "campaign_id": "120236311971890504",
  "ad_id": "120233281654680682",
  "adset_id": "120233281060550682",
  "affid": "1000",
  "fbp": "fb.1.1764081009618..."
}

Email/CRM sale parameters:

{
  "_gl": "1*ve7c...",
  "source": "email_crm",
  "campaign": "direct-crm-ticket",
  "email_id": "msg_abc123",
  "original_utm_source": "fb",
  "original_campaign_id": "120236311971890504"
}

SMS sale parameters (disguised as Google Ads):

{
  "_gl": "1*19k±...",
  "sms_campaign": "sms_prospect_nov_20",
  "sms_message_id": "msg_sms_xyz",
  "original_utm_source": "google",
  "original_campaign_id": "22740491004"
}

These parameters reveal

Original traffic source (first-touch)
Current traffic source (last-touch)
Campaign identifiers
Ad identifiers
Email/SMS identifiers
All tracking codes

Without reading parameters, you’d think: “This is a Google Ads sale”

With parameters, you discover: “This is an SMS retargeting sale, originally from Google Ads”

Multiple Events Per Sale

Each sale has at least 3 events in the system

Event 1: Landing (Day 1)

{
  "event": "conv_landing_page_view",
  "user_id": "019a974d0bb877f19905e4e10c3d5892",
  "timestamp": "2025-11-18T10:23:45Z",
  "parameters": {
    "utm_source": "google",
    "utm_campaign": "weight-loss-q4",
    "campaign_id": "22740491004",
    "landing_page": "/medication-tirzepatide-discount-01"
  }
}

Event 2: Lead Capture (Day 1, 15 minutes later)

{
  "event": "achieve_screen_viewed",
  "user_id": "019a974d0bb877f19905e4e10c3d5892",
  "timestamp": "2025-11-18T10:38:22Z",
  "properties": {
    "email": "user@example.com",
    "phone": "+1-555-0123"
  },
  "parameters": {
    "utm_source": "google",
    "campaign_id": "22740491004"
  }
}

Event 3: Sale (Day 3, via SMS retargeting)

{
  "event": "purchase_completed",
  "user_id": "019a974d0bb877f19905e4e10c3d5892",
  "timestamp": "2025-11-20T14:52:33Z",
  "properties": {
    "email": "user@example.com",
    "amount": 499,
    "transaction_id": "txn_abc123"
  },
  "parameters": {
    "utm_source": "google", // Original source (preserved)
    "campaign_id": "22740491004", // Original campaign
    "sms_campaign": "sms_prospect_nov_20", // Last-touch
    "sms_message_id": "msg_sms_xyz"
  }
}

All 3 events share the same user_id → Connected journey

This enables multi-touch attribution: You can see the full path from landing → lead → sale

Understanding the “Footprints”

Every event has a schema - the JSON structure with all parameters

PostHog event schema:

user_id (or distinct_id)
event (event name)
timestamp
properties (event-specific data)
$set (user properties)
utm_source, utm_medium, utm_campaign, etc.
Custom parameters (campaign_id, ad_id, affid, etc.)

Your marketing events have their footprints

Facebook ads: campaign_id, ad_id, adset_id, affid, fbp
Google Ads: gclid, campaign_id, g_campaignid
Email: email_id, campaign, source: email_crm
SMS: sms_campaign, sms_message_id
Organic: utm_source: organic, referrer data

These footprints are how you see things - the ONLY way you see things.

If you don’t understand the footprints

You can’t read the data correctly
You’ll misattribute sales (thinking SMS = Google Ads)
You’ll miss patterns
You’re “a bad hunter”

The Practical Workflow

No meetings needed. No complex analysis. Just

Notice what IDs exist: “Oh, we have PostHog IDs. Fine.”
See what parameters are captured: “We have affid, campaign_id, utm_source. Cool.”
Use them: Query by these fields, segment by them, attribute by them
Done: Takes 1 minute to pull sales attribution

This isn’t magical

The system stores parameters on every event
Parameters persist through the funnel
Query pulls parameters with sales data
You read the parameters to understand attribution

Common pattern

See PostHog tracking → Know you have user IDs and events
See affid field → Know you can segment by traffic source
See campaign_id → Know you can attribute to campaigns
Just use what’s there

No need to overthink it - if the parameter exists, use it.

Key Insights from the November 20th Analysis

Traffic Sources Driving Sales

Facebook (affid 1000): 2 sales
Google Ads (affid 1001): 2 sales
Email/CRM retargeting: 1 sale
SMS retargeting: 1 sale (discovered via parameters)
Test traffic: 4 sales (internal testing)
Alternative funnel: 1 sale

Retargeting Impact

Multiple sales came via email/SMS retargeting
But all attributed back to original paid campaigns
Parameters reveal both original and last-touch sources

Parameter Reading Reveals Hidden Patterns

One “Google Ads” sale was actually SMS
Email sales traced back to original Facebook campaign
Multi-touch journey visible through parameter history

Principles Applied

End-to-End Attribution is Non-Negotiable ✓
PII is the Attribution Anchor ✓ (email connects events)
First-Touch Attribution is Sacred ✓ (parameters preserve original source)
Know What You Want to Check ✓ (understand event schema)
Understand Your Event Footprints (Schema) ✓
Parameters Reveal the Truth ✓

Tool Features That Enable This

Event export with all parameters
User ID linking across events (landing → lead → sale)
Parameter preservation through funnel
JSON parameter storage (flexible schema)
Multi-touch tracking (original + current source)
Fast querying (1 minute to answer “what happened on Nov 20?”)

The Bottom Line

With properly instrumented events and parameter tracking, you can answer complex attribution questions in minutes, not days. Just understand your data schema and use it.

Previous: Case 9: Multi-Touch Attribution
Next: Case 11: Cross-Validating Attribution
Back to Index: All Cases