How to Build a Custom Attribution System for Shopify

How to Build a Custom Attribution System for Shopify

Building a custom attribution system for Shopify means creating your own infrastructure to collect touchpoint data, stitch user journeys together, and assign conversion credit across marketing channels.

The Five Core Components

1. Data Collection Layer

You need to capture every interaction a visitor has with your marketing and your store:

• Click data: UTM parameters, click IDs (gclid, fbclid, ttclid), referral URLs
• Session data: Pages viewed, cart actions, checkout steps
• Conversion data: Order ID, revenue, products, customer email (hashed)
• Ad platform data: Campaign spend, impressions, and clicks from Meta Ads, Google Ads, and TikTok Ads APIs

On Shopify, click data is captured through JavaScript on your storefront. When a visitor lands, a script reads URL parameters and stores them in a first-party cookie.

2. Identity Resolution Layer

Identity resolution connects multiple sessions from the same person into a single profile. Common identifiers include:

• First-party cookie ID: Works within a single browser but fails across devices
• Hashed email: Available once the customer enters their email at checkout or via Klaviyo signup
• Shopify customer ID: Assigned after account creation or first purchase
• Click IDs: Platform-specific identifiers matchable to platform data

The most reliable approach combines cookie-based tracking with deterministic matching on hashed email. When a customer identifies themselves, you backfill their anonymous sessions with their known identity.

3. Data Storage Layer

Store raw touchpoint events in an append-only table. Each row represents one touchpoint: timestamp, user identifier, channel, campaign, and metadata. Conversion events go in a separate table linked by user ID.

4. Attribution Modeling Layer

Standard attribution models include:

• Last-click: Simple but undervalues upper-funnel channels
• First-touch: Useful for acquisition analysis but ignores conversion paths
• Linear: Equal credit to all touchpoints, a reasonable default
• Time-decay: More credit to touchpoints closer to conversion

Start with linear or time-decay. Moving to data-driven models requires expertise in causal inference or marketing mix modeling, which is where build-vs-buy becomes serious.

5. Reporting Layer

Build dashboards answering three questions: What is the true ROAS of each channel? Where should spend increase or decrease? How do customer acquisition costs vary by channel?

Implementation on Shopify

Step 1: Deploy server-side cookie tracking. Use a Cloudflare Worker or Shopify App Proxy to set first-party cookies via HTTP headers.

Step 2: Capture events via Shopify webhooks. Subscribe to orders/create and orders/paid. When an order fires, retrieve the customer's touchpoint history and run your attribution model.

Step 3: Pull ad platform spend data. Schedule daily jobs to pull spend from Meta, Google, and TikTok APIs. Match spend to attributed revenue for true ROAS.

Step 4: Build the attribution pipeline. Create a batch job that processes raw events into attributed conversions nightly.

Step 5: Validate against platform data. Your attributed conversions per channel should be lower than platform self-reports but sum to actual revenue.

Build vs. Buy

Browser privacy changes, API updates, and Shopify's evolving checkout architecture all require continuous adaptation.

Common Mistakes

1. Relying solely on client-side JavaScript: Browser restrictions degrade data within months. Use server-side tracking as the foundation.
2. Ignoring view-through conversions: Click-only tracking undervalues social and video channels.
3. Not testing incrementality: Multi-touch models still assign credit to touchpoints that may not have been necessary. Without incrementality testing, you cannot distinguish correlation from causation.
4. Underestimating maintenance: The hardest part is keeping it accurate as browsers, platforms, and Shopify evolve.