Attribution Modeling

Attribution Modeling

Overview

Attribution modeling determines which marketing touchpoints receive credit for a conversion, enabling informed decisions about where to allocate ad spend. Every ad platform (Meta, Google, TikTok) reports attribution using its own model — typically claiming 100% credit — which means the sum of all platform-reported revenue routinely exceeds your actual revenue.

This skill guides you through setting up first-party attribution on your platform, comparing attribution models side by side, and using dedicated attribution tools that do this automatically without building custom pipelines.

When to Use This Skill

• When marketing channels (Google Ads, Meta, email) each claim different shares of the same revenue
• When needing to make budget allocation decisions across acquisition channels
• When moving beyond last-click attribution to understand the full customer journey
• When building a marketing analytics report that compares channel performance under multiple attribution models
• When implementing first-party attribution to replace data lost from iOS tracking changes
• When affiliate, influencer, and paid search all contributed to the same order and each claims 100% credit

Core Instructions

Step 1: Determine your platform and choose the right attribution tool

Step 2: Set up UTM tracking correctly across all channels

Good attribution starts with consistent UTM parameters. Without them, 40–60% of traffic appears as "direct" (dark traffic).

UTM naming conventions to enforce across your team:

Platform-specific setup:

• Google Ads: Enable auto-tagging (adds gclid automatically); also add UTM parameters under Campaign → Settings → Additional settings → Campaign URL options
• Meta Ads: Add UTM parameters under Ad Set → Website URL, or use Meta's URL parameters feature at the account level
• Klaviyo: Go to Account → Settings → UTM Tracking — enable automatic UTM appending for all email and SMS campaigns
• Affiliates/influencers: Generate unique UTM links per creator using a URL builder; track in a spreadsheet

Step 3: Configure attribution in your chosen tool

---

Shopify

Option A: Built-in Shopify Analytics (last-click, free)

1. Go to Analytics → Reports → Sessions over time 2. Filter by Referral source to see which UTM sources drove sessions 3. Go to Analytics → Reports → Sales by traffic source for revenue by channel (last-click attribution) 4. Limitation: Shopify's built-in attribution is last-click only with a 30-day window; it cannot show multi-touch paths

Option B: Triple Whale (recommended for DTC brands spending $50K+/mo on ads)

1. Install Triple Whale from the Shopify App Store 2. Triple Whale's pixel fires on every page load, capturing the full click path before purchase 3. Go to Triple Whale → Attribution to see revenue under four models side by side: first-click, last-click, linear, and Triple Whale's own blended model 4. Connect your ad accounts (Meta, Google, TikTok) under Integrations — Triple Whale then compares your platform-reported ROAS against first-party-attributed ROAS 5. Use the Summary Dashboard for a daily view of true ROAS by channel

Option C: Polar Analytics (mid-market, more affordable)

1. Install Polar Analytics from the Shopify App Store 2. Connect ad accounts under Integrations 3. Polar provides first-party attribution with multiple models and a "blended ROAS" metric that accounts for all channels

---

WooCommerce

Using Metorik + GA4

1. Install Metorik (metorik.com) and connect to your WooCommerce store 2. Metorik captures UTM parameters on every order automatically — no additional plugin needed once configured 3. In Metorik, go to Reports → UTM to see orders and revenue by utm_source, utm_medium, and utm_campaign (last-click) 4. For multi-touch attribution, pair Metorik with Google Analytics 4:

• Install GA4 via Site Kit by Google or MonsterInsights plugin
• In GA4, go to Advertising → Attribution to configure the attribution model (last click, first click, linear, data-driven)
• GA4's data-driven attribution requires 400+ conversions per month to activate

Alternative: WooCommerce Google Analytics plugin (free)

• Tracks ecommerce events automatically; provides last-click channel attribution in GA4

---

BigCommerce

1. In BigCommerce, go to Storefront → Script Manager and add your attribution tool's pixel 2. Rockerbox (recommended): Installs via script; pulls BigCommerce orders via API to match to ad touchpoints; provides first-party attribution dashboard 3. Built-in BigCommerce analytics (Analytics → Marketing) shows traffic sources by last-click session; limited to top-level source/medium 4. For deeper analysis, connect BigCommerce to Google Analytics 4 via the official BigCommerce GA4 integration in Apps → Google

---

Custom / Headless

For headless storefronts, capture touchpoints server-side and store them with orders. Then model attribution in a data warehouse:

Step 1 — Capture UTM touchpoints on every visit:

// Client-side: capture and store UTM params in localStorage on every page load
function captureUTMTouchpoint() {
  const params = new URLSearchParams(window.location.search);
  if (!params.get('utm_source') && !params.get('gclid') && !params.get('fbclid')) return;

  const touchpoints = JSON.parse(localStorage.getItem('utm_touchpoints') || '[]');
  touchpoints.push({
    source: params.get('utm_source') ?? inferSource(document.referrer),
    medium: params.get('utm_medium') ?? 'organic',
    campaign: params.get('utm_campaign') ?? '(none)',
    touchedAt: new Date().toISOString(),
    landingPage: window.location.pathname,
  });
  // Keep last 10 touchpoints (30-day look-back)
  localStorage.setItem('utm_touchpoints', JSON.stringify(touchpoints.slice(-10)));
}

Step 2 — Attach touchpoints to the order at checkout:

// Send stored touchpoints with the order creation request
const touchpoints = JSON.parse(localStorage.getItem('utm_touchpoints') || '[]');
await createOrder({ ...orderData, marketingTouchpoints: touchpoints });

Step 3 — Store and model attribution in your data warehouse:

Export to BigQuery or Snowflake via Fivetran or Stitch, then build attribution models in dbt. Use a dbt package like dbt-attribution or write your own last-click, linear, and time-decay models against your orders + touchpoints tables.

Step 4: Compare attribution models and act on the data

Run these comparisons monthly to guide budget decisions:

Rule of thumb: If Meta claims $200K in attributed revenue and Google claims $180K, but your total revenue was $250K, you have significant attribution overlap. Use a first-party tool (Triple Whale, Rockerbox) to de-duplicate and get a realistic picture.

Best Practices

• Always compare multiple models — no single attribution model is "correct"; the comparison reveals which channels are over/under-credited in your current setup
• Use first-party data for attribution — iOS privacy changes have made third-party pixel attribution unreliable; server-side UTM tracking is now essential
• Standardize UTM naming conventions strictly — utm_source=google and utm_source=Google are treated as different channels; enforce lowercase and a controlled vocabulary
• Apply a 30-day look-back window — most e-commerce conversions occur within 30 days of first touch; use this as your standard window
• Benchmark total attributed revenue against actual revenue — the sum of first-party attributed revenue should equal total order revenue; a large gap indicates tracking gaps
• Report blended ROAS (total revenue / total ad spend) alongside channel ROAS — blended ROAS is harder to manipulate and gives a true picture of overall marketing efficiency

Common Pitfalls

Related Skills

• @influencer-tracking
• @affiliate-program
• @sales-reporting-dashboard
• @customer-analytics
• @ab-testing-ecommerce

Marketing Attribution Engine

Problem/Feature Description

The growth team at a mid-size e-commerce company has been running campaigns across Google Ads, Meta, email newsletters, and affiliate partners for the past quarter. The head of marketing is frustrated: the Google Ads dashboard claims $120K in revenue, the Meta dashboard claims $95K, and the email platform claims $80K — yet total store revenue was only $150K. Every channel is overclaiming because they each use last-click attribution internally.

The team needs a TypeScript attribution engine that processes historical conversion path data and produces channel-level revenue reports under different attribution models. The goal is to produce a side-by-side comparison showing how each channel's attributed revenue changes depending on the model used, so the team can have an informed discussion about budget allocation.

You have been provided with a JSON file containing sample conversion path data. Build the attribution logic, compute the results under each supported model, and produce a comparison report. The comparison should specifically highlight channels where different models disagree the most.

Output Specification

Produce a TypeScript implementation file attribution.ts that:

• Defines the relevant TypeScript types
• Implements all supported attribution model functions
• Implements a channel aggregation function
• Implements a multi-model comparison function that returns per-channel revenue under each model and identifies the largest discrepancies

Also produce a results.json file with the actual computed attribution output when run against the provided sample data. Run the code against the input data and save the output.

Include a run.sh script that shows how to execute the code.

Input Files

The following files are provided as inputs. Extract them before beginning.

=============== FILE: inputs/conversion_paths.json =============== [ { "orderId": "ord-001", "orderRevenue": 120.00, "touchpoints": [ { "source": "google", "medium": "cpc", "campaign": "brand", "touchedAt": "2026-02-10T09:00:00Z" }, { "source": "email", "medium": "newsletter", "campaign": "spring-sale", "touchedAt": "2026-02-15T14:00:00Z" }, { "source": "google", "medium": "cpc", "campaign": "brand", "touchedAt": "2026-02-18T10:00:00Z" } ] }, { "orderId": "ord-002", "orderRevenue": 85.50, "touchpoints": [ { "source": "meta", "medium": "paid_social", "campaign": "retargeting", "touchedAt": "2026-02-12T11:00:00Z" }, { "source": "meta", "medium": "paid_social", "campaign": "retargeting", "touchedAt": "2026-02-19T16:00:00Z" } ] }, { "orderId": "ord-003", "orderRevenue": 200.00, "touchpoints": [ { "source": "affiliate", "medium": "referral", "campaign": "partner-a", "touchedAt": "2026-01-20T08:00:00Z" }, { "source": "google", "medium": "cpc", "campaign": "non-brand", "touchedAt": "2026-02-01T12:00:00Z" }, { "source": "email", "medium": "newsletter", "campaign": "spring-sale", "touchedAt": "2026-02-10T09:00:00Z" }, { "source": "google", "medium": "cpc", "campaign": "brand", "touchedAt": "2026-02-19T15:00:00Z" } ] }, { "orderId": "ord-004", "orderRevenue": 55.00, "touchpoints": [] }, { "orderId": "ord-005", "orderRevenue": 310.00, "touchpoints": [ { "source": "influencer", "medium": "social", "campaign": "collab-jan", "touchedAt": "2026-02-05T10:00:00Z" }, { "source": "email", "medium": "newsletter", "campaign": "spring-sale", "touchedAt": "2026-02-14T08:00:00Z" }, { "source": "meta", "medium": "paid_social", "campaign": "retargeting", "touchedAt": "2026-02-19T20:00:00Z" } ] }, { "orderId": "ord-006", "orderRevenue": 75.00, "touchpoints": [ { "source": "google", "medium": "cpc", "campaign": "brand", "touchedAt": "2026-02-18T11:00:00Z" } ] }, { "orderId": "ord-007", "orderRevenue": 145.00, "touchpoints": [ { "source": "organic", "medium": "search", "campaign": "(none)", "touchedAt": "2026-02-08T14:00:00Z" }, { "source": "email", "medium": "newsletter", "campaign": "spring-sale", "touchedAt": "2026-02-16T10:00:00Z" }, { "source": "google", "medium": "cpc", "campaign": "brand", "touchedAt": "2026-02-20T09:00:00Z" } ] } ]

Data-Driven Channel Attribution for High-Volume Store

Problem/Feature Description

A large e-commerce retailer processes tens of thousands of orders per month and has accumulated over two years of server-side UTM touchpoint data. Their current rule-based attribution reports have a known problem: email consistently appears to get almost zero credit because most email-attributed journeys end with a branded Google search click, so last-click assigns everything to paid search. The analytics team suspects email is being dramatically undervalued in the budget planning process.

The team wants to move to a probabilistic model that infers each channel's true contribution to conversions based on historical path patterns rather than arbitrary rules. They need a TypeScript implementation of this model, along with a validation report that compares the probabilistic model's channel revenue allocations against the rule-based models to quantify the discrepancy.

The comparison report should also verify data integrity by checking whether the total attributed revenue matches actual revenue.

Output Specification

Produce a TypeScript file markov_attribution.ts containing: 1. A function to build a transition probability matrix from an array of conversion paths 2. A function to calculate each channel's contribution score based on its counterfactual impact on the overall conversion rate 3. A comparison function that runs both the probabilistic model and the rule-based models, producing a unified report

Produce a report.json file with the computed results run against the provided sample data, including the comparison between models and the revenue integrity check.

Include a README.md that documents when this model is appropriate to use vs. rule-based models.

Input Files

The following files are provided as inputs. Extract them before beginning.

=============== FILE: inputs/paths.json =============== [ { "orderId": "o-001", "orderRevenue": 150.00, "touchpoints": [ { "source": "google", "medium": "cpc", "campaign": "awareness", "touchedAt": "2026-01-15T08:00:00Z" }, { "source": "email", "medium": "newsletter", "campaign": "promo-feb", "touchedAt": "2026-02-01T10:00:00Z" }, { "source": "google", "medium": "cpc", "campaign": "brand", "touchedAt": "2026-02-10T14:00:00Z" } ]}, { "orderId": "o-002", "orderRevenue": 220.00, "touchpoints": [ { "source": "meta", "medium": "paid_social", "campaign": "prospecting", "touchedAt": "2026-01-20T09:00:00Z" }, { "source": "email", "medium": "newsletter", "campaign": "promo-feb", "touchedAt": "2026-02-03T11:00:00Z" }, { "source": "meta", "medium": "paid_social", "campaign": "retargeting", "touchedAt": "2026-02-12T16:00:00Z" } ]}, { "orderId": "o-003", "orderRevenue": 90.00, "touchpoints": [ { "source": "email", "medium": "newsletter", "campaign": "promo-feb", "touchedAt": "2026-02-02T08:30:00Z" }, { "source": "google", "medium": "cpc", "campaign": "brand", "touchedAt": "2026-02-09T11:00:00Z" } ]}, { "orderId": "o-004", "orderRevenue": 175.00, "touchpoints": [ { "source": "organic", "medium": "search", "campaign": "(none)", "touchedAt": "2026-01-25T13:00:00Z" }, { "source": "google", "medium": "cpc", "campaign": "brand", "touchedAt": "2026-02-11T10:00:00Z" } ]}, { "orderId": "o-005", "orderRevenue": 310.00, "touchpoints": [ { "source": "affiliate", "medium": "referral", "campaign": "partner-b", "touchedAt": "2026-01-18T07:00:00Z" }, { "source": "email", "medium": "newsletter", "campaign": "promo-feb", "touchedAt": "2026-02-04T09:00:00Z" }, { "source": "affiliate", "medium": "referral", "campaign": "partner-b", "touchedAt": "2026-02-13T15:00:00Z" } ]}, { "orderId": "o-006", "orderRevenue": 65.00, "touchpoints": [ { "source": "google", "medium": "cpc", "campaign": "brand", "touchedAt": "2026-02-14T10:00:00Z" } ]}, { "orderId": "o-007", "orderRevenue": 400.00, "touchpoints": [ { "source": "influencer", "medium": "social", "campaign": "macro-collab", "touchedAt": "2026-01-10T12:00:00Z" }, { "source": "meta", "medium": "paid_social", "campaign": "retargeting", "touchedAt": "2026-02-05T14:00:00Z" }, { "source": "email", "medium": "newsletter", "campaign": "promo-feb", "touchedAt": "2026-02-08T08:00:00Z" }, { "source": "google", "medium": "cpc", "campaign": "brand", "touchedAt": "2026-02-15T16:00:00Z" } ]}, { "orderId": "o-008", "orderRevenue": 130.00, "touchpoints": [ { "source": "email", "medium": "newsletter", "campaign": "promo-feb", "touchedAt": "2026-02-06T10:00:00Z" }, { "source": "google", "medium": "cpc", "campaign": "non-brand", "touchedAt": "2026-02-13T12:00:00Z" } ]}, { "orderId": "o-009", "orderRevenue": 88.00, "touchpoints": [ { "source": "meta", "medium": "paid_social", "campaign": "prospecting", "touchedAt": "2026-01-28T11:00:00Z" }, { "source": "google", "medium": "cpc", "campaign": "brand", "touchedAt": "2026-02-16T09:00:00Z" } ]}, { "orderId": "o-010", "orderRevenue": 55.00, "touchpoints": [] } ]

First-Party Marketing Touchpoint Tracking

Problem/Feature Description

A direct-to-consumer skincare brand has been relying on Meta Pixel and Google Tag to measure ad performance, but since iOS 14+ tracking changes gutted their conversion visibility, their reported ROAS from ad dashboards has dropped dramatically even though actual orders haven't. The engineering team has decided to move to a fully server-side, first-party attribution system based on UTM parameters.

They need two pieces of server-side TypeScript code: (1) a client-side function that fires on page load to record a marketing touchpoint when the user arrives via a tracked link, and (2) a server-side function that, when an order is placed, assembles the full conversion path for that order by fetching the user's historical touchpoints and persisting it.

The client-side capture must handle Google Ads, Meta Ads, and standard UTM-tagged campaigns. The conversion path builder must correctly resolve anonymous visitors (who may have browsed before logging in) to their orders.

Output Specification

Produce a TypeScript file tracking.ts containing: 1. The client-side touchpoint capture function 2. The TypeScript type/interface definition for a marketing touchpoint 3. The server-side conversion path builder function

Also produce a short design-notes.md explaining:

• How anonymous visitors are tracked and linked to orders
• What data retention approach is recommended and why

Input Files

No input files required — implement the tracking system from scratch.