Back to Knowledge

Attribution Models

Method
Attribution

Definition

Answer the question: “Where did this customer come from?”

Related Knowledge

Attribution WindowsIncrementality Attribution ModelROAS (Return on Ad Spend)Marketing Mix Modeling (MMM)First/Last touch Position-based Attribution ModelMTA (Multi-Touch Attribution)

Common Attribution Approaches

  • MMM (Marketing Mix Modeling): Top-down statistical model that estimates the impact of each channel on overall results, often using aggregated data over time.
  • MTA (Multi-Touch Attribution): Bottom-up model that assigns credit across multiple touchpoints in a user journey.
  • Incrementality: Measures the lift caused by a channel or campaign versus what would have happened anyway (the counterfactual).

    • Growth Analyst Responsibilities

    Step 1: Audit and Choose a Source of Truth

    Your first job is to map out where the data is coming from and define what you trust.

    The conflict:

  • Facebook Ads: “We drove 100 sales.”
  • Google Analytics: “Facebook drove 40 sales.”
  • Internal database: “We had 85 total sales from all sources.”

    • You might discover that:

  • Facebook counts view-through conversions (people who saw but did not click an ad).
  • Google Analytics only counts click-through conversions.

    • The refinement:

    You define clear rules, for example:

  • “For official reports, we only count a Facebook sale if it happens within 1 day of viewing an ad or 7 days of clicking.”

    • Step 2: Implement Hard Tracking (The Technicals)

    Make sure every channel can be tracked reliably.

  • Ads: Ensure utm_source, utm_medium, and utm_campaign are on every single link.
  • Influencers / creators: Since they often drive views more than clicks, give them:
    • Vanity URLs (for example, company.com/mkbhd).
    • Unique promo codes (for example, MKBHD20).

    • The refinement:

  • “If a user uses the code MKBHD20, that YouTuber gets 100% of the credit, even if Google Analytics says the user came from ‘Direct’.”

    • Step 3: Solve the Dark Social Mystery

    “Dark social” is traffic that analytics tools cannot easily track:

  • Links shared in WhatsApp, iMessage, Slack, Discord, or a screenshot seen on TikTok followed by a Google search.

    • The problem: Your dashboard shows a huge spike in Direct or Organic Search traffic.

    The solution (HDYHAU): Add a short “How did you hear about us?” question to the signup or checkout flow.

    Example analysis:

  • Digital data: User came from "Google Search".
  • Survey response: “Saw a tweet from @levelsio.”

    • The refinement:

  • You conclude that a share of “Organic Search” is actually driven by Twitter.
  • You might decide: “We will attribute 30% of this organic search spike to the creator or channel that users self-report most often.”

    • Step 4: Weekly Defense and Reporting

    Every week, you review performance and defend how spend and credit are allocated.

  • You: “Affiliate traffic is up, but the conversion rate has dropped.”
  • Stakeholder: “Why?”
  • You: “Attribution analysis shows one coupon site ranking for ‘Company coupon’. Users already intend to buy, then search for a discount, click that affiliate link, and complete purchase. The affiliate is capturing credit without adding value.”

    • Decision:

  • Recommend banning that affiliate or reducing their commission. This protects budget and improves the accuracy of your attribution model.