Tech & AIBeginnerPreview
Marketing Attribution
A practical, evidence-driven course that teaches you what marketing attribution really is, how the standard single- and multi-touch models assign credit, and how to choose, build, and sanity-check an attribution approach that does not lie to you in a post-cookie, walled-garden world.
Beginners, marketers, founders, and analysts who spend across multiple channels and want to know which ones truly drive conversions instead of trusting each platform's self-reported numbers.
Course content
Workbook & downloads
Put the course into practice — a printable workbook plus editable templates you can fill in and reuse.
Preview the workbook
This workbook turns the course into reps you can run on a real business. Each section mirrors one course module with hands-on exercises, fill-in worksheets, and checklists you apply to your own marketing. Pick one product or business you run that spends across at least two channels and carry it through every section. You will finish with a mapped customer journey, a clean UTM convention, a reconciled cross-tool conversion count, a chosen attribution model with a justified lookback window, and a planned incrementality test that will tell you which channels truly drive sales.
What Attribution Is and Why It Is So Hard
Build an accurate model of attribution — the multi-touch journey, the creator-versus-capturer lens, and the platform double-counting trap — before you trust a single number.
Worksheet: Map a Real Customer Journey
Pick two or three real customers (talk to them, or reconstruct from your data) and write out every marketing touchpoint they had before buying, in order. This makes the multi-touch problem concrete for your own business.
- Customer A — ordered list of touchpoints from first contact to purchase
- Customer B — ordered list of touchpoints from first contact to purchase
- Total touchpoints and total days from first touch to purchase (per customer)
- For each touchpoint, label its funnel stage: awareness / consideration / conversion
- For each touchpoint, label its role: demand creation or demand capture
Exercise: Expose the Platform Double-Counting Gap
Pull the real conversion total from your own back end, then pull each ad platform's self-reported conversions for the same window. Quantify the inflation so you stop trusting summed platform claims.
- Record the TRUE order/conversion count from your own system (Shopify / Stripe / CRM) for one week.
- Record each platform's claimed conversions for that same week (Meta, Google, TikTok, etc.).
- Sum the platform claims and compute the inflation: (summed claims − real total) / real total as a percentage.
- Which platform is most likely over-claiming via view-through, and how would you check?
Checklist: Attribution Reality Gut Check
- I have mapped at least one real journey and confirmed it is multi-touch, not a single click.
- I have labelled each channel as demand creation or demand capture, especially branded search as capture.
- I am anchoring on my own back-end order count, never on summed platform claims.
- I understand each ad platform grades its own homework with its own model and lookback.
- I can name at least two reasons (cross-device, dark funnel, view-through, time lag) my tracking is incomplete.
The Standard Attribution Models
Learn the six models by applying them to your own journey, see how the same path tells different stories, and choose a model and lookback window that fit your sales cycle.
Exercise: Run One Journey Through Four Models
Take one mapped journey from Section 1 and split 100% of the conversion credit across its touchpoints under four different models. Watch how the 'winning' channel changes with the model.
- First-touch: assign 100% to the first touchpoint. Which channel wins?
- Last-touch: assign 100% to the final touchpoint. Which channel wins?
- Linear: split credit evenly across all touchpoints (e.g. 4 touches = 25% each).
- Position-based (U-shaped): 40% to first, 40% to last, remaining 20% split across the middle. Which story feels most true, and why?
Worksheet: Measure Your Sales Cycle and Set the Lookback Window
The lookback window silently changes every result, so set it deliberately to cover your real buying timeline. Pull the data from your journeys and analytics.
- Average days from first touch to purchase (your real sales cycle)
- Typical number of touchpoints per converting journey
- Chosen lookback window in days (must comfortably exceed your cycle)
- Click-through window vs view-through window decision (include view-through? Y/N + why)
- Where this window is configured (GA4 attribution settings, ad platform settings)
Worksheet: Choose Your Attribution Model
Use the course decision framework to commit to one model for day-to-day reporting. Write down the reasoning so every later report is comparable and defensible.
- Sales cycle length and number of touchpoints (from above)
- Selected model (last-touch / linear / time-decay / position-based / data-driven)
- One-sentence justification tying the model to how your customers actually buy
- Do you have enough conversions for data-driven attribution? (rough monthly conversion count)
- Where the model is set and who else needs to use the same one for consistency
Checklist: Model Selection Audit
- I have run at least one real journey through multiple models and seen the winner change.
- I am not using a single-touch model as my default for a multi-channel budget.
- My lookback window comfortably covers my measured sales cycle.
- I have decided whether view-through credit is included and discounted it appropriately.
- My chosen model and window are documented so all reports are comparable.
Tracking the Data and Reading the Tools
Attribution is downstream of clean data. Build a UTM convention, work inside GA4's attribution reports, and reconcile the conflicting numbers across your tools.
Worksheet: Write Your UTM Naming Convention
Define a fixed, lowercase vocabulary for your UTM parameters so GA4 never splits one channel into three. This convention is the foundation every model downstream depends on.
- utm_source allowed values (e.g. google, facebook, newsletter, linkedin)
- utm_medium allowed values (fixed list: cpc, email, social, display, affiliate, organic)
- utm_campaign naming pattern (e.g. {season}_{promo}_{year} → spring_sale_2026)
- utm_content and utm_term usage rules (when to use each)
- Rules: always lowercase? never tag internal links? where the shared UTM spreadsheet/builder lives?
Exercise: Compare Models in GA4
Open GA4 → Advertising → Attribution → Model comparison. Put last-click next to data-driven and record how credit moves between your channels. This is where attribution clicks.
- Confirm which attribution model your GA4 property uses by default (data-driven is GA4's default).
- In Model comparison, record conversions for your top 5 channels under last-click vs data-driven.
- Which channels GAIN credit under data-driven (usually upper-funnel: organic social, display, video)?
- Which channels LOSE credit (usually branded search and direct), and what does that imply for budget?
Worksheet: Reconcile Your Conflicting Tools
List the conversion count each tool reports for the same campaign and week, then assign each tool to the decision it is actually best for. The goal is not to make them match.
- Back-end count (Shopify/Stripe/CRM) — use for: total revenue & the denominator
- GA4 count — use for: neutral cross-channel comparison
- Google Ads count — use for: optimising Google campaigns
- Meta count — use for: optimising Meta campaigns
- Decision: one model + one lookback + one dashboard you will anchor everything to
Checklist: Tracking & Tooling Quality Check
- My UTM convention is documented, lowercase, and uses a fixed medium vocabulary.
- No paid or email links go out untagged, and internal links are never UTM-tagged.
- I know which attribution model and conversion window my GA4 property uses.
- I have compared last-click vs data-driven in GA4 and seen the credit shift.
- Every report anchors on the back-end order count, and I know which tool answers which question.
Data-Driven Attribution and Proving Causality
Move beyond fixed rules to algorithmic attribution, account for privacy changes, and plan the incrementality test that proves which channels truly cause sales.
Exercise: Apply the Removal-Effect Question to Every Channel
For each major channel, ask the question at the heart of both data-driven attribution and incrementality: if this channel vanished tomorrow, how many conversions would actually disappear? This separates real drivers from passengers.
- List your top 5 channels by spend.
- For each, estimate honestly: if removed tomorrow, would conversions drop a lot, a little, or barely?
- Flag any channel (often branded search or retargeting) you suspect is taking credit it did not earn.
- Pick the one channel where you are least sure whether spend is incremental — this is your test candidate.
Worksheet: Privacy-Readiness Audit
Tracking signal is eroding from iOS ATT, third-party cookie loss, and consent rules. Record where you stand and what to fix so your measurement survives.
- First-party data you collect (logged-in accounts, email list size, back-end records)
- Server-side tracking / Conversions API status (Meta CAPI, Google server-side tagging): live? Y/N
- Consent Mode implemented for users who decline cookies? Y/N
- Estimated share of traffic that is iOS-app or unconsented (rough %)
- Top 2 gaps to close (e.g. install CAPI, grow first-party email capture)
Worksheet: Design One Incrementality Test
Plan a controlled experiment to prove (or debunk) whether your test-candidate channel actually causes sales. A holdout or geo test is the gold standard that attribution alone cannot give you.
- Channel under test and the hypothesis (e.g. branded search is mostly non-incremental)
- Method: conversion-lift holdout (Meta/Google) OR geo experiment (GeoLift/Robyn)
- Exposed group vs holdout/control group definition (and how regions are matched for geo)
- Test duration and primary metric (back-end conversions/revenue)
- Decision rule: what lift result would make you keep, cut, or shift this budget?
Checklist: Causality & Measurement-Stack Check
- I understand my attribution model is correlational and cannot prove a channel caused a sale.
- I have applied the removal-effect question and flagged a likely non-incremental channel.
- I have audited my privacy readiness (first-party data, server-side tracking, Consent Mode).
- I have a concrete incrementality or geo test planned with a clear decision rule.
- My long-term plan triangulates multi-touch attribution, incrementality, and marketing mix modelling.
Your Action Plan
- Choose one product or business that spends across at least two channels to carry through the whole plan.
- Map two or three real customer journeys touchpoint by touchpoint, labelling each as demand creation or demand capture.
- Pull your true order count from your own back end and compare it to summed platform claims to expose the double-counting gap.
- Run one real journey through first-touch, last-touch, linear, and position-based to see how the winning channel changes.
- Measure your real sales cycle, then set a lookback window that comfortably covers it and decide on view-through credit.
- Choose one attribution model with a written justification and apply it consistently across every report.
- Write and document a lowercase UTM naming convention with a fixed medium vocabulary, and stop sending untagged links.
- Open GA4 Model comparison, put last-click next to data-driven, and record which channels gain and lose credit.
- Reconcile your tools by assigning each one to the decision it is best for, anchoring everything to the back-end count.
- Audit your privacy readiness, then design and run one incrementality or geo test with a clear keep/cut/shift decision rule.
Pairs well with
Courses members commonly take alongside this one.
Flagship CoursePreview
Freelance Business Foundations: Position, Price, Sell, and Deliver High-Value Services
Freelancing · Beginner · 16h
Self-pacedPreview
Client GrowthPreview
Freelance Client Acquisition: Outreach, Leads, Referrals, and Deal Flow
Freelancing · Beginner · 15h 30m
Self-pacedPreview
Sales SystemPreview
Freelance Sales & Proposals: Discovery Calls, Scoping, Objections, and Closing
Freelancing · Intermediate · 16h
Self-pacedPreview