Key Takeaways

  • Last-touch overcredits brand search by 20–40% (2024 GA4 data[1]), starving prospecting budget.
  • Data-driven attribution (DDA) requires 400 conversions / 28 days in GA4 (2024 threshold[3]) — not 1,000/month as older docs claim.
  • Time decay with 7-day half-life is optimal for 1–4 week sales cycles (2024 GA4 default[5]).
  • U-shaped (40/10/10/40) remains the gold standard for B2B lead gen with 2–8 week cycles (2024 Hypha/LeadSources benchmark[8][9]).
  • Switching models resets your baseline — change quarterly max; use GA4's Attribution Paths report (24–48h delay[1]) to verify.

The $10K Question: Which Channel Deserves Credit?

A customer buys a $10,000 DTC bundle. Their journey: TikTok Spark Ad (Day 1) → Meta retargeting carousel (Day 3) → Brand search click (Day 5) → Purchase.

Last-click (GA4 default[1]) gives 100% credit to brand search. Budget gets funneled into branded search. TikTok and Meta retargeting — the actual demand creators — get zero credit. Next quarter, prospecting budget gets cut. ROAS looks great; profit tanks.

This is the attribution trap. The model you choose doesn't just change reports — it rewrites your budget.

Quick Comparison: 6 Attribution Models at a Glance

Model First Touch % Middle % Last Touch % Complexity Best For (Sales Cycle)
Last-Touch (Last Non-Direct Click) 0% 0% 100% ⭐ < 7 days, 1–2 touches, impulse e-comm
First-Touch 100% 0% 0% ⭐ Brand awareness measurement, top-funnel audit
Linear 25% 25% (each) 25% ⭐⭐ Equal-touch journeys, simple multi-channel audit
Time Decay (7-day half-life) ~9% ~30% (distributed) ~61% ⭐⭐⭐ 1–4 week cycles, DTC, lead gen nurture
Position-Based (U-Shaped) 40% 10% (each middle) 40% ⭐⭐⭐ B2B lead gen, 2–8 week cycles, clear first/last
Data-Driven (GA4 DDA / Shapley) Algorithmic Algorithmic Algorithmic ⭐⭐⭐⭐

Quick calc: Plug your last-touch revenue into the ROAS Calculator — then re-run with DDA credit splits to see the real ROAS gap.

Deep Dive: How Each Model Distributes Credit

1. Last-Touch (Last Non-Direct Click) — GA4 Default

Logic: 100% credit to the last non-direct click before conversion. Direct visits are ignored; credit rolls back to the last campaign touch.

4-touch example (2024 GA4 default[1]):
TikTok (Day 1) → Meta Retargeting (Day 3) → Email (Day 5) → Brand Search (Day 7) → Purchase
Credit: TikTok 0%, Meta 0%, Email 0%, Brand Search 100%

Pros Cons
Default in GA4, Google Ads, Meta Ads Manager Overcredits brand/navigational search by 20–40% (2024 GA4 data[1])
Simple, stakeholder-friendly Starves prospecting/awareness budget
Accurate for <7-day, 1–2 touch journeys Fails for any multi-touch journey

When to use: Pure impulse e-comm (<7 day cycle, 1–2 touches). Never for considered purchases.

ROAS Calculator — calculate ROAS with last-touch data

2. First-Touch

Logic: 100% credit to the first touchpoint in the journey.

4-touch example:
TikTok (Day 1) → Meta Retargeting (Day 3) → Email (Day 5) → Brand Search (Day 7) → Purchase
Credit: TikTok 100%, Meta 0%, Email 0%, Brand Search 0%

Pros Cons
Reveals true demand-gen channels Ignores nurture & close — overcredits top-funnel
Simple to implement & explain Misleads budget if used alone for allocation
Great for channel-audit: "What started the journey?" Useless for ROAS optimization alone

When to use: Channel-mix audit for awareness spend. Pair with CPA Calculator to measure acquisition cost by first touch.

CPA Calculator — measure acquisition cost by first touch

3. Linear

Logic: Equal credit to every touchpoint in the path.

4-touch example (25% each):
TikTok 25% → Meta Retargeting 25% → Email 25% → Brand Search 25%

Pros Cons
Zero channel bias First blog visit = demo request = equal credit (unrealistic)
Easy to calculate & explain Overcredits low-value middle touches
Good baseline for model comparison Rarely reflects true influence

When to use: Quick audit when you have no prior model baseline. Compare against First/Last to see the "credit gap."

4. Time Decay (7-Day Half-Life)

Logic: Credit decays exponentially by time. Touchpoints 7 days before conversion get ~50% of the credit of a touchpoint at conversion (2024 GA4 default half-life[5]).

4-touch example (7-day half-life, 2024 GA4 math[5]):
TikTok (Day 1, 6 days out) → Meta Retargeting (Day 3, 4 days out) → Email (Day 5, 2 days out) → Brand Search (Day 7, 0 days out) → Purchase
Credit: TikTok ~9%, Meta ~22%, Email ~38%, Brand Search ~31%

Pros Cons
Matches psychology: recency = influence No universal half-life; GA4 uses 7-day default[5]
Ideal for 1–4 week cycles Aggressive decay undervalues awareness
Balances simplicity & accuracy Requires timestamped touch data

When to use: DTC e-comm (1–2 weeks), lead gen nurture (2–4 weeks). Best default for most performance marketers.

5. Position-Based (U-Shaped)

Logic: First touch 40%, last touch 40%, middle touches split remaining 20%.

4-touch example (classic 40/10/10/40):
TikTok 40% → Meta Retargeting 10% → Email 10% → Brand Search 40%

Pros Cons
Honors both demand creation & capture Middle nurture touches severely undercredited (10% each)
B2B gold standard (2024 Hypha/LeadSources[8][9]) Assumes first & last are always most important
Stakeholder-friendly visual Fails for journeys with >4 touches (middle gets <5% each)

When to use: B2B lead gen, agency client reporting, 2–8 week cycles with clear first/last milestones.

6. Data-Driven Attribution (GA4 DDA / Shapley Value)

Logic: Shapley game theory — computes marginal contribution of each touchpoint across all observed paths. Credit = average lift when touchpoint is present vs. absent (2024 GA4 implementation[5][16]).

Thresholds (2024 verified[3][4]):

  • GA4: 400 conversions / 28 days (property-level)
  • Google Ads: 600 conversions / 30 days (account-level[4])

4-touch example (hypothetical DDA output):
TikTok 28% → Meta Retargeting 12% → Email 15% → Brand Search 45%

Pros Cons
Most accurate — reflects your data Black box; hard to explain to stakeholders
Adapts automatically to behavior shifts Requires 400+/28d (GA4) or 600+/30d (Ads)
No arbitrary rules (half-life, 40/20/40) Not available in all tools (GA4, Ads, HubSpot Enterprise[10])

When to use: You hit thresholds and have complex journeys (5+ touches, cross-device). Default choice for mature accounts.

Visual: Credit Distribution Comparison

How to Choose: Decision Framework

By Sales Cycle Length

Sales Cycle Recommended Model Why
< 7 days (impulse e-comm) Last-Touch / Time Decay Few touches; recency = influence
1–2 weeks (DTC, low-consideration) Time Decay (7-day half-life) Recency-weighted; simple enough for daily ops
2–4 weeks (lead gen, mid-market) Time Decay / U-Shaped Nurture matters; first & last both critical
2–8 weeks (B2B, agency) U-Shaped / Linear Clear first-touch (lead source) + last-touch (demo/meeting)
2–6 months (Enterprise SaaS) Data-Driven / W-Shaped* Complex, multi-stakeholder; needs algorithmic credit
10k+ conv/mo (High-volume brand) Data-Driven (GA4 + Ads) Threshold met; model self-optimizes

W-Shaped (30/10/10/10/30) adds "opportunity created" milestone — available in HubSpot Enterprise, Hypha, LeadSources, Factors.ai, 6sense[8][9][10][12].

By Business Type (Ads Calculator Audience Focus)

Business Type Typical Cycle Recommended Model Calculator to Use
E-comm DTC < 2 weeks Time Decay / DDA ROAS Calculator
Lead Gen / Agency 2–8 weeks U-Shaped / Linear CPA Calculator
B2B SaaS 2–6 months W-Shaped / DDA Campaign Funnel Calculator
High-Volume Brand 10k+ conv/mo Data-Driven Break-Even ROAS / ROI/LTV Calculator

GA4 Setup: Switch Attribution in 3 Steps

Step 1 — Open Attribution Settings
Admin ▸ (Property column) ▸ Attribution Settings ▸ Reporting Attribution Model

[1]

Step 2 — Select Model
Choose Data-driven (if eligible) or Position-based / Time decay / Linear / First click / Last click.
Lookback window: Set to 90 days for B2B/long-cycle; 30 days for e-comm[5].

Step 3 — Verify (wait 24–48h)
Reports ▸ Advertising ▸ Attribution ▸ Attribution Paths.
Confirm credit redistributes per new model. Compare "Model Comparison" tab[1].

Pro tip: Run Model Comparison for 2 weeks before switching primary model. Export CSV → plug revenue splits into Campaign Funnel Calculator to model budget impact.

The Attribution Trap: 5 Mistakes That Waste Ad Budget

  1. Defaulting to Last-Touch → Overcredits brand search by 20–40% (2024 GA4[1]), starves prospecting. Fix: Switch to Time Decay or DDA immediately if eligible.
  2. Wrong Lookback Window → 30-day window for 90-day B2B cycle = 60% of journey invisible (2024 Think with Google[2]). Fix: Match window to sales cycle + 30 days buffer.
  3. Ignoring Assisted Conversions → Blog/email assist 60% of conversions but get 0% credit in last-touch (2024 Factors.ai[10]). Fix: Check Advertising ▸ Attribution ▸ Assisted Conversions monthly.
  4. Cross-Device Blindness → Mobile → Desktop → Tablet = 3 users in GA4 without User-ID/Google Signals[6]. Fix: Enable Google Signals + User-ID; server-side GTM (WeltPixel[6]).
  5. Comparing Models Wrong → CPA changes per model; comparing last-touch CPA to DDA CPA = apples-to-oranges (2024 LeadsRx[11]). Fix: Pick one primary model/quarter. Use others for insight only.

FAQ (Schema-Ready)

Which attribution model gives the highest ROAS?

Last-touch inflates ROAS by crediting only the final click (often brand search). Data-driven gives the *true* ROAS — typically 15–30% lower but actionable. Use the ROAS Calculator with DDA splits to see the real number.

Can I use different attribution models for different channels?

Don't. Pick one primary model for budget allocation. Mixing models makes cross-channel CPA/ROAS incomparable (2024 Digital Applied[11]). Use secondary models only for diagnostic reports (e.g., "First-Touch for TikTok audit").

How many conversions do I need for Data-Driven Attribution?

GA4: 400 conversions / 28 days (property-level[3]). Google Ads: 600 conversions / 30 days (account-level[4]). Older docs cite 1,000+/month — outdated as of 2024.

Is last-touch attribution ever okay?

Yes: <7-day sales cycle, 1–2 touchpoints, impulse e-comm (flash sales, low-AOV DTC). For anything considered (>2 weeks, >3 touches), last-touch misallocates 20–40% of budget (2024 GA4[1]).

What about iOS14+/cookieless attribution?

Use DDA (models missing paths), Server-side GTM (WeltPixel[6]), Consent Mode v2 (GA4[1]), and Geo experiments (Andava Digital[13]) for incrementality validation. Don't rely on client-side cookies alone.

How often should I change attribution models?

Quarterly max. Changing models resets your baseline — YoY and QoQ comparisons break. Pick one, lock for 90 days, evaluate in Model Comparison, then decide (2024 Optimize Smart[5]).

Related Calculators (CTA Box)

Related Articles

Sources & References

  1. Google Analytics Help: Attribution Overview — https://support.google.com/analytics (retrieved 2026-07-18)
  2. Google Ads Help: Attribution Models (MB Adv Agency) — https://support.google.com/google-ads (retrieved 2026-07-18)
  3. Seresa: GA4 Data-Driven Attribution Thresholds — https://seresa.io/blog/ga4-data-driven-attribution-thresholds/ (retrieved 2026-07-18)
  4. 1ClickReport: GA4 Conversion Attribution Analysis — https://1clickreport.com/blog/ga4-conversion-attribution-analysis/ (retrieved 2026-07-18)
  5. Optimize Smart: GA4 Attribution Models Explained — https://www.optimizesmart.com/blog/ga4-attribution-models-explained-how-to-choose-the-right-one/ (retrieved 2026-07-18)
  6. WeltPixel: Server-Side Tracking Impact — https://weltpixel.com/blog/server-side-tracking-ga4/ (retrieved 2026-07-18)
  7. Growth Method: Data-Driven Attribution & Time Decay — https://growthmethod.com/data-driven-attribution/ (retrieved 2026-07-18)
  8. Hypha HubSpot Development: W-Shaped & Full-Path Attribution — https://hypha.com/hubspot-development/w-shaped-attribution/ (retrieved 2026-07-18)
  9. LeadSources.io: W-Shaped Attribution Breakdown — https://leadsources.io/w-shaped-attribution/ (retrieved 2026-07-18)
  10. Factors.ai: Marketing Attribution Models Guide — https://www.factors.ai/blog/marketing-attribution-models/ (retrieved 2026-07-18)
  11. Digital Applied: Server-Side GTM & Signal Loss — https://digitalapplied.com/blog/server-side-gtm-cookieless/ (retrieved 2026-07-18)
  12. The Paradigm Shift: Cookieless Attribution Research — https://theparadigmshift.io/research/cookieless-attribution/ (retrieved 2026-07-18)
  13. Andava Digital: GeoLift Incrementality Testing — https://andavadigital.com/geolift-incrementality-testing/ (retrieved 2026-07-18)
  14. Stella: Incrementality Benchmarks — https://stella.ai/blog/incrementality-benchmarks/ (retrieved 2026-07-18)
  15. Triple Whale: Triangulation (MMM + MTA + Incrementality) — https://www.triplewhale.com/blog/triangulation-mmm-mta-incrementality (retrieved 2026-07-18)
  16. PROANALYTICS: Shapley Value & Markov Chains Attribution — https://proanalytics.ro/shapley-value-markov-attribution/ (retrieved 2026-07-18)
  17. Whitehat SEO: AI Search Traffic Attribution (ChatGPT) — https://whitehatseo.co.uk/ai-search-traffic-attribution/ (retrieved 2026-07-18)
  18. Koehn AI: Shapley vs Markov Attribution Comparison — https://koehn.ai/shapley-vs-markov-attribution/ (retrieved 2026-07-18)