MarketingAdvanced7 min read

Multi-Touch Attribution

Multi-touch attribution is the practice of distributing credit for a conversion across the multiple marketing touchpoints a customer interacted with on their journey — instead of giving 100% credit to the last click. Common models: Linear (equal credit), Time-Decay (more credit to recent touches), U-Shaped (40% to first, 40% to last, 20% to middle), and Data-Driven (algorithm assigns weights based on actual lift). The B2B average customer touches 7-13 marketing assets before buying. If you only credit the last touch, you'll underfund every channel except the demo-request form — and slowly bleed your top-of-funnel pipeline.

Also known asMTAAttribution ModelingMarketing AttributionTouchpoint Attribution

Challenge a friend Browse library

The Trap

The trap is treating attribution as a math problem when it's a political problem. Whoever 'owns' the last touch (usually paid search or sales) advocates for last-click attribution because it makes their ROI look great. Whoever owns top-of-funnel (content, brand, PR) advocates for first-touch. The right answer is rarely either extreme, but the conversation is usually decided by org politics, not data. The second trap: attribution accuracy is fundamentally limited by tracking — iOS 14, cookie deprecation, and dark social mean 30-60% of touches are now invisible.

What to Do

Pick a multi-touch model that REWARDS the channels you want to grow. If brand and content matter to you, use U-shaped or time-decay (not last-click). Run an incrementality test — pause one channel for 4 weeks in a controlled region and measure the actual conversion drop. The gap between attributed conversions and incremental conversions tells you how wrong your attribution model is. Most companies discover their last-click model overcredits paid search by 30-60% and undercredits brand by similar amounts.

Formula

Channel Credit = Σ (Touchpoint Weight × Conversion Value) across all touchpoints in the journey

In Practice

HubSpot famously moved from last-click attribution to a custom data-driven model around 2019. The shift revealed that their content/SEO investments — which last-click attribution had been valuing at ~15% of pipeline credit — were actually contributing closer to 40% of incremental revenue. Paid search had been overcredited (often the last click before a demo request, but rarely the source of awareness). Reallocating budget from paid to content following this insight reportedly drove a 30%+ improvement in pipeline efficiency over 18 months.

Pro Tips

01
Incrementality > attribution. Attribution tells you what the model thinks happened. Incrementality (controlled experiments where you turn channels on/off) tells you what actually happened. The two often disagree by 30-100%. If you have to choose, run incrementality tests on your top 3 channels and use attribution for everything else.
02
Triangulate three signals: (1) attribution model output, (2) self-reported attribution ('How did you hear about us?' on signup forms), (3) incrementality tests. When all three agree, you have signal. When they disagree, run more experiments before making budget decisions.
03
Brand and PR are systematically undercredited by every attribution model. They lift everything (organic search volume, paid CTR, direct traffic) but show up as 'Direct/Unknown' in the data. The smartest CMOs accept that 20-30% of their growth will be attribution-invisible and budget accordingly.

Myth vs Reality

Myth

“Data-driven attribution is the most accurate model.”

Reality

Data-driven models (Markov, Shapley, ML-based) are more sophisticated, but they're only as accurate as the data they ingest. With cookie deprecation, iOS 14, and dark social, even the best models are working with 40-70% of the actual journey. Sophistication doesn't fix missing data — it just produces more confident wrong answers.

Myth

“Last-click attribution is always wrong.”

Reality

For some conversion types (urgent intent like 'plumber near me'), last-click is roughly correct because the journey IS short. The trap is using last-click for considered B2B purchases with 7+ touchpoints, where it dramatically misrepresents the channel mix. Match the model to the journey, not to fashion.

Try it

Run the numbers.

Pressure-test the concept against your own knowledge — answer the challenge or try the live scenario.

🧪

Knowledge Check

A B2B prospect's journey: (1) Read a blog post via Google Organic, (2) Saw a LinkedIn ad, (3) Attended a webinar, (4) Got a sales email, (5) Clicked a Google Ad and filled out the demo form. They convert. Under last-click attribution, which channel gets credit?

Industry benchmarks

Is your number good?

Calibrate against real-world tiers. Use these ranges as targets — not absolutes.

Average Touchpoints to B2B Conversion

B2B SaaS with 90+ day sales cycle

Enterprise SaaS

13-27 touches

Mid-Market B2B

7-13 touches

SMB B2B

4-7 touches

B2C Considered

3-5 touches

B2C Impulse

1-2 touches

Source: Forrester / Gartner Buyer Journey Research

Real-world cases

Companies that lived this.

Verified narratives with the numbers that prove (or break) the concept.

🎯

HubSpot

2018-2021

success

HubSpot operated for years on last-click attribution, which heavily favored paid search and bottom-funnel channels. After moving to a data-driven multi-touch model, they discovered content marketing was contributing roughly 40% of incremental pipeline — more than 2x what last-click had credited. Paid search was overcredited because it tended to be the last touch on already-warm leads. Reallocating budget from paid acquisition to content investment over 18 months drove pipeline efficiency improvements and reduced blended CAC.

Last-Click Content Credit

~15% of pipeline

Multi-Touch Content Credit

~40% of pipeline

Budget Reallocation

Paid → Content

Pipeline Efficiency Lift

30%+ over 18 months

Attribution model choice quietly determines budget allocation. If you're running last-click on a B2B journey with 10+ touches, you're systematically underfunding the channels that build awareness and overfunding the ones that capture demand.

Source ↗

💬

Drift

2017-2020

success

Drift built their entire growth strategy around being unmeasurable in traditional attribution models. They invested heavily in podcasts, founder-led brand content (CEO David Cancel's media presence), and conversational marketing — all channels that show up as 'Direct' or 'Unknown' in attribution dashboards. While competitors optimized to measurable bottom-funnel ads, Drift built a brand moat that drove organic and direct traffic at scale. By the time the company sold for $1.2B, ~60% of pipeline came from sources their own attribution model couldn't track.

Direct/Brand Traffic %

~60% of pipeline

Outbound % of Pipeline

<10%

Acquisition

$1.2B (Vista Equity)

The attribution-invisible channels (brand, PR, podcasts, dark social) are often the most valuable. The companies that win long-term invest in channels they can't perfectly measure — because that's exactly where competitors won't compete.

Source ↗

Decision scenario

The Attribution Model Switch

You're CMO of a $40M ARR B2B SaaS. CFO has been using last-click for 3 years and uses it to allocate $8M annual marketing budget. Currently 65% goes to paid search/retargeting, 35% to content/brand. You believe content is undercredited and the budget mix is wrong.

Annual Marketing Budget

$8M

Paid Channels %

65%

Content/Brand %

35%

Current Model

Last-Click

Avg Touchpoints to Convert

Decision 1

You want to shift to multi-touch attribution and reallocate budget. The CFO won't approve any change without proof. You have one quarter and $200K experiment budget.

Build a sophisticated data-driven multi-touch attribution model in-house with the data team — present the new credit allocation as the proofReveal

After 4 months of engineering work, you present the new model. The CFO asks 'How do you know this model is more accurate than the old one?' You don't have a good answer — both are mathematical interpretations of the same incomplete data. The CFO points out that ~50% of touches are missing due to iOS 14 and cookie deprecation. Your new model is rejected as 'a different opinion, not better data.' You spent $200K and gained no political ground.

Budget Reallocation: NoneTime Lost: 4 months

Run an incrementality test: pause 50% of content spend in two regions for 60 days while holding paid search constant. Measure actual revenue delta vs attribution prediction.Reveal

After 60 days, paused-content regions show organic traffic down 31%, paid search CTR down 12%, and pipeline down 24% — much worse than last-click attribution predicted (which would have predicted 5-8% pipeline drop). The data is causal, not correlational. The CFO can't dispute it. You secure approval to shift budget toward content (now 50/50 instead of 65/35) and adopt time-decay attribution as the default model. Within 12 months, blended CAC is down 18% and pipeline is up 22%.

Budget Mix: 65/35 → 50/50CAC: Down 18%Pipeline: Up 22%

Related concepts