Marketing Attribution

The practice of determining which marketing channels and touchpoints actually drive a customer’s decision to purchase.

---

What is it?

A customer sees a social media ad on Monday, reads a blog post on Wednesday, receives an email on Friday, and buys on Saturday after clicking a Google search result. Which of these touchpoints “caused” the purchase? That is the attribution problem --- and it is one of the hardest measurement challenges in marketing.¹

Marketing attribution is the practice of assigning credit for a conversion (usually a sale) to the marketing channels and touchpoints that influenced it. The goal is simple: understand what is working so you can spend more on it, and understand what is not working so you can stop wasting money on it. But the execution is anything but simple, because customers rarely interact with just one channel before buying.²

The default in most analytics platforms is last-click attribution --- giving 100% of the credit to the last touchpoint before the purchase. In the example above, Google search gets all the credit. The social ad, the blog post, and the email get nothing. This is convenient to measure but systematically wrong: it overvalues end-of-funnel channels (search, retargeting) and undervalues top-of-funnel channels (brand awareness, content) that initiated the journey.³

The alternative is multi-touch attribution (MTA), which distributes credit across multiple touchpoints. Various models exist --- linear (equal credit), position-based (more credit to first and last touch), time-decay (more credit to recent touches), and algorithmic/ML-based (credit weighted by statistical contribution). Each model has trade-offs, but all are more accurate than last-click for understanding the full customer journey.⁴

In plain terms

Attribution is about answering “what made the customer buy?” honestly. The simplest answer (the last thing they clicked) is almost always misleading. The real answer involves understanding the full chain of interactions that led to the purchase --- and giving appropriate credit to each one.

---

At a glance

Attribution models compared (click to expand)

graph LR
    subgraph LastClick["Last-Click"]
        LC1["Ad: 0%"] --> LC2["Email: 0%"] --> LC3["Search: 100%"]
    end
    subgraph Linear["Linear"]
        LN1["Ad: 33%"] --> LN2["Email: 33%"] --> LN3["Search: 33%"]
    end
    subgraph PositionBased["Position-Based"]
        PB1["Ad: 40%"] --> PB2["Email: 20%"] --> PB3["Search: 40%"]
    end
    subgraph TimeDecay["Time-Decay"]
        TD1["Ad: 15%"] --> TD2["Email: 30%"] --> TD3["Search: 55%"]
    end

Key: The same three-touchpoint journey, credited four different ways. Last-click ignores everything except the final interaction. Multi-touch models distribute credit more realistically.

---

How does it work?

Last-click attribution: the default

Last-click gives 100% of the conversion credit to the final touchpoint before purchase. It is the default in most analytics tools because it is easy to implement: one click, one conversion, one channel credited.³

The problem is that last-click systematically distorts reality. Top-of-funnel activities (brand advertising, content marketing, social media awareness) that introduce customers to the brand receive zero credit, because they rarely appear as the last click. End-of-funnel activities (search ads, retargeting, email reminders) receive disproportionate credit because they often appear right before the purchase.³

This creates a dangerous feedback loop: marketers see that search and retargeting “drive” the most conversions, so they shift budget toward those channels. But those channels only work because earlier touchpoints created the demand. Cut the top-of-funnel spend, and eventually the bottom-of-funnel channels stop performing too --- but by then, the cause is invisible.¹

Think of it like...

Last-click attribution is like crediting only the striker who scored the goal. The midfielder’s through pass, the defender’s clearance that started the counter-attack, and the goalkeeper’s throw that reached the halfway line --- none of them get credit. Over time, the manager might bench the midfield because they “don’t score goals.” The team collapses, and nobody understands why.

Multi-touch attribution models

Multi-touch attribution (MTA) distributes credit across all touchpoints in the customer journey. The main models are:⁴

Model	How it works	Best for
Linear	Equal credit to every touchpoint	Simple overview of channel involvement
Position-based (U-shaped)	40% to first touch, 40% to last touch, 20% split among middle	Businesses that value both awareness and conversion
Time-decay	More credit to recent touchpoints, less to earlier ones	Short sales cycles where recency matters most
Algorithmic / ML-based	Statistical model determines each touchpoint’s contribution	High-volume businesses with strong data infrastructure

No model is perfectly accurate. The value of MTA is not precision --- it is directional correctness. Even a rough multi-touch model gives a far more honest picture of channel performance than last-click.⁴

Think of it like...

Multi-touch attribution is like dividing credit for a successful recipe. The ingredients (first touch), the technique (middle touches), and the final plating (last touch) all matter. Different models just disagree about how much credit each step deserves. But all of them are better than crediting only the oven.

The data infrastructure requirement

Attribution is only as good as the data behind it. To run any attribution model, a business needs:²

1. Cross-channel tracking --- the ability to follow a single user across multiple platforms and devices (social media, search, email, website)
2. Identity resolution --- connecting anonymous sessions to known customers (this is increasingly difficult with cookie deprecation and privacy regulations)
3. Event logging --- recording every meaningful interaction (ad impression, click, email open, page view, purchase) with timestamps
4. A sufficiently large data set --- statistical models need volume to produce reliable results

For small businesses with limited traffic, sophisticated attribution models may not be worthwhile. Simple approaches --- asking “how did you hear about us?” at checkout, using unique discount codes per channel, or tracking UTM parameters --- can provide directional insights without the infrastructure investment.²

Think of it like...

The data infrastructure is like the stadium’s camera system. With one camera, you only see the goal (last-click). With cameras covering the entire pitch (cross-channel tracking), you can rewind and see the full sequence of play. But installing that camera system has a cost, and for a Sunday league match, it may not be worth it.

---

Why do we use it?

Key reasons

1. It prevents systematic misallocation of marketing budget. Last-click attribution funnels money toward end-of-funnel channels and starves the top-of-funnel channels that create demand. Proper attribution reveals the full chain of influence, enabling smarter budget decisions.¹

2. It produces accurate per-channel CAC. Customer-acquisition-cost can only be measured per channel if you know which channels contributed to each conversion. Without attribution, per-channel CAC is guesswork.³

3. It reveals the true performance of every marketing investment. Brand campaigns, content marketing, and social media often look like they “don’t convert” under last-click. Multi-touch attribution shows their actual contribution to the purchase path, justifying continued investment in channels that work but are invisible under simple models.⁴

---

When do we use it?

• When marketing spend is spread across multiple channels and you need to know which are performing
• When deciding whether to cut or increase budget for a specific channel
• When top-of-funnel metrics (impressions, clicks, engagement) look strong but bottom-of-funnel conversions are attributed elsewhere
• When you suspect that last-click reporting is distorting your understanding of channel performance
• When building a data infrastructure for marketing analytics and need to define what to track

Rule of thumb

If you are spending money on more than one marketing channel, you have an attribution problem. The only question is whether you are measuring it or ignoring it.

---

How can I think about it?

The football assist

Attribution is like crediting a goal in football. Last-click attribution gives all the credit to the striker --- the last player to touch the ball. But the goal would not have happened without the midfielder’s pass, the winger’s run that pulled the defender out of position, and the defensive clearance that started the counter-attack.

Position-based attribution is like giving credit to both the player who started the move and the player who finished it, with smaller credit to those in between. Time-decay is like giving more credit to the pass before the goal and less to the clearance that started the move.

No model is perfect. But crediting only the striker --- which is what last-click does --- will eventually lead to under-investment in midfield, defence, and build-up play. The team stops creating chances, and the striker has nothing to finish.

The recipe

Think of a customer’s journey to purchase as a recipe, and attribution as the question: what made this cake good?

Last-click says “the oven” --- it was the last thing that touched the cake. But the quality of the ingredients (first touchpoint), the mixing technique (middle touchpoints), and the oven temperature (last touchpoint) all contributed. Crediting only the oven would lead you to buy a more expensive oven while ignoring bad ingredients.

Multi-touch attribution acknowledges that every step mattered. Linear attribution says they all mattered equally. Position-based says the ingredients and the oven mattered most. Algorithmic attribution analyses hundreds of cakes to determine statistically which steps actually predict quality. None is perfect, but all are better than “it was the oven.”

---

Concepts to explore next

Concept	What it covers	Status
customer-acquisition-cost	The total cost to gain one customer, which requires attribution to measure per channel	complete
conversion-rate-optimisation	Improving conversion rates, which interacts with how credit is distributed across the funnel	complete
channel-mix-strategy	Diversifying across acquisition channels, informed by attribution data	stub

Some cards don't exist yet

A broken link is a placeholder for future learning, not an error.

---

Check your understanding

Test yourself (click to expand)

1. Explain why last-click attribution systematically overvalues end-of-funnel channels.
2. Name four multi-touch attribution models and describe how each distributes credit differently.
3. Distinguish between last-click and position-based attribution. In what business context would you prefer each?
4. Interpret this scenario: a business using last-click attribution sees that paid search drives 70% of conversions and social media drives 5%. They cut social media budget by 80%. Three months later, paid search conversions decline by 30%. What happened?
5. Connect attribution to customer-acquisition-cost. Why is per-channel CAC unreliable without a multi-touch attribution model?

---

Where this concept fits

Position in the knowledge graph

graph TD
    MS[Marketing & Sales] --> MA[Marketing Attribution]
    MS --> CAC[Customer Acquisition Cost]
    MS --> CRO[Conversion Rate Optimisation]
    MS --> CMS[Channel Mix Strategy]
    CMS --> MA
    MA --> CAC

    style MA fill:#4a9ede,color:#fff

Related concepts:

• customer-acquisition-cost --- attribution is the prerequisite for accurate per-channel CAC; without it, you cannot know what each channel truly costs
• conversion-rate-optimisation --- CRO improvements interact with attribution by changing where in the funnel conversions happen

---

Sources

---

Further reading

Resources

• Methods and Models: A Guide to Multi-Touch Attribution (Nielsen) --- The most authoritative overview of MTA models, from rule-based to algorithmic
• Marketing Attribution: What It Is, Why It Matters (Adobe Marketo) --- Accessible introduction covering the business case and common models
• The Complete Guide to Marketing Attribution (Segment) --- Practical guide focused on the data infrastructure and implementation side of attribution
• Multi-Touch Attribution and the Impact of Privacy Changes (AppsFlyer) --- How cookie deprecation and privacy regulations are reshaping attribution methodology

Footnotes

1. Marketo. (2024). Marketing Attribution: What It Is, Why It Matters. Adobe Marketo Blog. ↩ ↩² ↩³

2. Segment. (2025). The Complete Guide to Marketing Attribution. Twilio Segment Blog. ↩ ↩² ↩³

3. Dalessandro, B. et al. (2012). Causally Motivated Attribution for Online Advertising. Proceedings of ADKDD, ACM. ↩ ↩² ↩³ ↩⁴

4. Nielsen. (2019). Methods and Models: A Guide to Multi-Touch Attribution. Nielsen Insights. ↩ ↩² ↩³ ↩⁴