What this is about
This article is part of our Marketing & Performance topic cluster — an overview of all articles on online marketing, attribution, omnichannel and video content for Swiss SMEs.
Multi-touch attribution models are used in campaign marketing analysis to assign value to the different touchpoints along a customer journey. There is a range of models — and each one has its own logic. For an SME with a tight marketing budget, picking the right model is decisive: without clear attribution, you allocate budget based on the wrong signals. To make the numbers reliable, a clean marketing reporting setup belongs in the foundation.
Theoretical grounding — where the attribution models come from
Multi-touch attribution has been part of digital marketing theory since the late 2000s. Chaffey, Hemphill and Edmundson-Bird systematise the models in Digital Business and E-Commerce Management (7th Edition, 2019) within the framework of e-commerce performance tracking, and emphasise that the attribution choice directly influences marketing-mix optimisation — a point that is often overlooked in SME reality.
Concretely: if you measure only last-touch, you systematically overestimate performance ads and underestimate brand and content investment. If you measure only first-touch, you make the opposite mistake. The honest answer almost always lies in a data-driven or W-shaped model — and that choice is part of the Control phase in the SOSTAC planning framework.
The most important attribution models
1. First-touch attribution
Logic: Full credit for a purchase goes to the first touchpoint the customer had contact with.
When it makes sense: When you want to understand which marketing channels drive initial attention — top-of-funnel analysis.
Advantages:
- Easy to implement and understand
- Highlights the initial touchpoints
- Valuable insights if your goal is to track initial interest
Disadvantages:
- Oversimplifies the customer journey
- Does not account for the contribution of other marketing activities
Example: A customer sees a Facebook ad for a speaker, clicks, but does not buy. Later she receives an email, clicks on it, and buys. First-touch: Full credit goes to the Facebook ad.
2. Last-touch attribution
Logic: Full credit goes to the last touchpoint before the purchase.
When it makes sense: For simple, linear sales funnels with a clear conversion path.
Advantages:
- Simple implementation
- Clear conversion mapping
- Good for direct-response campaigns
Disadvantages:
- Ignores all earlier touchpoints
- Over-credits bottom-of-funnel channels (e.g. brand search)
- Distorts investment decisions — awareness channels lose credit
Example: Same scenario as above — last-touch: Full credit goes to the email. The Facebook ad gets no credit.
3. Linear attribution
Logic: Credit is distributed evenly across all touchpoints.
When it makes sense: When you assume every touchpoint contributed equally to the conversion.
Advantages:
- Considers the entire customer journey
- Easy to explain
Disadvantages:
- Reality is rarely that uniform
- Overrates less influential touchpoints
- Underrates decisive touchpoints
4. Time-decay attribution
Logic: Touchpoints closer to the moment of conversion receive more credit than early touchpoints.
When it makes sense: For longer sales cycles with multiple touchpoints, where the late stage is more decisive.
Advantages:
- Weights more recent, purchase-relevant touchpoints higher
- Realistic for B2B with longer sales cycles
Disadvantages:
- Still underrates initial awareness
- Configuration effort (decay rate)
5. Position-based / U-shaped attribution
Logic: First and last touchpoint each receive 40%, the middle touchpoints share the remaining 20%.
When it makes sense: When you want to weight both awareness and conversion honestly.
Advantages:
- Balances first- and last-touch
- Practical default model for SMEs
Disadvantages:
- Static — does not fit every funnel
- Middle touchpoints are underrated
6. Data-driven attribution
Logic: An algorithm learns from your data which touchpoints actually contributed to the conversion. Individual per customer journey.
When it makes sense: With enough data (at least 600 conversions / 30 days in Google Ads) and complex multi-channel journeys.
Advantages:
- Data-based weighting per touchpoint
- Adaptive
- Realistic for complex funnels
Disadvantages:
- Black box (hard to explain)
- Data requirements often too high for small SMEs
- Setup complexity
Which attribution model for which SME?
| SME type | Recommended model |
|---|---|
| Small SME, little data | Last-touch + qualitative insights |
| SME with awareness focus | First-touch or position-based |
| B2B with longer sales cycles | Time-decay |
| Mature SME with good tracking | Position-based (U-shaped) |
| Data-driven SME | Data-driven (Google Ads / GA4) |
AI-driven attribution modelling
A concrete example of how data-driven attribution played out for a two-sided platform is documented in our art24 marketplace mandate.
In 2026, AI is fundamentally reshaping multi-touch attribution:
- GA4 data-driven attribution: Google offers data-driven attribution directly in GA4 — no external tools, based on machine learning.
- AI agents for cross-device attribution: Tracking customer journeys across multiple devices remains difficult — AI agents help link cross-device touchpoints by behaviour patterns.
- Predictive attribution: Instead of just attributing retrospectively, AI models predict prospectively which investment will have which impact.
- Privacy-compliant attribution: With the disappearance of 3rd-party cookies and ever-stricter privacy rules, privacy-by-design attribution solutions (e.g. marketing mix modelling, conversion modelling) become more important. AI is the central lever here.
Tool recommendations for SMEs in 2026:
- GA4 with data-driven attribution (free, if you have enough data)
- Triple Whale, Northbeam, Rockerbox for e-commerce attribution
- Custom setup via SQL/BigQuery + ChatGPT Code Interpreter for tailored SME solutions
Practical tip: look at multiple models in parallel
Pros do not use just one attribution model — they compare several. Per channel, different models show different ROI — and the truth usually lies somewhere in between. It helps to mirror each model against the STDC customer journey to see which phase it over- or underrates.
Example setup: GA4 reports with 3 attribution models in parallel (first-touch, last-touch, data-driven). Compare per channel. Channels that underperform in all 3 models are clearly droppable. Channels that shine in only one model deserve critical questioning.
Q&A — multi-touch attribution for SMEs
Which attribution model is the best? There is no "best" model — only fitting ones. It depends on your funnel, data volume, and strategic questions. Position-based is a good default for SMEs. Ideally, the model choice is embedded in larger digital transformation programmes that bring data and strategy together.
How much data do I need for data-driven attribution? Google Ads data-driven attribution: from 600 conversions / 30 days and 3,000 ad interactions / 30 days. GA4 attribution is less restrictive but becomes unreliable below roughly 100 conversions / month.
How does attribution differ from marketing mix modelling (MMM)? Attribution works at user level (which touchpoints did this customer see). MMM works at aggregate level (which channel produced how much revenue in Q3). MMM gains relevance as cookies disappear. In practical execution, a marketing data agent increasingly handles the ongoing analysis of both levels.
Can I switch attribution models without losing data? The data stays. But switching changes the interpretation — channels that shone under last-touch suddenly look weaker under position-based. Set up a clean communication process with the team. What an external view of setup and channels looks like is shown in our VCS marketing screening.