What Is an Attribution Model?
An attribution model is a set of rules that decides which marketing touchpoints receive credit when a customer converts. Modern buyers rarely encounter a brand once and immediately purchase. A typical South African online shopper might first see a brand's Instagram Reel, later search Google and click an organic result, then receive a promotional email and finally complete a purchase after clicking a retargeting ad. Four different channels contributed to that sale, but most platforms by default credit only one.
The choice of attribution model fundamentally changes how you understand your marketing performance. The main models are: first-touch attribution, which credits the channel that first brought the customer to your site; last-touch attribution, which credits the final channel before conversion; linear attribution, which splits credit equally across all touchpoints; time decay attribution, which gives more credit to touchpoints closer to the conversion; and position-based (U-shaped) attribution, which gives 40% credit each to the first and last touch and divides the remaining 20% among middle touches. Google Analytics 4 also offers a data-driven model that uses machine learning to assign credit based on actual observed patterns in your data.
Each model produces a different picture of your channel mix. Under last-touch attribution, Google paid search might appear to generate most conversions because it is often the final click. Under first-touch attribution, social media or display advertising might receive more credit because it introduces customers to the brand. Neither view is complete on its own, but together they reveal where different channels contribute along the funnel.
Choosing the right model for your campaign tracking setup depends on your business model and the length of your customer journey. A business with short, impulse-driven purchase cycles might be well served by last-touch attribution. A business selling a high-consideration service, such as commercial legal advice or custom manufacturing equipment, with a weeks-long decision period benefits more from a model that values awareness and nurture touchpoints.
Attribution Model In Practice
A Johannesburg property developer running a six-week pre-launch campaign uses a combination of Facebook awareness ads, Google Display retargeting, and a WhatsApp broadcast to their existing database. Under last-touch attribution, the WhatsApp broadcast looks like the hero channel because most serious enquiries come after receiving that message. But removing Facebook from the mix would have meant most of those recipients had never heard of the development. A linear or time-decay model shows that the Facebook awareness phase played a meaningful role in warming audiences before the WhatsApp push converted them.
South African marketers often discover that SEO is undervalued under last-touch attribution because organic search frequently initiates the journey but rarely closes it. When they switch to a multi-touch model in GA4, SEO's contribution to the conversion funnel becomes clearer, often justifying a higher content and optimisation budget than a pure last-touch report would suggest.
The models, and what each one flatters
| Model | Credit goes to | Flatters |
|---|---|---|
| Last click | The final touch before conversion | Branded search and retargeting |
| First click | The first touch | Awareness channels and top-of-funnel content |
| Linear | Split evenly across all touches | Channels that appear often but persuade little |
| Time decay | Weighted towards recent touches | Late-funnel activity |
| Position based | 40/40 to first and last, 20 split between | Neither extreme; a reasonable compromise |
| Data driven | Modelled from your own conversion paths | Nothing deliberately, but needs volume |
Every model is wrong in a specific direction. The useful discipline is knowing which direction, so you can read a report without being misled by it.
Choosing one, and the limits you cannot model away
Pick based on your sales cycle and channel mix. Short cycles with few touches can live with last click. Longer considered purchases need a multi-touch view or the awareness channels get defunded on paper while still doing the work.
Three limits no model removes:
- Cross-device and cross-browser gaps. Tracking prevention breaks paths that genuinely happened.
- Offline influence. Word of mouth, a billboard, a conversation. None of it appears.
- Correlation, not causation. Attribution shows what was present, not what caused the sale.
The only method that answers the causal question is a holdout test: pause a channel for a defined period and see whether total conversions move.
See multi-touch attribution and our analytics service.
FAQ
Which attribution model should I use in GA4?
GA4 defaults to a data-driven attribution model, which uses machine learning to distribute credit based on how each touchpoint statistically influences conversions in your data. For businesses with limited data volume, a linear or position-based model may produce more stable, interpretable results than data-driven attribution.
Why does my ROAS look different in Google Ads vs GA4?
Google Ads typically uses last-click or data-driven attribution within its own platform, while GA4 uses its own attribution model for the same conversions. Because each platform applies different rules for assigning credit, the conversion counts and ROAS figures will differ. Neither is wrong - they reflect the same reality through different lenses.
Which attribution model is most accurate?
None is accurate in an absolute sense; each distributes credit by a rule. Data-driven attribution is usually the most defensible when you have the conversion volume for it, but the only causal test is a holdout experiment.
Why do my platform numbers not match Google Analytics?
Each platform claims conversions under its own model, window and identity graph, so totals overlap and exceed reality. Reconcile against your actual sales ledger monthly and treat platform figures as directional.