Credit: 0% 0% 100%
In a last-click model, the consumer's entire purchase journey collapses to a single point: the last ad they interacted with before buying. A campaign that ran three months of upper-funnel awareness ads gets zero credit if a retargeting ad was the final touchpoint. The retargeting campaign claims 100% of the attributed revenue.
This is not a bug — it is a deliberate design choice. Last-click is simple, deterministic, and easy to audit. These properties made it the dominant attribution standard as retail media scaled. Its limitations become critical when comparing performance across platforms with different model defaults.
Last-Click vs. First-Click: The Criteo Problem
Criteo, one of the largest retail media platforms, uses a first-click model by default — assigning 100% credit to the first ad interaction rather than the last. This creates a structural incompatibility with Amazon's last-click ROAS that is entirely independent of campaign performance.
| Scenario | First-Click ROAS | Last-Click ROAS | True Return |
|---|---|---|---|
| Criteo introduces consumer (Day 1), Amazon closes sale (Day 5) | High ↑ | 0x (no credit) | Shared |
| Amazon introduces consumer (Day 1), Criteo retargets (Day 8), sale (Day 8) | 0x (no credit) | High ↑ | Shared |
| Single platform, single touchpoint | Identical | Identical | Accurate |
The model difference systematically distorts cross-platform ROAS comparisons. An agency comparing Criteo (first-click) with Amazon (last-click) without model correction is comparing apples to oranges — not because campaigns perform differently, but because the measurement rules are different.
How Last-Click Biases Budget Allocation
Last-click attribution systematically over-credits lower-funnel campaigns that appear close to conversion and under-credits upper-funnel campaigns that build awareness and intent. In a retail media context:
Sponsored Products retargeting — which reaches consumers who already searched for the product category — captures last-click credit on purchases that may have been heavily influenced by earlier brand awareness advertising. Its reported ROAS appears high.
Sponsored Brands and display — which introduce consumers to products earlier in the journey — lose their credit to subsequent retargeting touchpoints under last-click. Their reported ROAS appears low, even if they drove the initial intent.
Agencies allocating budget based purely on last-click ROAS will systematically over-invest in retargeting and under-invest in awareness, potentially leaving growth on the table by starving the top of the funnel that feeds the retargeting pool.
Last-Click in the Normalization Model
Because last-click is the default for Amazon — the largest retail media platform — it serves as the model baseline in normalized ROAS calculations. Platforms using first-click (Criteo) receive a downward model correction factor (M) that accounts for the systematic difference in attribution behavior between first-click and last-click models under typical purchase journey distributions.
The normalization model correction for Criteo is not a judgment that first-click is wrong — it is a mathematical correction for the fact that first-click and last-click produce structurally different ROAS figures for identical campaign performance. Normalization makes both numbers speak the same language.
FAQ
Amazon Sponsored Products, Amazon Sponsored Brands, Walmart Connect, Target Roundel, Instacart Ads, and Kroger Precision Marketing all use last-click as their default model. Criteo is the major exception, using first-click by default. Some platforms offer model configuration at the campaign level, but default last-click is the industry norm.
Multi-touch attribution (MTA) distributes conversion credit across multiple ad interactions in the consumer journey, rather than assigning it all to one touchpoint. Models include linear (equal credit to all touches), time-decay (more credit to touches closer to conversion), and data-driven (algorithmic credit distribution). MTA is more methodologically complete than last-click but requires cross-platform identity resolution that is difficult to achieve in closed retail media environments where each platform operates its own walled garden.
Yes, this is a well-documented structural bias. Because retargeting ads appear close to purchase, they capture last-click credit at high rates. Budgets managed purely on last-click ROAS tend to concentrate in retargeting over time, which can reduce the size of the audience pool that retargeting reaches — eventually constraining overall program growth. Incrementality testing is the most reliable way to identify when retargeting investment has reached diminishing returns in terms of true causal impact.
RetailNorm's normalization model includes a model-type correction (M factor) that adjusts for the difference between first-click and last-click attribution. Criteo ROAS figures are adjusted to last-click equivalent before cross-platform comparison, making budget allocation decisions model-agnostic.