What is last-touch attribution?

Last-touch attribution gives the entire credit for an order to the final paid touchpoint before the purchase. If a customer clicked a search ad three weeks ago, a social ad last week, and a retargeting ad an hour before buying, last-touch hands the retargeting ad 100% of the order's contribution margin and the other two nothing.

Formula

last-touch margin (channel) = Σ full order margin, over orders whose
                              last paid touch belongs to that channel

Worked example

An order carries €80 of contribution margin after a journey of three paid touches: search → social → retargeting. Last-touch books €80 to retargeting, €0 to search and social. Linear books €26.67 to each. Summed over a month, the gap between those two readings per channel is exactly the Δ last-vs-linear column on the Attribution tab.

How Saldo Metrics computes it

The attribution engine rebuilds each order's journey from pixel events: every purchase event is traced back through the same customer's page views carrying UTM parameters in the 30 days before the purchase, one touch per session, chronologically ordered. Under last_touch, the whole order — its revenue and its contribution margin, both EUR — is written to fact_attribution against the last of those touches. An order with no paid touch in the window is attributed to organic, so channel totals always add up to total attributed margin. Uncosted orders are excluded entirely rather than attributed zero margin, which would understate whichever channel sold them. v_attribution_model_comparison then sums the last_touch rows per channel, next to the same sum under every other model.

Why it matters

Last-touch is what most ad platforms approximate, so it is the model to look at when reconciling this app against an ads manager. It systematically over-credits bottom-of-funnel channels — retargeting and brand search "close" journeys that other channels started — and the model comparison exists to show how much of a channel's apparent performance is exactly that.

Common mistakes

• Budgeting on last-touch alone. Cutting a top-of-funnel channel that rarely closes can starve the retargeting channel that does — the journeys it closed no longer start.
• Reading last-touch as "what really happened". It is one deliberate rule among four; the honest signal is where the models disagree.
• Forgetting the 30-day window. A journey that started 35 days before purchase contributes only its recent touches, on every model.

Where you see this in the app

Marketing → Attribution, as the last-touch column of the model comparison and as a selectable model in the per-model breakdown.