First-touch attribution credits the earliest eligible recorded touchpoint; last-touch attribution credits the latest eligible one before the chosen outcome. Neither reconstructs every influence on a traveler. For a rental business, the useful question is not which model is universally right, but whether discovery and booking completion are being confused.
This guide combines Google's public measurement definitions with a review of Orcar's documented reservation and payment attribution design on 14 September 2026. The design observations below are not a production coverage audit or a claim about channel performance. The hero is conceptual artwork, not a customer journey chart.
1. Define the touch before choosing the model
“First” means first within your observable history and eligibility rules, not necessarily the first time someone heard about you. A traveler might receive a recommendation in a private chat before opening any measurable link. “Last” also needs a definition: the final visit, final tagged interaction, and final non-direct click can be different things.
Google Analytics distinguishes user, session, and event scope. First user source describes acquisition of the observed user; Session source describes the source of a session; event-scoped attribution assigns credit to a key event. Changing the reporting attribution model does not rewrite user- or session-scoped dimensions. [1][3]
Do not tell a colleague to select “first click” in today's GA4 attribution model menu. Google removed that reporting model in November 2023. First user dimensions still exist, and a first-party system can retain its own first-touch record. These are different concepts. [4]
Consent choices, blocked storage, cleared cookies and switching devices can interrupt observable history. A missing earlier touch must remain unknown; do not bypass privacy choices to reconstruct it.
2. Follow one booking through three visits
Consider this fictional example, not an Orcar customer record. In the same browser, a traveler opens a tagged partner article link, later returns through a paid search ad, then types the rental site's address and pays. Assume both tagged touches remain eligible and linked to that booking.
| View | What receives credit in this example? | What the view helps explain |
|---|---|---|
| First eligible recorded touch | Partner article | Observed discovery |
| Last non-direct click | Paid search | Latest eligible non-direct return |
| Literal final visit | Direct visit | How the payment session began, not necessarily the credited channel |
| Business outcome | One qualifying paid booking | The deduplicated outcome all views describe |
Google's attribution models exclude direct visits from credit unless the path is entirely direct. That rule is why “the customer came back directly” does not automatically make the booking a direct-channel conversion. A custom first-party rule must be documented separately. [4]
The article and ad have not each produced an additional booking. They are two views of the same outcome. Adding first-touch and last-touch totals doubles the accounting rather than improving the evidence.
3. What Orcar's booking design makes visible
In the documented Orcar implementation, the browser builds separate first- and last-touch attribution records. Reservation creation and payment creation both pass attribution and analytics-session context to the backend. Payment creation can update the reservation's submitted attribution, and payment completion can link the reservation to the payment session.
This creates a rental-specific interpretation problem: the information available when a reservation starts need not match the information available when payment completes. A later visit can bring new evidence. Before comparing reports, identify which snapshot and outcome time each report uses. Do not assume a field called “first touch” is an immutable record of a person's lifetime discovery.
The documented acquisition funnel derives its final booking step from confirmed paid reservations linked to measured sessions, rather than accepting a browser payment-success event alone as the business outcome. This is a description of the inspected design, not proof that every live booking has a complete link. A client event, a payment attempt, and a confirmed paid reservation answer different questions.
4. Compare the same rental cohort, not two totals
Use this review checklist before deciding that one channel is outperforming another:
- Outcome: choose a deduplicated confirmed paid booking, a completed rental, or another clearly named outcome. Do not mix them.
- Time: state whether the cohort is grouped by visit, booking payment, or pickup date, plus the time zone and data cutoff.
- Eligible history: record the lookback window and direct-traffic rule. A cookie lifetime is not automatically the reporting lookback window. Google defines the latter as how far back a touch can qualify for credit. [3]
- Snapshot: state whether attribution was captured at booking creation, payment creation, or a later reconciliation.
- Coverage: show linked bookings divided by all qualifying bookings in the same outcome cohort. Within linked bookings, show missing source labels separately. A valid session link does not guarantee a known channel.
- Value: apply one currency and refund policy consistently. Paid amount, net revenue after refunds, and contribution profit are not interchangeable.
Keep an explicit unlinked bucket. If coverage changes between weeks, changes in attributed bookings may partly reflect better matching rather than better marketing. Restrict access to reconciliation identifiers; do not put traveler names, document details, or booking-access tokens into campaign labels.
5. Use the disagreement to decide the next test
If a channel appears often in first-touch views but rarely in last-touch views, inspect whether it introduces travelers who return later. If the reverse occurs, inspect whether it helps people finish a decision. These are hypotheses, not proof that either channel creates incremental demand.
Compare the same booking cohort under both rules, preserve the unknown portion, and review representative paths only with appropriate access. A controlled experiment or other defensible causal design is needed before claiming that removing a channel would remove its attributed bookings. Our rental-results measurement guide explains the difference between a measured change and an effect.
For offline partnerships, our QR campaign guide covers the separate problem of proving each handoff from scan to booking. This article instead asks how to allocate credit once an outcome is linked. Explore Orcar's rental-partner offering when planning your own traveler acquisition and booking flow.
Frequently asked questions
Is first-touch or last-touch attribution more accurate?
Does a direct return erase the earlier source?
Can I call an unlinked booking organic or direct?
Sources