TiNoleggio is Italy's leading car rental comparison and booking platform. They aggregate offers from 75+ rental partners, serving both leisure travelers (tinoleggio.it) and international markets (tinorent.com). Revenue comes from booking commissions and insurance product sales, with additional B2B/API channels for tour operators like Voyage Privé.
Donux operates as TiNoleggio's embedded product design team. The engagement includes 12 Data Sprints per year (2-week cycles focused on a single measurable objective) and 3 Design Sprints per year for larger structural projects.
Every sprint follows the same rhythm: the first week is co-design between Donux and TiNoleggio, the second week is handoff to the development team for implementation. Biweekly Outcome Meetings keep both teams aligned on data, decisions, and next moves.
TiNoleggio had two clear problems. SERP click-through rates were stagnating, leaving revenue on the table at the top of the funnel. And insurance product attach rates, the company's main margin lever, were underperforming year-over-year.
Both problems needed more than design intuition. With 3,000+ monthly bookings and dozens of variables affecting conversion, guessing wasn't an option. TiNoleggio needed a systematic, experiment-driven approach to identify what actually moves the numbers.
On top of that, organic traffic was declining (-30,000 clicks over 3 months), making every visitor more valuable. The pressure to convert existing traffic was high.
— Increase CTR from SERP to booking: move more users from search results to the checkout funnel
— Increase insurance product sales: improve attach rates on the platform's highest-margin product
— Build a data infrastructure that supports continuous experimentation and measurement
Setting up the experimentation stack
Before running any tests, we needed reliable measurement. We configured GrowthBook for A/B testing, built a Looker Studio dashboard with daily-updated conversion metrics, and established statistical significance thresholds. When the GA-to-BigQuery pipeline broke later in the project, we migrated the testing infrastructure to PostHog.
SERP optimization (top of funnel)
We started with the search results page, where every percentage point of CTR improvement compounds across thousands of daily sessions.
The first experiment was straightforward: testing CTA label copy. "Vedi l'offerta" vs "SELEZIONA" reached 99.2% statistical confidence in one week. The winning variant became the new default.
We then tested a single-button modal vs. a dual-button layout. Result: +10-11% visits to the driver data page and +3.8% bookings. Shipped to production.
Not every test won. A 9-variant matrix testing 3 labels against 3 colors showed the control (green "Vedi l'offerta") outperformed all alternatives. We closed the test and moved on, confirming the baseline was already strong.
We also reorganized the SERP filters based on usage data, which increased filter engagement.
Insurance upsell optimization (margin lever)
Insurance was the longest and most complex workstream, spanning 10 months.
Starting point: upsell rate was around 2%, and a legacy product ("Cambia il Pianeta" at €4) had zero sales in two years. Less than 50% of users even saw the insurance section. The existing pre-selected checkbox was a regulatory risk (Booking.com had been fined for a similar pattern).
We designed and tested a dedicated insurance selection modal. It worked: upsell rates went from ~2% to a 5% baseline, with peaks above 10% during high season. Average ticket value rose from €15 to €17.
When absolute insurance numbers started declining (driven by B2B channel growth diluting the percentage), we dug into the data. In relative terms, insurance attach rates were actually up 5% year-over-year. The B2B channel simply didn't support insurance sales due to a technical limitation.
We then designed a FOMO banner for the checkout page, inspired by Ryanair and Discover Car, using persuasive statistics (claim rates, average deductibles, deposit blocks). Tracked live from day one.
The latest experiment, three visual variants of the insurance section (Comics/illustrated, Car/damaged vehicle, Control), showed the Comics variant outperforming at ~90% significance. Still running at the time of writing.
Every design decision had to respect EU regulations: no "recommended" labels, no pre-selected opt-out, and precise wording around deductible reimbursement vs. coverage.
Billing flow redesign (conversion protection)
A critical operational problem surfaced: ~5% of daily bookings had broken billing data, costing the dev team ~2 hours/day in manual fixes. We redesigned the billing form, moving it before payment and making fields mandatory.
We monitored the impact closely. Conversion dipped initially, but Clarity session recordings (30+ sessions analyzed) revealed UX issues: users spent up to 5 minutes searching for their SDI code, and the SDI/PEC alternative wasn't communicated clearly.
After iterations, the data stabilized with no statistically significant conversion loss. The operational cost dropped to near zero.
Landing page optimization
Analysis revealed that CPC campaigns were sending users to pages with specific promises ("rent without credit card") but the product didn't reflect them: no pre-applied filters, no aligned messaging. Landing pages used an outdated layout with legacy HTML stored in the database.
We designed a new modular system: reusable sections with dynamic content per city, a redesigned hero section with dynamic titles, median pricing, Trustmate reviews, and a search widget. The 40% of users who scrolled the full page without interacting signaled the above-the-fold experience needed the most attention.
— Button color on Tinorent: higher conversion rate
— SERP filter reorganization: higher filter engagement
— CTA label on SERP card: higher click-through rate
— Single-button modal vs dual-button: +11% page visits, +3.8% bookings
— CTA matrix (9 variants: 3 labels x 3 colors): no improvement over existing design
— Insurance selection modal: upsell rate from 2% to 5%+
— Billing page redesign: zero conversion loss, ~150 fewer manual fixes/month
— Insurance visual design (3 variants): in progress, illustrated variant leading
Regulation constrains design choices
EU consumer protection rules prohibit pre-selected insurance opt-outs and "recommended" labels. Every insurance UX decision had to be legally validated. The winning approach: positive emotional framing ("you're covered up to €6,000") instead of fear-based copy.
B2B growth can mask consumer wins
Insurance attach rates appeared to drop, but the real story was B2B channel growth (where insurance can't be sold via API). Segmenting the data by channel revealed consumer insurance was actually up 5% YoY.
Operational costs hide in plain sight
The billing form wasn't a UX problem initially, it was an ops problem. But the fix (mandatory fields before payment) created a UX problem. Solving one without creating the other required session recordings, segmented analysis (orders above/below €400), and iterative refinement.
Not every test needs to win
The 9-variant CTA matrix confirmed the existing design was already optimal. That's a valuable outcome: it freed the team to focus effort elsewhere instead of endlessly iterating on a solved problem.
SERP performance
CTR from SERP: +2 percentage points YoY.
Insurance upsell
Upsell rate: from 2% baseline to 5% average (150% relative uplift). Revenue per booking: from €15 to €17 average ticket. YoY insurance attach rate: +5% on consumer channel (when properly segmented).
Operations
Billing errors requiring manual intervention: reduced from ~150/month to near zero. Dev team time saved: ~2 hours/day freed from manual billing fixes.
Infrastructure
Looker Studio dashboard with automated daily metrics. PostHog configured as the new experimentation and analytics hub. GrowthBook Bandit testing for multi-variant experiments.
The partnership continues with three major workstreams:
— Personal Area: a three-phase project adding user accounts with magic link login, saved preferences, and one-click rebooking. Recurring users already represent ~25% of bookings and ~40% of revenue.
— Landing page system: modular, template-based pages with dynamic content per city, replacing the legacy layout.
— PostHog migration: full transition of A/B testing from the GrowthBook/GA/BigQuery pipeline to PostHog for more reliable experimentation.
The first step is a quick chat.
