Tiago Cardoso

High-consideration trips sold across the US and Canada. A large share of bookings close offline or by phone with a travel agent, often days or weeks after the first ad click.

Last-click attribution credited only a fraction of the bookings paid media actually influenced. Budget was being set on numbers that understated the truth and hid where profit really came from.

Built a server-side tracking foundation (server-side GTM, Meta CAPI, Google enhanced conversions) to recover touchpoints the browser was losing, then added a multi-touch model and reconciled everything against confirmed revenue in the CRM: a three-lens view comparing platform-reported, multi-touch and CRM-confirmed outcomes.

On identical ad spend, the data-driven view surfaced far more bookings than last-click had credited, and a far higher true revenue ROAS, exposing channels that last-click was undervaluing and letting budget move to where it earned out.

Paid search runs across three brands and two markets (US and Canada) on a mature but sprawling Google Ads account. The brief: grow bookings without letting cost outrun margin.

The account had grown organically: overlapping campaigns, uneven coverage and bidding that chased platform-reported conversions rather than confirmed profit.

Rebuilt the account architecture across brands and markets, restructured Search and Performance Max coverage, tightened query-level optimisation, and aligned bidding to CRM-confirmed conversions by feeding offline bookings back into the platform, so bids learn from real revenue.

Scaled spend significantly while keeping the account profitable on a margin basis, growing bookings across the portfolio with budget concentrated where it actually earned out.

The group relies on a large email-marketing database to nurture future travellers. The channel needed a steady, low-cost flow of qualified subscribers.

One-off lead pushes were costly and hard to judge. Lead volume looked healthy, but nobody could see which leads became trips.

Built an evergreen lead-generation funnel: a set of lead magnets (destination guides, checklists, VIP offers) feeding a fully automated capture pipeline, with attribution that matches subscribers to confirmed bookings, so the channel is judged on revenue and margin, not sign-ups.

Over 100,000 leads captured in a single year and, measured on margin rather than revenue, a six-figure profit after ad spend: one of the most efficient channels in the mix.

Meta is the primary paid-social channel for direct package sales across the three brands and both markets, on a six-figure annual budget.

Meta reports generous, view-through ROAS that overstates its real contribution. Optimising to that number risks pouring budget into spend that looks efficient on the dashboard but does not show up in the bank.

Owned campaign structure, audiences, creative and budget pacing across the conversion funnel, and judged every campaign against CRM-confirmed bookings rather than Meta's reported attribution, so the channel scales on what it genuinely touches.

A conversion channel that Meta touched on 1,500+ purchases in 2025 and 750+ in the first quarter of 2026, with budget allocated on verified revenue rather than platform-reported ROAS.

Reporting captured spend, conversions, revenue and margin, but the quality of what campaigns produced lived in scattered emails, SMS threads and call recordings that no dashboard could see.

Two campaigns could look identical on cost and conversions yet differ completely in lead quality and intent. Judging on numbers alone missed half the picture, and reading thousands of messages and calls by hand was not feasible.

Built automated pipelines in Python (with AI assistance) that pull the numbers from the ad platforms, GA4 and the CRM, and connect a large language model (Claude) to Twilio, the Gmail API and BigQuery to gather and analyse the emails, SMS and call transcripts tied to each campaign, scoring it on both the numbers and the quality of demand.

A single, decision-ready view of each campaign on both dimensions, surfacing which campaigns brought genuinely qualified demand versus cheap but empty volume, and turning days of manual reading into an automated analysis.