Results

Six hotels in Warsaw and Kraków, one shared Google Ads account and Meta campaigns for every property. When I took over, the account optimised toward a shared pool of conversions: phone clicks counted as bookings, and one hotel reported 79 conversions where 33 real reservations existed.

What I changed: one real conversion per hotel (a confirmed booking with its value), campaign-level goals instead of account defaults, a foreign tracking container removed from the booking engine, and budget moved from Performance Max campaigns that did not pay to brand search and event campaigns that did.

August 2026 against July: 15% more room bookings from Google campaigns and 14% more booking value, with 3% less spend. The brand campaign returned 15 times what it cost. Event and conference inquiries more than doubled. The booking funnel report from the same period showed 65% more reservations on 15% less traffic, because every step of the funnel improved. Reports go to the client every month with decisions, not just numbers.

Two city hotels sold weddings, birthdays, communions and conferences through a contact page and a phone number. Nobody could say which campaign brought which inquiry.

I built dedicated pages for each occasion, in Polish and English: offer, gallery, FAQ, a short form, and a “gate” where a guest leaves a phone number or email to download the menu. Every form, call over 30 seconds and download is tracked in one container, so campaigns optimise toward real inquiries.

August 2026 against July: inquiries from ads grew 144% in one hotel and 113% in the other. In the last week of the month, after the gates went live, an inquiry cost 38% less than the monthly average. The pages also started showing up in ChatGPT answers: visits from AI assistants rose 156% and brought inquiries at no cost.

The monthly report for six hotels used to mean a day of exporting numbers from Google Ads, Meta, Google Analytics and Search Console, pasting them into slides and hoping nothing was copied wrong.

I wrote a pipeline that pulls the full previous month from all four APIs on the first of the month, builds the tables (spend, conversions, cost per inquiry, bookings, top queries, best posts) and renders them to a PDF per city. An AI step drafts the commentary from the numbers; I write the decisions myself.

Since August 2026 the data collection takes zero hours. What I spend time on now is what the client pays for: reading the numbers, deciding what to scale and what to pause. The same pipeline runs a booking-funnel report that found a 46% mobile drop-off at the payment step, which the hotel is fixing.

Agency years, 2014. I managed the team that ran the Facebook sales campaign for Wizz Air: 100 routes in 15 countries at the same time, 26 of them from Poland, with the goal of selling tickets, not collecting likes.

How it worked: audiences built from people who had flown with the airline before, lookalike audiences on routes with little traffic, at least four creatives per route and more than 400 creatives a month, translated into local languages down to Azerbaijani and Hebrew. Every day the budget moved toward the routes that were selling below plan.

Six months later: over 200 000 tickets sold and a return of about 30 000% on ad spend. On Warsaw to Tel Aviv the campaign brought 19 times more conversions than search ads at 3.5 times lower cost. Facebook published it as one of 13 airline success stories worldwide, and it won a Mixx Award in the Direct Response category.