The problem

The network was growing faster than the advertising system could keep up with it. Wimbledon was opening, Covent Garden was getting ready, and advertising was working on the logic of "show as many people as possible and hope that someone will come."

Meta Ads did not have a funnel that led people to sign up. Three thousand pounds a month went to coverage, and the connection between impressions and bookings remained opaque. In Google Ads, the budget was spread across campaigns that yielded nothing, instead of focusing on the ones that were effective.

The most acute problem was in analytics. The Altegio booking system ran on a separate domain, and cross-domain tracking in GA4 was not configured. What this meant in practice was that when a customer moved from the So.Shell website to the Altegio booking page, GA4 lost them. The session was interrupted, and the booking was not recorded as a result of the advertising campaign. In addition, the website cut off UTM tags when switching between pages, and even bookings made through the website itself were partially lost in the reports.

In other words, money was spent on advertising, people signed up, but we couldn't see which campaigns brought these people in. Budget decisions were made blindly.

Implementation

We divided the work into four stages of ~30 days each. Not for the sake of aesthetics, but because in advertising, each subsequent step depends on the data from the previous one.

The first thing we did in Meta Ads was to turn off all message campaigns. Both at the top and bottom of the funnel. The decision was not easy: 15 "quality responses" in correspondence looked like a result. But it was impossible to track how many of these dialogues became real records. iOS 14.5+ restrictions and privacy rules in Europe made such campaigns opaque. Money comes in, but what comes out of it is unknown.

Instead, we launched a lower funnel campaign optimized for bookings through the website and a pilot campaign through Altegio (an online booking system). At the same time, we built the middle funnel: clicks to the Instagram profile (CTR 5.25%), interactions with posts, and video views. The top level shows ads to everyone within a 10 km radius of the salon with a frequency of 2+, the middle level warms up interested users through content, and the bottom level converts the warm audience into bookings.

We also immediately suggested that the client add a pop-up window with a promo code for the first discount to the website. This was an identification tool: when a person leaves their contact details to receive a promo code, we can see which campaign they came from and track whether they made a booking afterwards. At that time, GA4 did not count bookings through Altegio correctly, and this promo code became one of the few ways to verify advertising data with actual bookings.

We turned off three campaigns in Google Ads that were spending the budget without results. We launched a remarketing campaign for website visitors. We identified a cross-domain tracking failure between the website and Altegio and began working on a fix so that GA4 could finally see the entire path from click to booking. We conducted a demographic analysis and found the core audience: women aged 25-34, with a 16.66% share of bookings and an application cost of £1.43.

The bottom of the funnel in Meta Ads started to work. 113 applications per month, application cost £7.23. The funnel broke even: approximately one-third of the budget for each stage. A separate campaign for Westfield through Altegio showed an application cost of £4.70, the best result among all campaigns for bookings. In the second half of the month, the cost per lead across all salons decreased by 6-49%: Meta's algorithm came out of the learning phase and began to optimize impressions.

In Google Ads, bookings increased by 39% (from 348 to 484) with only a 4.5% increase in budget. The cost per booking fell from £4.8 to £3.6, and the cost per identified application from £4.0 to £2.3. The new PMAX for Westfield generated 19 bookings in the first month, and the optimized location-based campaign reduced the cost per application from £9.6 to £4.0.

The Christmas period is a difficult time for beauty salon advertising: auctions become more expensive and user behavior changes. We used this time for structural changes.

In Meta Ads, the main campaign for appointments showed a 64% increase in applications and a 32% decrease in the cost per application. The client finally implemented the pop-up window with a promo code that we recommended at the start, and it immediately resulted in a noticeable increase in identified applications. We increased the lower funnel budget by 84%, and bookings increased by 61% to 135 per month. The share of those who sign up decreased from 8.96% to 5.66%: we started to attract colder audiences. But the absolute number of signups has almost doubled, and that's what matters most for the business.

We launched PMAX for three new salons in Google Ads and rebuilt the competitor campaign from search to PMAX (6 bookings in the first month). We turned off the remarketing campaign after three periods without results. We moved some of the search campaigns to broad match.

This period also showed where we were wasting money. Search for "nails" - 3 bookings at £88.9 each. Search for "manicure eyebrow" - one booking at £146.8. We turned off both and redirected the budget to PMAX, where the cost of a booking is £5.6-9.6.

The most important decision of this period seems inconspicuous: removing price per click restrictions in the Google Ads brand campaign. Previously, we manually limited the bid to control costs. This also limited the algorithm: it could not select the optimal price for each auction. When we let it work independently, bookings through branded search increased by 43% (from 319 to 456), and the cost of bookings fell by 31%. This was the biggest single improvement for the entire project.

The pop-up window with the promo code had been running for the second period and peaked at 244 identified applications. PMAX Wimbledon grew by 160%. The total cost of bookings fell to £3.4.

Meta Ads operated on a limited scale during this period: the client stopped some of the campaigns, and bookings fell by 81%. At the same time, the quality of records improved: the share of those who book increased to 9.6%, and the cost of an application remained at £7.53. This confirmed our third hypothesis. The problem was not the quality of the advertising, but the breaks: the algorithm learned well, but without a flow of visitors, it simply had no one to convert into customers.