Engagement diary · Gym chain · 23 sites · 8 weeks · Jun–Aug 2026
Three sites were about to lose their budgets. The phones disagreed.
Written by Saoirse Brennan · 14 August 2026 · Figures representative, chain anonymized
A 23-site gym chain paid €38 per trial membership on €19,000 of monthly spend and was ready to cut three sites from paid search entirely. Eight weeks later trials cost €19, and the three condemned sites finished inside the chain's top ten. The rescue was a measurement fix, never a bidding trick, and the diary keeps the failed Performance Max test in the record where it belongs.
€38 → €19
trial membership CPL
3
written-off sites rescued by call data
231
negatives added in the first month
The starting position
The chain ran competent creative and a healthy trial offer; the problem was epistemic. Conversion data was web forms only, in a business where a meaningful share of prospective members, and nearly all of the over-45s, pick up the phone. The whole account was being graded on the behavior of its youngest customers. Add a blended budget and a two-year-old negative list, and the €38 blended CPL was less a performance figure than a rumor. The quarterly review had already drafted the budget cut for the three "worst" sites; the engagement started eleven days before that meeting.
Week 1
The audit table
Twenty-three sites, €19,000 a month, trial-membership CPL ranging from €21 to €67 by branch, blended to a comfortable-looking €38. Three suburban sites flagged for losing their ad budgets entirely at the next quarterly review. The audit also found the day-pass problem: high-volume 'day pass' and 'gym near me price' queries polluting the trial campaigns' conversion data at every site.
Weeks 2–3
Calls counted
Forwarding numbers and per-site GA4 installed, with the booking system's trial-start data imported back as offline conversions. The three written-off sites turned out phone-heavy: their real CPL sat near €24, not the €55-plus the form-only data claimed. Their members skewed older and simply rang the front desk, exactly the pattern the wider call data predicts, with phone leads converting at rates web forms never touch. Budgets survived.
Weeks 4–5
Query hygiene
Weekly reviews started; 231 negatives across job-seeker, class-timetable, and day-pass themes, built as themed shared lists. Day-pass queries got their own cheap campaign with its own landing page instead of polluting trial-membership data, and it quietly became profitable on its own terms: low value, near-zero management, honest numbers.
Weeks 6–7
Landing and offer split
Trial pages split by site cluster with local class schedules, coach photos, and each gym's actual floor shots; tCPA reset per cluster from corrected data. One failed test, honestly logged: Performance Max for trials underperformed search at every site cluster and was killed at its cap, with the channel-level report showing the spend had drifted to Display placements that clicked but never trialed.
Week 8
The close
Blended trial CPL closed at €19 against €38 at the start, on corrected data both times. The per-site spread narrowed from 3.2× to 1.8×, the three rescued sites finished inside the top ten, and the join-rate column confirmed the cheaper trials were converting to memberships at the same rate as before, so the volume was real demand rather than discount tourists.
Where the €19 came from
| Lever | Contribution | Nature |
|---|---|---|
| Counting calls + booking imports | €38 → €29 | measurement, zero media change |
| 231 negatives + day-pass quarantine | €29 → €23 | query hygiene |
| Cluster landing pages + reset tCPA | €23 → €19 | relevance + bidding on real data |
The measurement lever paid the most, which is typical and worth internalizing: the account did not get cheaper leads first, it got honest arithmetic first. The failed PMax test also earned its keep as information; the channel report, standard in Google Ads since November 2025, showed exactly which placements ate the test budget, which made the kill decision a two-minute conversation instead of a debate.
The benchmark context
Health and fitness search converts at a 6.94% median with cost per lead at $67.36 in the 2026 LocaliQ data. A €19 trial CPL sits well under that band because a gym trial is a lighter commitment than the vertical's blended lead mix, and we say so instead of framing it as heroics. The honest star of the diary is week two: 44% of the chain's trial starts arrived by phone at sites with older member demographics, and form-only data had been starving exactly those sites. The wider numbers make the mechanism unsurprising, with phone leads converting during the call at 37% across industries while web forms average near 1.7%.
What generalizes
Never cut a location's budget on form-fill data alone, and never let a high-volume low-intent query family (day passes, timetables, jobs) share a campaign with your money conversion. The counting layer comes first, per call and lead tracking, the weekly hygiene per the core service, and failed tests get written verdicts rather than quiet burials. The free audit builds the week-one table that started this diary, including the phone-share estimate per site that predicted which branches were being miscounted.