Case studies
What the work
actually looks like.
Representative engagements, each structured the same way: the situation, what we built, and the result.
Representative engagements based on the work we do and the results this work typically produces.
Connecting website behaviour to bookings
improvement in new-member LTV-to-CAC within two quarters
- Attribution
- BigQuery
- Meta
Situation
A studio was spending across Meta, Google, and referral but couldn't tell which channels produced members who stuck around, not just which produced sign-ups. So budget kept flowing to the cheapest lead, regardless of whether those people stayed.
What we built
We tied front-end CMS and website behaviour to the back-end booking platform, creating a single view from first website touch, to first class, to lifetime value. Every channel could now be judged on the members it produced, not the clicks.
Result
Spend was reallocated away from the cheapest-lead channel toward the highest-LTV channel.
Automated Mariana Tek sync for live dashboards
efficiency in reactivation spend vs. cold acquisition
- Data Sync
- Mariana Tek
- Klaviyo
- Snowflake
Situation
A multi-location operator was exporting data from Mariana Tek by hand every month. Dashboards were always stale, and ad audiences were built manually, late, and inconsistently across locations.
What we built
Automated, ongoing sync from Mariana Tek into a warehouse feeding live dashboards, plus auto-refreshing Meta and Klaviyo audiences, including lapsed members, high-value members, and trial non-converters, rebuilt on a schedule.
Result
The manual export work disappeared, dashboards stayed live, and reactivation spend went to audiences that were always current.
Pre-churn detection & intervention
reduction in voluntary churn among flagged members
- Lifecycle
- Klaviyo
- Mariana Tek
Situation
Cancellations always arrived as a surprise. By the time a member emailed to cancel, they were already gone, mentally checked out weeks earlier, with no signal anyone acted on.
What we built
An attendance-decay signal that flags members whose frequency is dropping (say, 3×/week down to 1×/week), feeding an automated win-back sequence triggered before cancellation, not after.
Result
At-risk members were reached 3–4 weeks before they would have cancelled, while there was still a relationship to save.
Turning the schedule into a retention lever
increase in effective class capacity utilisation without adding classes or locations
- Capacity
- Scheduling
- Mariana Tek
- BigQuery
Situation
A multi-location studio had classes waitlisting during peak evening hours while entire off-peak time slots ran a third full. Scheduling decisions were made instructor by instructor, on feel, with no visibility into which slots actually drove revenue and retention versus which just filled a calendar.
What we built
We joined class-session and reservation data by location, instructor, and time slot to calculate fill rate and waitlist depth for every class, every week. A recurring scheduling report surfaced the slots worth adding capacity to and the ones quietly losing money, and members waitlisted on a full class got an automatic nudge toward the nearest open alternative instead of falling off the calendar entirely.
Result
Peak-hour capacity opened up without adding a single class, off-peak slots filled from the same member base that previously had nowhere else to go, and members turned away from a full class stopped disappearing.
Flagging trial members before the conversion window closes
typical lift in trial-to-membership conversion rate
- Trial Conversion
- Mariana Tek
- Lifecycle
Situation
A studio running trials on Mariana Tek could see who signed up and who converted at the end, but nothing useful in between. By the time a trial expired unconverted, there was no time left to change the outcome.
What we built
A query against the reservation data isolating the single behavioral signal that predicts conversion better than anything on the intro offer itself: whether a trial member books a second class within 7 days of the first. Anyone who hasn't booked by day 5 or 6 gets flagged automatically, feeding a specific, timely nudge instead of a reminder sent after the trial has already ended.
Result
Studios running this kind of mid-trial intervention typically see trial-to-membership conversion improve by roughly 25 to 35 percent, not from a better offer, just from reaching people inside the window where a nudge still changes the outcome.
Turning paused memberships into a duration signal
of memberships paused 60+ days recovered to active
- Retention
- Mariana Tek
- Lifecycle
Situation
A multi-location studio's frozen-membership bucket had grown to a few hundred accounts with no way to tell who had been paused two weeks versus who had been paused eight months. The booking platform tracked freeze as a status, not a duration, so a member on a short genuine break sat in the exact same bucket as someone who had quietly decided not to come back.
What we built
We treated pause duration the same way we treat attendance decay for pre-churn detection: as a rolling signal, not a static flag. Any membership paused past a set threshold, tuned to the studio's own patterns, got surfaced automatically and fed into a recurring outreach queue with a short, direct message: come back this week, or close it out.
Result
A meaningful share of long-paused members either reactivated or converted to a clean cancellation instead of drifting for months in a status that counted as neither active revenue nor a lost member. The freeze bucket stopped quietly growing every month, because every paused account now had a clock running on it.
Finding which members would hurt most to lose
of at-risk top-decile revenue members retained through early, escalated intervention
- Retention
- Revenue
- Mariana Tek
Situation
A studio's monthly revenue looked healthy on paper. But nobody could answer a simple question: which members make up most of that revenue, and are any of them quietly slipping away? The studio already had a system flagging members showing early warning signs of leaving. The problem was that this warning system and the revenue numbers lived in two separate reports that had never been compared. So a top-spending member could start fading out and get the same generic win-back email as anyone else, if they got noticed at all.
What we built
We answered the question directly. First, we ranked every member by how much they'd spent over the past 12 months, to find the roughly top 10% who account for most of the studio's revenue. Then we checked that list against the same early-warning system already used to catch churn. Any high-revenue member who tripped that warning, for example a top-decile member whose weekly visits had quietly halved over the past month, or who let a package lapse without rebooking, got moved to the front of the line, with a more personal, higher-touch follow-up instead of the standard generic win-back email.
Result
Most studios have never actually answered a basic question: which twenty or so members would hurt the most to lose, and are any of them already trending down? This closes that gap. Once a high-revenue member is flagged early instead of after they've already left, catching them before they cancel becomes possible. In this kind of engagement, a meaningful share of the top-revenue members already at risk get retained, because they're caught with enough time to act.
Recognise your studio in any of these?
Most owners see at least one. Let's talk about which one is costing you the most right now.