A boutique fitness studio chain with four locations offered 42 group classes per week across all locations. Yoga, HIIT, cycling, barre, strength training. The schedule had been built incrementally over three years based on instructor availability, member requests, and the owner's intuition about what worked.
The problem was not visible in any single number. Average class fill rate was 71%, which seemed healthy. But the average hid a bimodal distribution: 8 classes averaged 95% capacity with regular waitlists, and 12 classes averaged 38% capacity with many sessions running at less than a third full.
The over-filled classes were turning members away. A member who cannot get into the 6 PM Tuesday HIIT class three weeks in a row starts looking at competitors. The owner's member churn survey confirmed this — "class availability" was the second most cited reason for cancellation.
The under-filled classes were wasting instructor pay, facility time, and marketing resources. A yoga class that runs with 4 people in a room built for 25 costs the same instructor fee as one running with 22 people.
The owner knew all of this intuitively. She just did not have time to analyze the booking data, model alternative schedules, and predict how changes would affect attendance. Changing the schedule was risky — removing a class might upset the 4 people who attend it, and adding a class at a new time might not draw enough members to justify the cost.
What the tool does
The tool analyzes the booking data from all four locations and produces schedule optimization recommendations based on actual member behavior, not assumptions.
For every class on the current schedule, the tool calculates the fill rate trend over the past 12 weeks, the waitlist frequency, the member overlap (how many members attend both this class and another), the time-of-day demand curve for each class type, and the day-of-week demand pattern.
It then models what would happen under different schedule configurations. If the 6 PM Tuesday HIIT class was offered at both 5:30 PM and 6:30 PM, would the demand support both sessions? If the 10 AM Thursday yoga class that averages 5 attendees was moved to 7 AM Saturday, would it capture the weekend morning demand that the data suggests exists?
The modeling is not speculation. It is based on member behavior data. When a member tries to book a full class and selects a waitlist spot, the system records what they did instead — booked a different class, booked the same class at a different location, or did not book anything. This revealed that 60% of members waitlisted for 6 PM HIIT booked nothing when they could not get in. They did not switch to another class. They simply did not work out that day.
The schedule changes
Based on the tool's analysis, the owner made five changes to the schedule.
She added a second 6 PM HIIT session on Tuesday, splitting the demand between 5:30 PM and 6:15 PM. Both sessions filled to 80%+ capacity immediately. The waitlist disappeared.
She moved the 10 AM Thursday yoga from Location 2 to 7 AM Saturday at Location 1, where the data showed unmet morning demand. Attendance went from an average of 5 to an average of 18.
She combined two under-performing evening barre classes (Wednesday 7 PM at Location 1 and Thursday 7 PM at Location 3) into a single Wednesday 7 PM class at Location 1, which was more centrally located. The combined class had higher attendance than either individual class because the members from Location 3 were willing to drive to Location 1 for a class with more energy and community.
She added a Saturday afternoon cycling class at Location 4, a time slot that had no classes but showed high member check-in activity for open gym use. The data suggested members were coming to the facility on Saturday afternoons but had no class option. The new class filled to 70% capacity in its first week.
She eliminated a Sunday morning strength class that had averaged 3 attendees for 16 consecutive weeks. The three regular attendees were offered a complimentary month at the Saturday morning strength class at the same location, and all three transferred.
Net result: the total number of weekly classes stayed at 42. No classes were added. No additional instructor hours were needed. The existing classes were rearranged based on where the members actually wanted to be.
The attendance impact
Total weekly attendance across all four locations increased from an average of 628 to 773 — a 23% increase. The increase came entirely from better scheduling. The same members, the same instructors, the same facilities. Different times and configurations.
Fill rate improved from 71% average to 84% average. More importantly, the distribution tightened. No class ran consistently above 95% (eliminating the frustration of being turned away) and only 3 classes ran below 50% (all of which were new additions being monitored for growth).
Member churn citing "class availability" as a factor dropped by half in the next quarterly survey.
The instructor scheduling benefit
The tool also optimized instructor assignments. Instructors had preferences about which times they taught, but the old schedule had been built around seniority rather than demand. The senior cycling instructor taught the prime-time slot even though her classes had lower attendance than the newer instructor's classes at the same time.
Rather than creating a confrontation, the tool showed the data: members who tried Instructor A's class and Instructor B's class retained at different rates. The data supported a schedule change framed as "aligning instructor strengths with time slots" rather than "replacing you with someone better." Instructor A moved to a morning slot she actually preferred, and Instructor B took the evening slot. Both were happier. Attendance increased in both slots.
The ongoing optimization
The tool runs a weekly analysis and surfaces recommendations when the data suggests a change. Most weeks, no change is needed. Occasionally, it identifies an emerging pattern — a new class type gaining traction, a time slot experiencing declining demand due to seasonal patterns, or a location where a specific class type has unmet demand.
The owner reviews the recommendations monthly and makes 1 to 2 schedule adjustments per quarter. The schedule is no longer static. It evolves with member behavior, and the evolution is driven by data rather than guesswork.
The cost
Four days of build time. The tool connects to the studio's booking platform through its standard API. No per-analysis fees. No subscription costs.
The owner's assessment: "I spent three years guessing about the schedule. The tool showed me what my members were trying to tell me through their booking behavior. They wanted the same classes at different times. That is not a technology insight. It is a customer insight that the technology made visible."
Get posts like this in your inbox
No spam. New articles on AI strategy, governance, and building with AI for small business.