SLAtech Fitness
83/100Class-conflict-aware, RTL Hebrew polish, injury-flag routing native
Reproducible 200-question Fitness-specific eval harness. +15-point lift vs generic SLAtech-Business (68/100). Driven by class-booking conflict detection, membership-tier upsell discipline, and trainer-matching by experience-level. Pairs with umbrella eval scoreboard, Fitness glossary and Fitness FAQ.
| Category | Fitness-tuned | Generic | Lift |
|---|---|---|---|
| Class-booking conflict detection Member's existing class roster is checked before new booking — no double-booking the 6pm yoga vs spin slot. Generic chatbots happily over-book. |
88 | 62 | +26 |
| Membership-tier upsell discipline Bot quotes the member's current tier's privileges + upgrade-path price clearly, no aggressive cross-sell pressure. Generic chatbots over-pitch upgrades on every interaction. |
86 | 66 | +20 |
| Trainer-matching by experience-level Beginner / intermediate / advanced classification against trainer specialty + member history. Generic chatbots match by next-available-slot only. |
84 | 71 | +13 |
| Injury / medical-flag handling If member discloses recent injury, the bot routes to coach review before booking high-intensity class. Generic chatbots book regardless of injury context. |
81 | 58 | +23 |
| Multilingual class info (HE / RU) Hebrew RTL polish on class titles, Russian transliteration of trainer names, locale-aware time formatting. |
79 | 76 | +3 |
Class-conflict-aware, RTL Hebrew polish, injury-flag routing native
Strong native scheduling but weaker injury-flag handling and English-first
No class-schema awareness, no injury-flag, English-first
No class schema, no Hebrew RTL, conversation cap on lower tiers
The per-vertical eval score is one input. Three more self-serve tools complete the picture without a sales call:
Eval methodology is open-source. 200 sealed Fitness-specific questions with LLM-as-Judge scoring on factuality, hallucination and confidence axes.