The infrastructure specification for moving ETS Performance from broadcast marketing to precision targeting. RFM segmentation, predictive scoring, look-alike audiences, and CDP architecture for a network of 50+ facilities and 10,000+ member families.
ETS Performance sends the same message to every parent in every market. A faith-driven father in Oklahoma receives the same email as a data-literate physician in Austin. A coach exploring the operator path gets the same nurture cadence whether they downloaded a guide yesterday or attended a webinar six months ago.
This is not a creative problem. It is an infrastructure problem. The data exists -- membership records in GymMaster, behavioral data in GA4, assessment data in VALD, engagement data in the email platform -- but it lives in disconnected silos with no unified view, no scoring, and no automated activation.
The goal: right message, right parent, right time. Personalization at scale requires infrastructure, not just intent. This document specifies that infrastructure.
Collection with partial unification. Data captured but siloed. No behavioral segmentation, no predictive scoring, no automated audience activation.
Full activation within 12 months. Segments sync to ad platforms automatically. Predictive scores trigger workflows. Suppression lists update daily. Every dollar of media spend informed by first-party behavioral data.
Visual audit of 10 data sources across the ETS ecosystem. Quality scores reflect completeness, accuracy, recency, consistency, and accessibility.
Scale: 1.0 (unreliable) to 5.0 (production-ready)
| Source | Data Type | Volume | Quality | Status | Consent |
|---|---|---|---|---|---|
| GymMaster | Members, billing, attendance, plan tiers | 10,000+ records | 4.0 |
Production-ready | Contractual |
| Booking System | Eval bookings, no-shows, child demographics | 500-1K/month | 3.6 |
Usable | Explicit |
| VALD Performance | Force plate data, bilateral asymmetry, re-tests | 10,000+ assessments | 4.4 |
Production-ready | Contractual |
| GA4 | Sessions, events, conversions, traffic | 50K+ sessions/mo | 3.8 |
Usable | Cookie consent |
| Email Platform | Contacts, opens, clicks, sequences | 15-25K contacts | 3.4 |
Usable | Opt-in |
| CRM | Contacts, deals, operator pipeline | 20K+ records | 2.4 |
Needs remediation | Contractual |
| Social Media | Followers, engagement, ad performance | Varies | 3.4 |
Usable | Platform ToS |
| Operator Applications | Apps, CSCS status, pipeline stage | 200-500/year | 3.2 |
Usable | Explicit |
| Facility Records | Attendance logs, class rosters, notes | Per-facility | 1.8 |
Unreliable | Operational |
| Quiz Platforms | Scorecard results, readiness quiz data | New source | 4.4 |
Production-ready | Explicit |
Cannot connect website visitor to member to VALD record. Fix: implement email as universal ID, CRM as system of record.
Cannot segment the full member base. Fix: standardize required fields, validation rules, deduplication across 50+ facilities.
Beyond page views, no engagement-based segments exist. Fix: deploy event taxonomy via GTM.
Cannot use performance improvement data in campaigns. Fix: API integration from VALD to CRM.
Progressive profiling: never ask for more than 3 fields at once. Always offer value proportional to data requested.
Recency, Frequency, Monetary value applied to a subscription + attendance model. Scored at the family level, not the individual. Recency uses attendance, not billing -- a family that pays but does not attend is R1 regardless of payment status.
Days since last attended session
Sessions per month (90-day rolling avg)
Monthly revenue per family
Estimated share of 10,000+ member base by segment
Referral program enrollment, early access to new programs, director relationship deepening. Request NPS and testimonials. These families are the seed audience for look-alikes.
In-person (director), email (personal), referral program
referral_rate >= 1 per quarter
Cross-sell: sibling enrollment, tier upgrade nudge. Community involvement via parent ambassador program. Target: 20%+ enroll a second child within 6 months.
Increase session frequency. Send attendance encouragement. Highlight VALD improvement data tied to training consistency. Invite to parent data walkthroughs.
Attendance declining. Trigger: director outreach -- "We noticed Maya hasn't been in this week. Everything okay?" Re-test reminder. Survey to surface hidden objections.
High-value families who stopped attending. Highest retention priority. CEO-level outreach if needed. These were Champions. Something changed. Director call, personal email, in-person meeting.
Win-back campaign. Feedback survey: "What would bring you back?" If no response after 30 days, reduce outreach. Move to re-engagement suppression after 90 days.
Every parent/family moves through a defined lifecycle. The stage determines content, outreach, and measurement. Data triggers drive transitions, not calendar time.
Known contact, not yet booked evaluation.
Scheduled. Target show rate: 80%+
Same-day enrollment target: 40-50%
First 90 days. Critical habit-formation window.
Past 90d. Attending regularly, completing re-tests.
12+ months. High retention probability.
R5 F4-5 M4-5 AND 1+ referral or NPS 9-10.
Previously active member showing churn indicators. Entry: churn score reaches 60+ OR RFM drops to R2 F2.
Cancelled membership. Win-back campaign at 90 days. Re-permission quarterly. Suppress from active marketing.
| Trigger Event | Automated Action | Timing | Escalation |
|---|---|---|---|
| Evaluation no-show (1st) | "We saved your spot" email + reschedule link | +2 hours | None. One more chance. |
| Attendance drops 50%+ | Director text: personal, not template | +48 hours | Director phone call at +7 days |
| Missed VALD re-test | Email: "Your child's 6-week re-test is due." | +1 day | Director mention at next session |
| Payment failure | Auto retry + soft email: "Quick heads-up" | D1: retry, D3: email | Director outreach at D7 |
| Application abandoned | "You started your application. No rush." | +48 hours | Recruiter email at +14 days |
Four models, each with defined inputs, weighting, thresholds, and triggered actions. Phase 1 is rules-based (implementable in CRM). Phase 2 transitions to statistical models once 6+ months of data accumulates.
Director outreach, same-day enrollment nudge
Continue nurture, send evaluation invite
Email nurture only, do not over-invest
Route to recruiter within 48 hours
Nurture + next webinar invite
Monthly newsletter only
Action: enroll high-propensity families in referral program. Ask for testimonials. Invite to "bring a friend" sessions.
Director phone call within 24 hours. Personal, not automated.
Director text + re-engagement email. Offer welcome-back session.
Automated check-in email. No escalation unless score rises.
No intervention. Standard engagement.
Five seed audiences built from first-party data. Start at 1% expansion. Refresh quarterly. Never exceed 5% for parent audiences.
Find more families like your best families.
Parents who enrolled because of VALD data and clinical proof.
Faith-driven families. Harder to target digitally -- build look-alikes for Sarah (spouse), not Matt.
Small but precise. Layer with title + company targeting on LinkedIn.
Victoria Chen archetype. IFA, PE firms, franchise media.
| Seed | Platform | Start % | Max % | Geography | Test Budget | Target CPA |
|---|---|---|---|---|---|---|
| Champions | Meta | 1% | 3% | 15mi radius per facility | $1,500/fac/mo | < $40 |
| Evidence Seekers | Meta | 1% | 2% | 15mi radius | $1,000/fac/mo | < $50 |
| Covenant Keepers | 1% | 2% | Midwest/Southeast | $800/fac/mo | < $35 | |
| Directors | 1-3% | 5% | US national | $3,000/mo | < $200/app | |
| Franchise | 1-3% | 5% | US national | $2,000/mo | < $30/dl |
Always start at 1%. The closest match is in the first 1%.
Test 14 days minimum before expanding. Need 30+ conversions minimum.
Expand only if CPA is within 1.5x target. Otherwise refine the seed.
Never exceed 5% for parent audiences. Interest targeting performs equally beyond that.
Refresh seed lists quarterly. Stale seeds degrade look-alike quality.
| Suppression List | Reason | Refresh |
|---|---|---|
| Active members | Do not spend acquisition budget on enrolled families | Daily sync |
| Active pipeline | Let email/SMS sequence work. Do not confuse with cold ads. | Daily sync |
| Churned <30 days | Too soon for win-back. Feels tone-deaf. | Monthly |
| Operator applicants | Already talking to recruiter | Weekly |
| Unsubscribed / opted-out | Legal compliance (CAN-SPAM) | Real-time |
| ETS employees + families | Not prospects | Quarterly |
| Competitors | D1 Training, Athletic Republic, Parisi staff | Quarterly |
Phased approach: CRM-led data hub first (months 1-6), full CDP only when scale demands it (100+ facilities or real-time personalization required).
HubSpot Marketing Hub Professional. Unified contacts, lifecycle stages, lead scoring, list segmentation, email automation, ad audience sync.
Add when: 100+ facilities, real-time sync needed, website personalization by segment, multi-location attribution.
| Integration | Source | Dest. | Method | Priority |
|---|---|---|---|---|
| Member records + attendance | GymMaster | HubSpot | API or Zapier | P1 |
| Eval bookings + show/no-show | Booking | HubSpot | API or Zapier | P1 |
| Behavioral events | GA4 | HubSpot | Tracking code + event sync | P1 |
| Custom audiences | HubSpot | Meta | Native ad audience sync | P1 |
| Quiz completions + scores | ScoreApp | HubSpot | Zapier or native | P1 |
| VALD re-test data | VALD | HubSpot | Custom API integration | P2 |
| Operator matched audiences | HubSpot | Native or CSV upload | P2 |
Create a "Family" company object in HubSpot. Associate all parent contacts and child records. RFM scoring, churn prediction, and billing all operate at the Family level.
COPPA considerations (minors ages 8+), state privacy laws across 20+ facility states, cookieless targeting, and server-side tracking architecture.
| Regulation | Applicability | ETS Requirement | Implementation |
|---|---|---|---|
| CAN-SPAM | All email marketing | Unsubscribe in every email, physical address, no deception | Standard email platform compliance |
| CCPA/CPRA | California residents | "Do Not Sell" link, right to delete, right to know | Consent banner + CRM deletion workflow |
| State Privacy Laws | VA, CO, CT, TX, etc. (20+ states) | Varies. Most require opt-out + privacy policy. | Universal banner (strictest standard) |
| COPPA | Athletes ages 8+ (children under 13) | Parental consent before collecting data on children | Parent is account holder. No direct marketing to children. |
Custom audiences from CRM (email, phone). Deterministic matching. You own the data. Most accurate.
Events sent server-to-server via GTM server container. Bypasses ad blockers and ITP. 15-25% more signal captured.
Platform algorithms find users who resemble seed audience. Black-box modeling. Seed data must be consented first-party.
Target based on content consumed (youth sports, ACL prevention, S&C career content). No personal data. Privacy-safe.
Youth sports, CrossFit, NSCA interests. Least precise. Top-of-funnel only.
Three phases over 12 months. Foundation, Activation, Intelligence.
Goal: unified contacts, basic segmentation, audience activation.
Goal: behavioral triggers, operator scoring, server-side tracking.
Goal: predictive models, persona refinement, advanced segmentation.