Measurement dictionary
Definitions, formulas, data sources, cadence and interpretation rules.
Metric definitions
All metrics must be calculated from versioned source data. Any definition change must be dated, documented and applied consistently across all gyms.
| Metric | Definition | Calculation | Required data | Reporting cadence | Interpretation |
|---|---|---|---|---|---|
| Exit capture accuracy | Share of observed eligible exits correctly captured and matched to the right member | Correct matched exit events ÷ independently observed eligible exits | Exit event, member ID, timestamp, audit sample | Daily, weekly | Hardware and identity-matching reliability |
| Completed visit rate | Share of valid entry events paired with a valid exit inside the visit window | Completed visits ÷ eligible entry events | Entry event, exit event, member ID, matching window | Daily, weekly | Whether Taango closes the visit loop consistently |
| Frequency change before cancellation | Difference between recent visit frequency and the member's historical baseline before cancellation | Recent 28-day visits ÷ baseline 28-day visits, reported as percentage change | Completed visits, baseline period, cancellation date | Monthly, final | Primary behavioural signal |
| Session duration change before cancellation | Difference between recent median session duration and historical median before cancellation | Recent median duration ÷ baseline median duration, reported as percentage change | Completed visits, duration, cancellation date | Monthly, final | Pilot hypothesis, not a proven predictor |
| Members flagged | Unique members who cross an approved signal threshold in the reporting period | Count of distinct flagged member IDs | Member ID, signal type, threshold, flag timestamp | Weekly | Signal volume and staff workload |
| Operator report open rate | Share of issued reports opened by at least one intended site user | Opened reports ÷ delivered reports | Report ID, delivery status, open event, site user | Weekly | Whether reports reach operator attention |
| Operator action rate | Share of flagged members with a recorded eligible action inside the action window | Flagged members with action ÷ eligible flagged members | Flag, action type, action timestamp | Weekly | Whether the workflow produces intervention |
| Post-intervention return rate | Share of contacted flagged members who complete a visit inside the return window | Contacted members with completed visit in 14 days ÷ contacted flagged members | Action, completed visit, member ID | Weekly, monthly | Immediate behavioural response after intervention |
| Control retention rate | Share of eligible control members still active at the evaluation date | Active control members ÷ eligible control members | Group assignment, active status, cancellation or freeze | Monthly, final | Baseline outcome without pilot intervention |
| Intervention retention rate | Share of eligible intervention members still active at the evaluation date | Active intervention members ÷ eligible intervention members | Group assignment, action status, active status | Monthly, final | Outcome for members exposed to intervention |
| Control vs intervention retention difference | Absolute percentage-point difference in retention between groups | Intervention retention rate − control retention rate | Matched group membership and status | Monthly, final | Estimated intervention effect, subject to design quality |
| Retained revenue estimate | Estimated membership revenue associated with incremental retained members | Incremental retained members × average monthly membership revenue × approved retention horizon | Retention difference, eligible cohort, revenue, horizon | Monthly, final | Commercial value estimate, not recognised revenue |
Standard supporting definitions
Eligible entry
A valid member entry event that occurs during site operating hours, is not a staff or test credential, and is not a duplicate within the de-duplication window.
Completed visit
One eligible entry matched to one valid exit for the same member within the approved visit window. The recommended initial window is five minutes to six hours, subject to site testing.
Baseline period
The eight complete weeks before the measurement period, excluding site closures, known access outages and membership freezes.
Recent period
The most recent complete 28 days before a signal date or cancellation date.
Cancellation
A membership status change to cancelled or terminated. Freezes and payment failures should be reported separately unless the analysis plan explicitly includes them.
Eligible action
A recorded, member-directed contact or service intervention completed within five business days of the flag. Bulk marketing messages do not count unless approved in the pilot protocol.
Return window
Fourteen days after the first eligible action. Also report seven-day and 28-day sensitivity views.
Incremental retained members
Eligible intervention members multiplied by the positive retention difference between intervention and control. Report zero if the difference is negative.
Data-quality dimensions
- Completeness: Required fields are populated.
- Validity: Values meet format and range rules.
- Uniqueness: Duplicate events are removed.
- Timeliness: Events arrive within the reporting service level.
- Consistency: Member identifiers and definitions align across sources.