Methodology Document

Menagerie Welfare Index
Methodology

How Menagerie measures captive exotic animal welfare

v1.0  ·  March 2026 Madcow Venture Company Foundation document
Contents
  1. What This Document Is
  2. Theoretical Foundation
  3. Why This Matters Beyond Individual Animals
  4. Domain Definitions and Data Mapping
  5. Species-Specific Weighting Profiles
  6. Decay Curves
  7. The Confidence Indicator
  8. Data Quality Flags
  9. The Collective Score
  10. Display by Mode
  11. Limitations and Honest Acknowledgments
  12. Future Development
  13. Research Data Access
  14. Changelog

What This Document Is

This document defines the methodology for the Menagerie Welfare Index (MWI), the care score system used throughout the Menagerie app to assess the welfare state of captive exotic animals.

It is written to serve three audiences simultaneously:

Keepers

So they understand what the score means and trust it as a genuine reflection of their animal's welfare, not a gamification element.

Researchers

So they can evaluate the methodology, identify its limitations honestly, and determine whether the dataset is suitable for scientific use.

Developers

So the implementation is consistent, documented, and maintainable as the platform grows.

Theoretical Foundation

The MWI is built on the Five Domains Model of Animal Welfare developed by David Mellor and colleagues, widely adopted in veterinary science, zoo management, and animal welfare research. The Five Domains are:

  1. Nutrition — adequacy of food and water provision
  2. Environment — appropriateness of physical living conditions
  3. Health — absence of disease, injury, and physiological dysfunction
  4. Behavior — ability to express species-appropriate natural behaviors
  5. Mental State — the subjective experience of the animal, integrating the above four

The MWI maps directly to this framework. Each domain is scored independently using available keeper log data, then combined into an overall welfare index score. The Mental State domain is treated as an emergent property of the other four rather than a directly measurable input, consistent with the original Five Domains framework.

This grounding in established peer-reviewed science means the MWI is not an invented metric. It is an implementation of a recognized framework using novel data collection methods.

Why This Matters Beyond Individual Animals

The MWI is designed to function at two levels simultaneously.

Individual level: A keeper understands how their specific animal is doing and what aspects of care need attention.

Population level: Aggregated across all animals of a species in Menagerie, the MWI produces the first continuous longitudinal welfare dataset for captive exotic species ever assembled at scale.

Population-level MWI data can answer questions that have existed in herpetology for decades but have lacked sufficient data to address:

Domain Definitions and Data Mapping

Domain 1

Nutrition

What it measures: Whether the animal is receiving appropriate food and water in appropriate quantities and frequencies for its species and life stage.

Data inputs
  • Feeding log frequency vs. species-appropriate schedule
  • Feeding response quality (eager / normal / reluctant / refused)
  • Weight trend over time (trajectory, not just recency of measurement)
  • Hydration logs
  • Supplement compliance logs
  • Active feeding strike flag
Key principles
  • Frequency is evaluated against species norms, not a universal standard. A tarantula not fed for three weeks is normal. A tegu not fed for three weeks during active season is a signal.
  • Feeding response is weighted heavily as a direct behavioral indicator of animal state, not just keeper behavior.
  • Weight trend matters more than weight measurement recency. A stable weight trend logged monthly is more meaningful than an unknown trend logged daily.
  • Brumation state modifies all Nutrition calculations. An animal in brumation has different nutritional requirements and the score reflects this.
Species-type weighting examples
  • Tegus, monitors, bearded dragons: high frequency feeding expected, response weighted heavily
  • Boas, ball pythons: lower frequency expected, refusal threshold higher before penalization
  • Tarantulas: very low frequency, long gaps normalized
  • Isopods, millipedes: colony-level provision tracking, not individual feeding logs
Domain 2

Environment

What it measures: Whether the animal's physical living conditions (temperature, humidity, lighting, enclosure size, substrate) are appropriate for the species.

Data inputs
  • Temperature readings (basking, warm side, cool side, ambient) vs. species target ranges
  • Humidity readings vs. species target ranges
  • UVB bulb installation date and replacement schedule
  • Photoperiod configuration vs. species requirements
  • Sensor data (source: govee) weighted higher than manual logs for accuracy
  • Enclosure dimensions vs. species minimum requirements
Key principles
  • Sensor data is treated as higher-quality evidence than manual logs because it is continuous, timestamped, and not subject to keeper observation timing.
  • Temperature gradient completeness matters. An animal with only a basking reading and no cool side reading has an incomplete environment picture. The score reflects uncertainty, not failure.
  • Calibration-flagged sensor data (source: govee_calibration) is excluded from Environment domain scoring.
  • The domain evaluates conditions over time, not just the most recent reading. A single out-of-range reading is different from a week of out-of-range readings.

The gradient completeness principle: For heliothermic species (tegus, monitor lizards, bearded dragons), a complete thermal gradient is a welfare requirement, not a preference. The Environment domain penalizes incomplete gradient data more heavily for these species than for thermoconforming species.

Domain 3

Health

What it measures: The absence of disease, injury, and physiological dysfunction, and the presence of positive health indicators.

Data inputs
  • Shed/molt quality and completion (complete, partial, stuck)
  • Presence of problem areas in sheds (retained eye caps, stuck shed on limbs)
  • Waste log consistency and frequency
  • Active health flags (infirmary status, feeding strike, injury)
  • Vet visit recency
  • Medication compliance
  • Body condition score when logged
  • Weight trend as a health indicator (distinct from its role in Nutrition)
Key principles
  • Shed quality is one of the strongest Health domain signals available in captive keeping. A complete clean shed indicates appropriate hydration, nutrition, and environmental conditions. Stuck shed is a multi-domain signal.
  • Waste consistency is a meaningful health indicator for many species. Unusual coloration, consistency, or frequency warrants attention.
  • The absence of vet visits is not penalized. Most healthy animals do not require veterinary care. The domain rewards documentation of vet visits when they occur.
  • Active infirmary status immediately reduces the Health domain score significantly.
Domain 4

Behavior

What it measures: Whether the animal is able to express species-appropriate natural behaviors and whether its behavioral patterns are consistent with positive welfare state.

Data inputs
  • Activity logs (out-of-enclosure time, handling sessions)
  • Enrichment logs and engagement
  • Last seen / sighting frequency
  • Behavioral observations (for mammals: friendly, vocal, playful, irritable, etc.)
  • Feeding response as a behavioral indicator (also evaluated in Nutrition)
  • Brumation behavior appropriate to season
Key principles
  • Behavior is the most species-variable domain. What constitutes appropriate behavior for a whip spider is entirely different from a tegu.
  • Hiding is normal for most exotic species and is not penalized. Hiding combined with feeding refusal and weight loss is a compound signal.
  • Enrichment is weighted differently by species. Enrichment for a tegu is meaningful welfare data. Enrichment for a tarantula is minimal signal.
  • The Behavior domain is the weakest in the current implementation due to data sparsity. It improves significantly as keepers log more behavioral observations.
Domain 5

Mental State

What it measures: The subjective experiential state of the animal and whether it is experiencing positive or negative affect.

Implementation approach: Mental state is not directly measurable through keeper logs. It is treated as an emergent integration of the other four domains, consistent with Mellor's original framework.

In the MWI, Mental State is represented by the overall score integration and is not calculated as a separate domain. Future development may incorporate proxies such as:

  • Feeding response eagerness as a positive affect indicator
  • Behavioral diversity as an enrichment indicator
  • Consistency of activity patterns as a stability indicator

This is an area where the MWI explicitly acknowledges its limitations and invites future methodological development.

Species-Specific Weighting Profiles

Domain weights are not equal across species. The following profiles reflect the relative importance of each domain for major species categories in Menagerie. All profiles sum to 100%. These weights are initial estimates based on husbandry literature and expert keeper knowledge, designed to be updated as population-level data accumulates and empirical validation becomes possible.

Heliothermic Lizards

Tegus, Monitor Lizards, Bearded Dragons, Uromastyx
Nutrition
28%
Environment
35%
Health
25%
Behavior
12%

Environment weighted highest because thermal gradient is physiologically critical for digestion, immune function, and all metabolic processes.

Colubrid and Boid Snakes

Boas, Ball Pythons, Kingsnakes, Corn Snakes
Nutrition
30%
Environment
32%
Health
28%
Behavior
10%

Behavior weighted low as cryptic species with limited behavioral expression in captivity. Feeding response is the primary behavioral signal.

Arboreal Chameleons

Veiled, Panther, Jackson's
Nutrition
25%
Environment
40%
Health
25%
Behavior
10%

Environment weighted highest of any profile. Chameleons are the most environmentally sensitive species in common captive keeping.

Tarantulas and Theraphosidae

 
Nutrition
15%
Environment
35%
Health
35%
Behavior
15%

Nutrition weighted lowest as tarantulas naturally fast for extended periods. Molt quality is the primary welfare indicator.

Amblypygi (Whip Spiders)

 
Nutrition
20%
Environment
40%
Health
30%
Behavior
10%

Humidity is the primary welfare variable. Behavioral assessment is limited by nocturnal, cryptic habits.

Myriapods

Giant African Millipedes, Centipedes
Nutrition
25%
Environment
40%
Health
25%
Behavior
10%

Humidity and substrate depth are primary welfare variables for millipedes.

Amphibians

Tree Frogs, Axolotls, Dart Frogs
Nutrition
25%
Environment
38%
Health
27%
Behavior
10%

Water quality, humidity, and temperature are critical for amphibian welfare. Skin health serves as a proxy health indicator.

Domestic Mammals

Cats, Dogs in Multi-Species Collections
Nutrition
30%
Environment
15%
Health
30%
Behavior
25%

Behavior weighted highest of any profile. Behavioral indicators are the primary welfare signal for cognitively complex mammals.

Exotic Mammals

Sugar Gliders, African Pygmy Hedgehogs
Nutrition
28%
Environment
30%
Health
27%
Behavior
15%

Environment more important than domestic mammals due to specific temperature and humidity requirements. Behavior more important than reptiles.

Decay Curves

Each domain score decays over time as care events age. Decay is not linear — recent events carry more weight than older ones. Decay rates are species-appropriate and state-aware.

Decay Principles

Recency weighting: A feeding log from yesterday is weighted more heavily than one from two weeks ago, even if both represent appropriate care. This reflects that current welfare state is more relevant than historical welfare state.

Species-appropriate decay rates: The decay rate reflects how quickly the absence of a care event becomes a welfare signal for that species.

Species Category Feeding Decay Half-Life Environment Decay Half-Life
Active heliothermic lizards4 days2 days
Boid snakes14 days3 days
Colubrids10 days3 days
Tarantulas21 days5 days
Amblypygi14 days3 days
Millipedes10 days4 days
Amphibians7 days2 days
Domestic cats/dogs2 days7 days

State-Aware Decay Modification

The Confidence Indicator

Every MWI score is accompanied by a confidence value representing how much data underlies the score. A score of 84 based on 847 data points over 18 months means something fundamentally different from a score of 84 based on 6 data points over 2 weeks. Both display as 84. The confidence indicator communicates the difference.

High confidence 30+ data points, 90+ days of history, all four domains represented
Moderate confidence 15 to 30 data points, 30 to 90 days of history, at least three domains represented
Low confidence Fewer than 15 data points or fewer than 30 days of history
Insufficient data Fewer than 5 data points. Score not displayed; message shown instead.

In Advanced mode, confidence is displayed as a percentage alongside the score. In Standard mode, confidence is shown as a subtle indicator. In Simple mode, confidence is not displayed — new animals show "Getting to know [name]" instead of a low-confidence score.

Data Quality Flags

Not all data is equal. The MWI tracks data quality through source flags on every log entry.

Source FlagMeaningScore Impact
manualKeeper-entered observationFull weight
goveeSensor auto-log, validatedFull weight, higher precision for Environment domain
govee_calibrationSensor data during calibration periodExcluded from score calculation
importedHistorical data imported from external sourceReduced weight (70% of manual)
demoDemo account dataExcluded from all aggregate calculations

Research queries automatically filter to source IN ('manual', 'govee') and exclude is_demo = true accounts.

The Collective Score

The Collective Score is the population-level aggregate of individual MWI scores for a species.

What it represents: The mean MWI score for all animals of a species in Menagerie, weighted by confidence level. Low-confidence individual scores contribute less to the aggregate than high-confidence scores.

What it answers: "How well are captive [species] being kept, on average, by Menagerie keepers?"

How it surfaces to keepers: In Advanced mode, a keeper can see their animal's score compared to the species population distribution. For example: "This animal scores 84. The Argentine Tegu population mean is 71. You are in the top 23% of tegu keepers by welfare score."

This is not gamification. It is genuine population context that makes the individual score meaningful.

Minimum population threshold: Collective Scores are only displayed when a species has at least 50 animals with high-confidence scores in the database. Below this threshold, the score is not shown as there is insufficient data for a meaningful population baseline.

Research use: Collective Scores and the underlying distribution data are the primary research output of the Menagerie dataset. They represent the first continuous longitudinal welfare assessment for captive exotic species at population scale.

Display by Mode

The MWI is presented differently depending on the keeper's selected experience level within the app.

Simple Mode
  • Large colored ring showing overall welfare state
  • Single word or short phrase: Thriving / Good / Needs attention / Getting to know [name]
  • No numerical score
  • No domain breakdown
  • One clear action prompt when a domain is significantly low
Standard Mode
  • Numerical score (0 to 100) with ring visualization
  • Domain breakdown showing four segments
  • Score trend over 30 days
  • Contextual interpretation for this animal right now
Advanced Mode
  • Full score with confidence percentage
  • Complete domain breakdown
  • Decay curve visualization per domain
  • Source flag distribution
  • Population comparison vs. species mean
  • Score velocity (7-day rate of change)
  • Raw component weights visible
  • Export button for complete dataset

Limitations and Honest Acknowledgments

The MWI is a tool, not a verdict. Several important limitations apply:

Observer effect

Keepers who use Menagerie more consistently produce higher-quality data and may show higher scores partly because of logging completeness rather than superior husbandry. The confidence indicator partially addresses this but does not eliminate it.

Self-selection

Keepers who use a care tracking app are likely more engaged than the general keeper population. MWI scores may not be representative of captive exotic animal welfare broadly. They represent the welfare of animals kept by engaged keepers who chose to track their care.

Species profile accuracy

The domain weightings are initial estimates based on husbandry literature and expert knowledge. They have not been empirically validated against health outcomes. Validation is a future research priority.

Mental state uncertainty

The Five Domains model acknowledges that mental state in non-human animals is difficult to assess. The MWI's treatment of mental state as emergent rather than directly measured is a principled choice but also a limitation.

Data sparsity in Behavior domain

Behavioral observations are the least consistently logged data type. The Behavior domain score is therefore the least reliable component of the MWI in most cases.

These limitations are published alongside the methodology, not hidden. Scientific credibility requires honest accounting of what a dataset can and cannot tell you.

Future Development

Empirical Weight Validation

As the dataset matures, domain weights can be tested against health outcomes. Do animals with consistently high Environment domain scores show fewer health events? Do animals with high Nutrition domain scores show better weight trajectories? This validation is the path from informed estimate to empirically grounded methodology.

Veterinary Outcome Correlation

With partner veterinary practices, MWI scores can be correlated with clinical health outcomes. This is the most rigorous validation pathway and the one most likely to result in peer-reviewed publication.

Keeper Experience Adjustment

A first-year keeper and a ten-year keeper keeping the same species under the same conditions may produce different log completeness. The MWI could eventually adjust for keeper experience level to isolate animal welfare signal from keeper behavior variation.

Temporal Pattern Analysis

The dataset will eventually support seasonal and longitudinal pattern analysis: how do species-level welfare scores vary by season, by geographic region, by keeper tenure? These patterns are research outputs, not app features.

The HERP Lab Connection

The HERP Laboratory at Eckerd College (herp.eckerd.edu) focuses on how captive environments affect reptile behavior and cognition. Their research questions and Menagerie's dataset are naturally complementary. A formal data sharing agreement would benefit both.

Research Data Access

Menagerie data is collected with keeper consent for research contribution. Keepers who opt in explicitly agree to their anonymized data being used for welfare research.

What Is Available to Researchers

What Is Never Released

Changelog

VersionDateNotes
1.0March 13, 2026Initial specification

Research Inquiries

If you are a researcher, veterinarian, or institution interested in accessing Menagerie welfare data or discussing a formal partnership, please reach out directly.

Contact the Research Team hello@mymenagerie.pet  ·  Subject: Research Inquiry