How matching works · v3.0

The honest story of
how we pick pets for you.

No black box. No “the algorithm decided.” Every score the Match Agent shows you is the sum of four human-readable decisions — and we'll walk you through all of them.

LAYER 01
Safety
Non-negotiables only
LAYER 02
Fit
Lifestyle, weighted
LAYER 03
Reach
Who's actually available
01What the research told us

Three ideas that changed how we match.

Before writing a line of matching logic, we read the studies. Three findings rewrote the rulebook.

9%
of a dog's behavior is explained by breed
We used to lean on breed to predict temperament. Turns out breed is a weak prior. The individual pet in front of you — their foster notes, their playgroup habits — matters roughly 3–5× more.
→ Morrill et al., Science 2022
94%
of owners who considered surrendering kept their pet — after getting support
Screening harder up front doesn't prevent returns. Post-adoption support does. So we match on fit, then hand you real help — not a 40-question gate.
→ Hill's State of Shelter Adoption 2024
3 days
median time from adoption to a nuisance-behavior return
Most early returns aren't about aggression. They're about barking, scratching, energy — expectations that never got set. So we're honest up front about what a pet is actually like.
→ HASS pilot, n=313,518
02Layer 1 · Safety

A very short list of non-negotiables.

Most adoption sites treat 20 things as deal-breakers — rental, fence, income, landlord call. The research doesn't back any of that up.

Our rule: a match is blocked onlywhen ignoring it would create a predictable safety incident, a legal problem, or at least a 2× return risk we can't fix with a conversation.

Everything else is a soft signal — surfaced honestly, never a slammed door.

Why this matters
Studies (HASS 2024) show that in the average shelter, only 1 in 3 interested visitors leaves with a pet. The top reason isn't unqualified applicants — it's a process that kept turning them away.
✕
Blocks a match
Documented bite history
on this specific pet, not their breed
Child-fear on record
when a child ≤5 lives with you
Species incompatibility
e.g. cat-reactive dog going to a cat home
Lease or breed law conflict
a legal problem, not a judgment
✓
Doesn't block a match
Renting vs. owning
meaningless to retention
No fenced yard
leash walks work for most dogs
First-time owner
we'll bundle you extra support
Apartment living
scored on barking + energy, not square footage
Income level
awareness of resources matters; income doesn't
03Layer 2 · Fit

Twelve dimensions of fit, weighted by what actually predicts joy.

Each pet gets scored 0–100 across twelve lifestyle dimensions. The weights aren't vibes — they come from return-reason data. Energy mismatch is the #1 reason dogs come back to shelters, so it's our heaviest weight. Coat color is zero.

The weights we use
Sum = 100%
Energy / exercise fit
18%
#1 behavioral return driver
Compatibility with children
15%
if applicable
Compatibility with other pets
14%
~22% of dog returns
Trainability × your experience
10%
Morrill-validated trait
Time alone tolerance
8%
early-week returns
Size × space × home type
8%
Hawes 2021
Medical / financial fit
6%
set expectations
Grooming / shedding
6%
adopter preference
Barking / vocality
5%
neighbor nuisance
Age / life stage
5%
Powell 2021
Drool / prey drive / other
3%
lifestyle quirks
Climate × cold/heat tolerance
2%
lower evidence
Evidence tier
We trust what we've seen more than what a breed book says
When a foster has lived with a pet for two weeks, we mark their notes as observed. A breed-level assumption gets flagged as a prior. Both count toward the score — but priors lower the confidence we show you.
Adaptive
Weights move when a category doesn't apply to you
No kids? The 'children compatibility' weight redistributes across the rest. You don't get penalized for things that aren't in your life.
Honest gaps
Missing data is called out, not smoothed over
If a pet's barking level was never assessed, we use a breed-based guess — and tell you we're guessing. You deserve to know what we know.
04Worked example

Why Buddy is a 94 for Maya.

Here's every number the Match Agent used. Nothing hidden.

Buddy
Buddy → Maya
3y Golden Retriever mix · Portland foster · 14 days in care
95
Strong match
DimensionWeightBuddyEvidencePointsWhy
Energy / exercise fit21.2%98Observed+20.8Foster logged 2h daily; Maya walks 90m + weekend hikes
Compatibility with children————Maya has no kids — weight redistributed
Compatibility with other pets16.5%95Observed+15.6Foster noted Buddy ignores cats; Maya has one
Trainability × experience11.8%94Observed+11.1Responds to 6 cues; Maya has adopted before
Time alone tolerance9.4%92Observed+8.7Maya works from home 4d/week
Size × space × home type9.4%92Observed+8.745 lb, medium; foster confirms apartment-settled
Medical / financial fit7.1%95Observed+6.7No chronic conditions; Maya has pet insurance
Grooming / shedding7.1%88Breed prior+6.2Moderate shed estimated from breed mix
Barking / vocality5.9%97Observed+5.7Foster: barks less than 3×/day
Age / life stage5.9%95Observed+5.63y, in the sweet spot
Drool / prey drive3.5%90Breed prior+3.2Low prey drive estimated from breed mix
Climate tolerance2.4%92Breed prior+2.2Portland wet; coat handles it
Final: 94.5 out of 100. Each row is score × weight, summed. Rounded to 94 for display.
Total
94.5
05Layer 3 · Reach

The page adapts to who's actually available.

A rigid algorithm says “no results” and leaves. We widen, substitute, and tell you what's going on.

30+
Plenty of candidates
Apply your full preferences. Show the top 10 ranked matches.
10–29
Narrowing in
Downgrade soft preferences to warnings (e.g. "a little larger than you wanted"). Keep only safety filters hard.
< 10
Thin pool
Surface similar-but-available substitutes using trait similarity — not breed labels.
0
Nothing today
Offer a waitlist, widen the geographic radius, and suggest a foster-to-adopt trial.
06How the score is shown

You'll see words, not just a percent.

A “76% match” implies precision we don't have. Instead, every pet gets a four-level label — and one concrete reason to say hello.

Austin Pets Alive! found that 75% of adopters said the in-person meeting — not any score — was their #1 decision factor. The score gets you in the room. The pet takes it from there.

Strong match
The agent is highly confident this pet fits multiple parts of your life.
90–100
Good match
A solid fit on most dimensions. A few trade-offs to discuss in the meet.
75–89
Worth meeting
Some real misalignment, but the pet has qualities that could win you over.
60–74
Likely not a fit
Surfaced for completeness. We'll tell you why this one is a stretch.
< 60
07What we don't do

The things we deliberately left out.

A list of screening practices that feel like quality control and actually just keep good homes away.

✕
Breed as a personality predictor
Breed explains ~9% of a dog's behavior. We treat it as a hint, not a label.
✕
In-shelter behavior tests
False-positive rates of 29–84% (Patronek & Bradley, 2016). We use foster and playgroup data instead.
✕
Landlord phone calls
No peer-reviewed evidence they predict anything — documented bias against renters and people of color.
✕
Minimum income checks
Fee-waived adoptions retain 93–95% at 6 months — same as full-fee. We share resource awareness instead.
✕
“Hypoallergenic” as a filter
No dog breed is scientifically hypoallergenic (Vredegoor 2012). We show a relative allergen level and suggest a foster trial.
✕
A percent-match that pretends to be precision
We round to a label and a reason. The meet-and-greet is the real decision.

Now that you know how it works —
let's close your loop.

The questionnaire takes about six minutes. Every answer you give gets treated the way we just described.