Skip to main content
Childery

How we rate daycares

What the 1–5 star Overall rating means, what data goes into it, what's missing, and where the limits are. Written for a parent who wants the honest version, not the marketing version.

Data current as of May 2026.

One number can't tell you whether a daycare is right for your kid. The rating tells you what we measured. It doesn't tell you what your child's specific teacher is like, what curriculum the room uses, or whether the place feels right when you walk in. Visit. Talk. Decide.

The short version

Each daycare gets up to three star ratings, all on a 1–5 scale:

  • Overall — the headline. A blend of Process and Structural quality.
  • Process — how rich the day-to-day care is. Built from the state's QRIS rating, third-party accreditations (NAEYC, NAC, NECPA, NAFCC), and Head Start CLASS observation scores. About 35% of US daycares have at least one Process signal; the rest don't because state QRIS participation is voluntary, accreditation is uncommon, and CLASS only covers Head Start.
  • Structural — how the facility actually performs against safety and quality benchmarks. Where we have inspection histories, license-status records, or teacher-credential data, we use that for a per-facility score. Where we don't, the rating reflects the state's regulatory baseline (ratios, group sizes, teacher-credential rules).

When both are available, the Overall rating is 0.7 × Process + 0.3 × Structural, rounded to the nearest star and clipped to the 1–5 range. When only one is available, that's the Overall.

What goes into Process

Six possible signals. We average whichever ones are present for a given daycare:

  1. State QRIS rating (mapped to 1–5). Most states run a Quality Rating and Improvement System; participation is voluntary in most, mandatory in a few. The 1–5 mapping uses each state's published tier definitions.
  2. NAEYC accreditation. National Association for the Education of Young Children. If a daycare is accredited, this contributes a 5; otherwise it doesn't contribute. Accreditation is a high bar — fewer than 10% of US childcare centers hold it.
  3. NAC accreditation. National Accreditation Commission for Early Care and Education Programs. Same scoring as NAEYC: present = 5, absent = doesn't contribute.
  4. NECPA accreditation. National Early Childhood Program Accreditation. Same scoring.
  5. NAFCC accreditation. National Association for Family Child Care, the main accreditor for family-based providers. Same scoring.
  6. Head Start CLASS scores. The Classroom Assessment Scoring System is the federally mandated observation tool for Head Start programs. It produces three subscores (Emotional Support, Classroom Organization, Instructional Support). We average them, then map the 1–7 scale linearly to 1–5. The CLASS evidence base is real but small — see class-validity for the literature.

What goes into Structural

Two layers, with the more specific one preferred when available.

Provider-level (preferred). When we have direct facility-level data, we use it. Three sub-signals:

  • Inspection / violation / complaint history. From state licensing portals — the share of inspection visits that found violations, complaints filed against the facility, and (where data exists) injury reports.
  • License status. Active and in good standing → 5 stars. Probation, restrictions, substantial-compliance flags → 1 star. Closed or suspended facilities are removed from the directory entirely.
  • Lead teacher credentials. Available for Florida and Louisiana so far, plus public-school pre-K classrooms in 11 other states (DOE-operated and charter pre-K facilities operate under teacher-certification rules that differ from private licensed centers).

State regulatory baseline (fallback). When provider-level data isn't available, the rating reflects the state's regulatory floor: child-to-staff ratios, group-size limits, teacher education and ongoing-training requirements. Each variable is anchored against fixed national benchmarks (NAEYC accreditation standards and the National Institute for Early Education Research's quality benchmarks). For public-school-hosted pre-K classrooms in eleven states, we substitute the documented program standards instead of the state floor — a Hawaii DOE pre-K classroom, for example, requires bachelor's-degree teachers with state Early Childhood licensure and meets the NIEER ratio benchmark, both of which are higher than the state's licensed-private-center floor.

What we don't include — and why

  • Parent reviews. The trust value is high; the fairness cost (review-bombing, fake astroturfing, unmoderated language) is also high. We don't run reviews until we can run them rigorously. See what-we-dont-track.
  • Subsidy participation flags. Whether a daycare accepts subsidies isn't a quality signal — it's a business-model and state-policy signal. We surface this on the daycare page as a filter for parents who need it, but it doesn't go into the rating.
  • Food program participation (CACFP). It's a nutrition signal, not a structural-quality signal.
  • A "better than predicted" residual bonus. An earlier version of the methodology added a +0.5 star bonus when a daycare's score was higher than a regression predicted from local demographics. The launch-gate audit showed this didn't meaningfully reduce demographic correlation in the ratings — and on a few demographics, slightly increased it. We removed it.
  • Parent demographics. We don't collect them. Period.

What the ratings can't tell you

Things that matter a lot for an individual child but aren't in the data:

  • Your child's specific teacher. Quality varies between rooms in the same daycare more than between daycares.
  • Curriculum or pedagogy. Reggio, Montessori, state-specific frameworks — none of this is in the licensing record.
  • Day-to-day classroom feel. Site visits are the only way to know.
  • Operating hours. State licensing rarely tracks this. We're working on it.
  • Capacity, age range, and license status — these vary by state in how completely they're captured. Where we don't have them, the daycare page shows — rather than a guess.

These limits surface explicitly on every daycare page in the "What we don't know" block, so the absence is visible at the moment of decision rather than buried.

How often the ratings update

Quarterly. The pipeline is reproducible end-to-end: every CSV input is public or scraped from public sources, every transformation is in code/calculate_ratings_v3.py in the project repo, and the segregation audit is regenerated alongside the ratings on every refresh.

Why we publish the audit

A rating system that ranks businesses based on neighborhood-correlated inputs is at risk of reinforcing segregation in childcare access. We checked. The audit is on its own page: segregation-audit. The short version: within any given state, no rating component correlates with any of 15 ZCTA demographic variables at |r| ≥ 0.03. The audit threshold is 0.25 — we have ~10× headroom on the within-state test.

Sources and reproducibility

The full audit, the Python pipeline, and the per-state structural-comparison CSV all live in the project repo. The exact list of audit tables produced on each rating refresh:

  • segregation_audit_v3_2026-05-07.csv — pooled and within-state |r| with each demographic, by score type
  • state_segregation_v3_2026-05-07.csv — per-state segregation diagnostic
  • state_variation_v3_2026-05-07.csv — which states have within-state variation on Process / Structural / both / neither
  • state_rating_distribution_2026-05-07.csv — star histogram per state
  • structural_v3_distributions_2026-05-07.csv — provider-level structural mean and standard deviation per state

If you find a number that looks wrong, submit a correction — we publish what we act on.