Data Methodology

1. Vacancy data sources

Primary source: MOE annual announcements

Vacancy, applied, and taken counts for each phase come from the Ministry of Education's official P1 Registration announcements, typically published 2-4 weeks after each phase's results day on moe.gov.sg/primary/p1-registration.

The MOE releases typically include:

• A national summary table (all schools, all phases)
• Per-school breakdowns (sometimes only schools with vacancies)
• Ballot outcome statistics for oversubscribed schools

Secondary source: data.gov.sg

For historical years (2023-2025), we cross-reference against data.gov.sg releases. The official dataset is published once per year, typically in Q1 of the following year.

Tertiary source: school websites

For supplementary seats (2CS) and Phase 3, MOE does not always publish school-by-school data. We contact schools directly via email or check their websites for opening notices during the registration window.

2. School identity data

3. Geographic lookup

When you type a postal code on the homepage, we go through this chain:

Step 1: OneMap Singapore API

OneMap is Singapore's authoritative geocoding service, maintained by the Singapore Land Authority (SLA). It works for specific building addresses — about 30% of Singapore postcodes. Examples:

• 521524 → 524A Tampines Central 7 (lat 1.3584, lng 103.9389)
• 238859 → ION Orchard (lat 1.3041, lng 103.8318)
• 039393 → Marina Bay Sands (lat 1.2834, lng 103.8607)

Step 2: Postal district centroid

For sector codes where no specific building exists (about 70% of postcodes), OneMap returns no result. We fall back to the postal district centroid — the average latitude/longitude of schools in that district. We mark these results as "approximate" in the UI.

Step 3: District centroids are derived from real data

For districts with 3+ schools in our dataset, we calculate the centroid from the actual school locations (averaged lat/lng). For districts with fewer schools, we use authoritative hand-curated coordinates.

Examples:

4. Distance calculations

We use the haversine formula for straight-line distance between two coordinates. This is a reasonable approximation for the kind of distances involved in P1 registration (typically 0-3 km).

Important: MOE uses building-outline distance (the walking path between two building entry points), not haversine. For distances under 1 km, the two methods typically agree within 0.1-0.3 km. For distances near the 1 km boundary, this can matter:

We do not have access to MOE's internal walking-path calculation, so we always label our distance as "straight-line" in the UI.

5. Data quality checks

Every refresh goes through these automated and manual checks:

1. Completeness: All 182 schools must have an entry for every year and every phase they participated in
2. Internal consistency: Total vacancy per school-year should sum sensibly across phases
3. Range checks: Vacancy between 0 and 600, applied between 0 and 1,000
4. Sanity checks: Schools with extreme oversubscription (>3x) are flagged for manual review
5. Cross-reference: Compare new data against previous year's value for the same school; flag changes >20% for review

6. Update schedule

7. Known limitations

What we don't have

• Individual student outcomes. We publish school-level statistics only. We do not have access to or publish individual student placement data.
• Application order within phase. MOE only publishes total applied count, not the order in which applications were received.
• Demographic breakdown. We cannot tell you how many PR vs citizen applicants were in a school's pool.
• Walking distance. As noted above, our distances are straight-line, not walking-path.
• Real-time MOE updates. We update within 7 days of MOE publishing, but not in real-time.

What can go wrong

• MOE revisions: Occasionally, MOE revises a number weeks after initial publication. We update within 14 days but there may be a brief inconsistency window.
• School data staleness: If a school changes its principal or contact info mid-year, our site may show the old value until our next refresh.
• OneMap outages: If the OneMap API is down, our postal lookup temporarily falls back to district centroids for all queries.
• Smart apostrophe schools: Singapore school names use both straight (') and curly (') apostrophes. We normalize both, but we may miss a school occasionally. Email data@primarysch.com if you spot one.

8. Version history

Questions?

For methodology questions or to report an issue: data@primarysch.com

For privacy-related questions: privacy@primarysch.com

General enquiries: Contact page →