The first two articles in this series covered the figures that decide the build and the figures that decide whether it fills. This one covers the seven that decide whether it lasts.
They are operating figures, and they are the ones investors look at least, because by the time they matter the deal is done and the school is someone else’s problem. A school that opens well and fills on schedule can still fail in year six, and when it does, the cause is almost always in one of the seven numbers below.
As in the first two parts, each figure carries a line stating where it comes from, and none travels above its basis.
9. Staff cost as a percentage of revenue
Healthy range: 45 to 55 per cent at maturity.
Below 45 per cent, look closely at class sizes and teacher quality, because the saving is coming from somewhere that will show up in results. Above 60 per cent, the school has either overstaffed against enrolment or is carrying a salary structure its fee level cannot support. In start-up years this figure will exceed 60 per cent by design, which is why the funding figure in Part One matters. It is also the largest single line in the operating cost figure there, so every point it moves is a point of margin.
Curro, which publishes the cleanest series, ran employee costs at about 52 per cent of revenue in 2018 and 46 per cent in 2024 as its estate matured, which is the band exactly.
Basis. GSE practitioner range. Curro ratios computed from its Annual Integrated Report 2024, party-published.
10. Teacher attrition
Typical international school turnover sits between 17 and 25 per cent a year, with wide variation between schools.
Three sources anchor that. A study of schools in the Near East South Asia region published in the Journal of Research in International Education found an average of 17 per cent between 2006 and 2009, with individual schools ranging from no turnover at all to 60 per cent, and the strongest predictor of who stayed was whether teachers saw the Head as supportive. ISC Research’s 2024 teacher movement survey, with 2,539 responses across 59 countries, found most schools can expect to keep a teacher for between one and six years. And GEMS Education, in a 2026 interview reported by Edarabia, put its own portfolio at about 17 per cent and described it as the lowest on record.
Treat 17 to 25 as the central band, not the full range. Regional studies have put the average anywhere from about 14 per cent across ECIS member schools in 2006 to about 32 per cent across East Asia in 2009 to 2010, so a school at the edge of the band is not necessarily an outlier for its region.
Two cautions. First, the group figures are self-reported with no standard methodology behind them, so comparisons across operators are indicative at best. Second, low attrition is not automatically good. A school with 5 per cent turnover and flat academic results has a different problem to one at 25 per cent. What matters is who leaves. Losing 20 per cent of your weakest teachers is a healthier year than losing 10 per cent of your strongest.
Basis. Published and attributed. Mancuso, Roberts and White, 2010; Odland and Ruzicka, 2009, for the ECIS figure; ISC Research, 2024; GEMS Education as reported by Edarabia, 2026.
11. The real cost of replacing one overseas teacher
USD 10,000 to 20,000 before the new appointment teaches a single lesson.
Recruitment fair fees, flights, shipping, settling-in allowance, visa and medical processing, plus the hidden cost of a term of reduced effectiveness while the new hire learns the school. Relocation industry data puts the move alone at USD 5,000 to 15,000 for a single teacher, before the school’s own recruitment cost is counted. Multiply the total by an attrition rate and retention stops being a human resources topic and becomes a line in the budget.
Basis. GSE estimate. No operator publishes this figure. The build-up is recruitment fee, flights, shipping, visa and medical, settling-in allowance; the relocation component is consistent with published industry data for international teaching moves.
12. Family retention
Strong performance sits at 85 to 90 per cent.
GEMS Education, in the same 2026 interview, reported around 85 per cent across its portfolio against 71 to 72 per cent two decades earlier. A five point improvement in retention is worth more than most marketing campaigns and costs less. It also compounds. In a 1,200 student school, moving from 82 to 87 per cent retention keeps sixty additional students enrolled each year, which is close to two full year groups over a decade.
Nord Anglia gave the same picture from the other direction when it listed: an average student tenure of three and a half years, which implies annual retention in the high eighties.
Basis. GSE practitioner range. The GEMS figures are as stated by GEMS, reported by Edarabia and Khaleej Times, 2026. Nord Anglia tenure from its F-1, 2014, party-published.
13. Fee collection and working capital
Arrears above 3 to 5 per cent of billed fees signal an admissions or affordability problem rather than an accounts problem.
Schools collecting termly rather than annually carry materially more working capital risk. Cash timing kills more schools than profitability does. Understand when the money actually arrives, not when it is invoiced.
The two published reference points bracket the line. Curro booked expected credit losses of about 3 per cent of revenue in the first half of 2024, at the threshold. Nord Anglia, at listing, had already received about 55 per cent of the coming year’s school revenue before the year began, which is what a fees-in-advance model looks like and why its arrears barely register.
Basis. GSE practitioner range. Curro from its H1 2024 results; Nord Anglia from its F-1, 2014. Both party-published.
14. Occupancy cost as a percentage of revenue
Sustainable range: 12 to 20 per cent for leased premises.
Set against the operating cost figure in Part One, that leaves a leased school with 10 to 20 per cent after rent in a good year. Above 25 per cent the school is working for its landlord, and the position rarely improves because rent escalates while fee increases are often regulated. There is a longer argument about why school rent cannot be benchmarked against commercial property, and it is worth reading before a number is agreed. In the GCC, fee caps tied to inspection outcomes make this constraint sharper than investors from other sectors expect. A school that owns its building does not escape the cost; it moves it into the asset, which is the subject of the PropCo and OpCo article.
Basis. GSE practitioner range. No operator publishes rent as a percentage of revenue; Nord Anglia books property costs inside cost of sales without breaking them out. The published evidence on school leases is on the yield side, covered in GSE’s cap rates article.
15. EBITDA margin at maturity
Realistic range: 25 to 35 per cent for a well-run single school, measured before occupancy cost.
State the basis before quoting the figure. The same school reads as 30 per cent to an owner-occupier and 15 per cent to a tenant paying market rent, and both numbers are correct. Established groups with scale and owned assets can exceed the range. Models projecting 40 per cent in year five are usually assuming a fee increase environment the regulator will not permit, or a staffing ratio inspection will not accept.
The listed groups sit almost exactly where the tenant band predicts. Nord Anglia reported adjusted EBITDA margins of 24.9 per cent in 2013 and 24.2 per cent in 2016; Curro has run at 22 to 25 per cent since 2018; ADvTECH’s schools division reported 24.7 per cent in 2025. All three are struck after property costs and blend ramping schools with mature ones. They validate the lower band, not the 25 to 35, and anyone who quotes them as a ceiling for a single mature school before rent has mixed two different numbers.
Basis. GSE practitioner range for a single mature school before rent, as published in GSE’s EBITDA benchmarks article. Group margins from Nord Anglia (F-1, 2014, and FY2016 results), Curro (Annual Integrated Report 2024) and ADvTECH (2025 results), party-published, after property costs.
How to use these numbers
Across the three articles, take any project in front of you and test three relationships before anything else.
- Does the fee level support the facility specification, with room left for rent and return?
- Is the funding sufficient to reach break-even occupancy, not just to complete construction?
- Does the enrolment ramp assume a year three that a real school could actually deliver?
Most failed school investments do not fail because the market was wrong. Demand for international education is well documented and still growing. They fail because a plausible-looking model contained one or two assumptions that no operator would have signed off on.
Read next
- International Education by the Numbers, Part One. The four figures that decide the build.
- International Education by the Numbers, Part Two. The four figures that decide whether the school fills.
- EBITDA Benchmarks: How Profitable Are International Schools Really?. The margin evidence behind figure fifteen.
- The PropCo / OpCo Model in School Development Explained. Where the rent goes when the building is owned.
- Governance Structures That Attract Education Investors. The board that holds the plan through the years these figures describe.
To talk about your school project, reach us through the contact page.