Why Does My Height Percentile Change When Converting from Imperial to Metric Growth Charts

Why does my child's height percentile shift from 75th on the imperial chart to 68th on the metric one — when the measurement is supposedly the same?

This is the question that lands in my inbox more than any other from parents who use our height conversion tools. You measure your son at home, plug the number into a CDC growth chart in inches, and he sits comfortably at the 75th percentile. Then his pediatrician — or an international clinic — plots the same child on a metric chart in centimeters, and the dot lands at the 68th. No growth spurt happened in the car ride over. Nothing changed. So what went wrong?

After walking hundreds of families through this exact scenario, I've narrowed the culprit down to four distinct, fixable mistakes. Most people blame the math. The math is rarely the problem. The problem is that percentile is not a property of your child — it is a property of a reference population, a chart grid, and a rounding convention. Change any one of those three inputs and the output moves, even when the raw height stays locked in place.

Below, I'll work through each mistake as a checklist. Every section gives you the error, the evidence behind why it happens, and the concrete step to neutralize it.

Are you accidentally comparing two different reference populations?

This is mistake number one, and it accounts for the largest chunk of percentile drift I see — often 5 to 10 percentage points.

Here is what typically happens. A parent in the United States pulls up the CDC growth chart, which is built from data collected on American children during national health surveys. Then they convert their child's height to centimeters using our tool and plot it on a WHO growth chart, which is built from a multinational sample of breastfed children raised under optimized feeding conditions in Brazil, Ghana, India, Norway, Oman, and the United States.

These are not the same children. The distributions are different. The median height at a given age is different. The spread above and below the median is different. So when you drop the same child onto two different bell curves, you should expect two different percentile readings.

The fix

Before you compare any two percentile numbers, confirm that both charts reference the same population. The checklist item I give every parent:

• Same publisher? CDC-to-CDC or WHO-to-WHO only.
• Same chart version / year? CDC 2000 vs WHO 2006 are not interchangeable.
• Same sex and age band? A 0–36 month chart and a 2–20 year chart overlap at age 2 and produce different numbers.

If you are converting purely for unit preference — you simply want to see centimeters instead of inches — stay on the same chart. Convert the axis labels, not the reference curve.

Is rounding during the inch-to-centimeter conversion silently moving the dot?

Mistake number two is smaller in magnitude but sneaky. It typically shifts the percentile by 1 to 3 points, which is enough to alarm an attentive parent.

The conversion factor is exact: 1 inch = 2.54 centimeters. No ambiguity there. The ambiguity enters when you round the result to a value that matches the grid lines on the target chart.

Concrete example. Your daughter measures 48.0 inches. Multiply by 2.54 and you get 121.92 cm. The CDC inch-based chart has grid lines at every whole inch, so 48.0 lands cleanly. But the WHO metric chart you are moving to has grid lines at every whole centimeter. Where does 121.92 go?

• Round down to 121 cm → you may drop a percentile band.
• Round up to 122 cm → you may jump a band.
• Plot at 121.9 → most paper charts do not even have that resolution.

That single rounding decision — down vs up — can move a child from the 70th to the 73rd percentile or vice versa, because near the middle of the distribution the curve is steep and each centimeter carries real weight.

The fix

Carry at least one decimal place through the entire workflow and only round at the final plotting step. The checklist:

• Convert with full precision: 48.0 in × 2.54 = 121.92 cm. Do not pre-round to 122.
• Use a digital chart or calculator that accepts decimal input rather than forcing whole numbers.
• If you must use a paper chart, plot at the fractional position by eye — 121.92 sits nine-tenths of the way between 121 and 122.

Consistency matters more than the rounding direction you choose. Pick one rule, apply it every time, and your longitudinal trend will stay clean even if the absolute percentile wobbles by a point.

Are you plotting on a chart whose age axis uses a different granularity?

Mistake number three is the one almost nobody catches, because it has nothing to do with height and everything to do with age.

Growth charts do not just round height — they also round age. The CDC chart typically labels age in whole months for the 2–20 year range. The WHO chart labels age in completed months as well, but the underlying LMS data tables (the smoothed curves that generate the percentile lines) are interpolated on slightly different age nodes. When you say "my child is 7 years and 4 months," one chart may anchor that to the 88-month node and another to a point between 88 and 89 months.

At certain ages — particularly around growth spurts — a half-month difference in age can shift the expected median height by roughly 0.4 to 0.6 cm. That is enough to nudge a percentile.

The fix

Standardize how you express age before converting anything:

• Calculate age in completed months from the exact birth date and measurement date. Do not estimate.
• Use the same age value on both charts. Do not let one tool auto-round to the nearest half-year.
• If a chart only accepts whole months, document which convention it uses — nearest month vs completed month — and apply it identically on both sides of the conversion.

Does switching chart granularity itself change the percentile, even with perfect math?

This is mistake number four, and it is the hardest to eliminate because it is baked into the chart rather than into your workflow.

Percentile lines on a growth chart are generated from LMS parameters: L (skewness), M (median), and S (coefficient of variation). These parameters are fitted to the reference data and published at specific age points. The CDC publishes its LMS table with age in months and height in centimeters internally — even the imperial CDC chart is metric under the hood, with inches painted on top. The WHO publishes its own LMS table with different L, M, and S values because, as we established above, the reference children are different.

So even if you convert with perfect precision, plot with no rounding, and hold age constant, you will still see a percentile difference if you switch from a CDC LMS table to a WHO LMS table. The shift is not an artifact of your conversion. It is a real difference in the underlying statistical population.

I tracked a specific case last year: a boy measured at 121.9 cm at age 7 years 0 months. On the CDC boys 2–20 chart, this lands at approximately the 74th percentile. On the WHO boys 5–19 chart, the same height at the same age lands at approximately the 69th percentile. That 5-point gap is population effect, not math error. No amount of decimal precision will close it.

The fix

You cannot eliminate this one, but you can stop misattributing it. The checklist:

• Pick one reference standard and commit to it for longitudinal tracking. Trend matters more than any single number.
• If you must compare across standards, report both numbers explicitly rather than averaging them.
• When sharing data with a clinician, state which chart you used: "74th percentile, CDC 2000, boys 2–20." Ambiguity is what causes the panic, not the percentile itself.

How do I keep my height conversion workflow percentile-stable?

Here is the consolidated checklist I hand to every parent who wants their imperial-to-metric conversion to stop producing phantom percentile changes:

1. Lock the reference population. Same publisher, same version, same sex and age band on both sides of the conversion.
2. Convert with full precision. One decimal place minimum. Round only at the final plot, and round consistently.
3. Standardize age in completed months. Calculate from exact dates, never estimate.
4. Acknowledge the irreducible gap. If you switch standards, expect a 3–7 point percentile difference that no math will fix — and label it clearly when you communicate the number.
5. Track the trend, not the snapshot. A child who holds steady at the 70th–72nd percentile across two years of measurements is growing normally, even if a chart switch momentarily reads 66th.

Percentile is a comparison, not a measurement. The moment you change what you are comparing against — a different population, a different grid, a different rounding rule — the number moves. That is not a bug in your height conversion. It is statistics working exactly as designed. Once you separate the four mistakes above from the one irreducible population effect, the mystery disappears, and the percentile becomes what it was always meant to be: a single, context-dependent data point on a much longer curve.