Baby Growth Percentiles: What the WHO Charts Do and Don’t Tell You

A standard, not an average

Most reference charts in medicine describe what a population happens to look like. The WHO Child Growth Standards do something more ambitious: they describe how healthy children should grow. Between 1997 and 2003 the WHO ran the Multicentre Growth Reference Study, following around 8,500 children in Brazil, Ghana, India, Norway, Oman and the United States — all raised under deliberately favourable conditions: breastfed, non-smoking households, adequate healthcare and nutrition. The striking finding was how similarly children grew across six very different countries when those conditions were met. Growth in the first two years, it turns out, is driven far more by environment and feeding than by ancestry — which is what justifies a single international standard (WHO Child Growth Standards).

The result, published in 2006, is the chart set used in child health records across more than 140 countries, including the UK's red book since 2009 — and it is the data behind the baby percentile calculator, which embeds the WHO's monthly tables for weight-for-age and length-for-age from birth to 24 months. The calculator gives you the number; this guide is about what the number does and does not mean — because the single most common way parents misread a growth chart is to treat a percentile as a score, when it is really one point on a trajectory.

How a measurement becomes a percentile

For every age in months and each sex, the WHO publishes three parameters known as L, M and S. M is the median — the weight or length of the exactly-average child of that age. S is the coefficient of variation, which sets how spread out the healthy range is. L is the interesting one: a Box–Cox power that corrects for skewness, using the method developed by Tim Cole (Cole 1990, PubMed 2354692).

Skewness matters because infant weight is not symmetrically distributed. Take girls at 12 months: the median is 8.95 kg, but the −2 SD boundary sits at 7.04 kg while the +2 SD boundary sits at 11.51 kg. The heavy tail stretches 2.56 kg above the median; the light tail only 1.90 kg below. A plain mean-and-standard-deviation model would put those boundaries an equal distance either side and misclassify children at both ends. The L parameter bends the distribution until it behaves normally, and then a measurement X converts to a z-score with one line of arithmetic: z = ((X/M)^L − 1) / (L × S). The percentile is simply the share of the standard normal distribution sitting below that z.

Two refinements are easy to miss. Length-for-age needs no skewness correction at all — the WHO models it as normally distributed, with L fixed at 1 (the published tables show this directly). And for weight-for-age, the WHO's own software applies a “restricted” adjustment beyond ±3 SD, scaling extreme values against the distance between the ±2 and ±3 SD curves so that a data-entry slip or a genuinely extreme measurement produces a stable, comparable z-score rather than an explosive one. The calculator implements both, matching the WHO's igrowup macro (WHO Anthro software) — so its output agrees with what a clinic's software would print.

Worked example: a 12-month-old girl

Suppose your daughter has just turned one. At her review she weighs 8.6 kg and measures 74.0 cm lying down. The WHO table for 12-month girls gives L = −0.2024, M = 8.9481 kg and S = 0.12268 for weight. Step by step:

X / M            = 8.6 / 8.9481   = 0.9611 (X / M)^L        = 0.9611^(−0.2024) = 1.00806 numerator        = 1.00806 − 1    = 0.00806 L × S            = −0.2024 × 0.12268 = −0.02483 z                = 0.00806 / −0.02483 = −0.32

A z-score of −0.32 corresponds to the 37th percentile: she weighs more than about 37% of healthy 12-month-old girls and less than the other 63%. Her length is even simpler because L = 1: the median is 74.015 cm with S = 0.03479, so z = (74.0 − 74.015) / (0.03479 × 74.015) = −0.01 — the 50th percentile almost exactly. Enter the same three numbers into the baby percentile calculator and you will see both percentiles, both z-scores, and the WHO medians for comparison. A 37th-percentile weight on a 50th-percentile length is an unremarkable, well-proportioned result — comfortably inside the −2 to +2 SD band that covers roughly 95% of healthy children.

Why percentiles move in the first two years

Now extend the example backwards. Suppose the same girl weighed 5.7 kg at 2 months (79th percentile) and 7.9 kg at 6 months (74th percentile) before arriving at the 37th at a year. A fall like that alarms many parents — she has crossed the 50th centile line heading down. In isolation, though, it is one of the most common patterns on any growth chart, and there are four distinct reasons why.

Birth size and genetic trajectory are different things

Size at birth mostly reflects conditions in the uterus — placental function, maternal size, gestation length (the pregnancy side of this is covered by the pregnancy weight gain calculator). Genetic growth potential asserts itself over the following one to two years, so babies born large relative to their genes drift down the chart (“catch-down” growth) while babies born small drift up. A majority of healthy infants cross at least one major centile line in their first two years for exactly this reason. Statisticians would add regression to the mean: children at the extremes of the chart tend, on average, to move toward the middle at the next measurement.

Feeding pattern shapes the curve

Breastfed and formula-fed babies grow differently in the second half of the first year — formula-fed infants tend to gain weight faster after about three months. Because the WHO standards were built on breastfed children, a breastfed baby who seemed to “falter” on older reference charts is often tracking the WHO median precisely, while an exclusively formula-fed baby may ride a somewhat higher weight centile late in the first year. Neither pattern is a problem in itself; they are simply different curves through the same chart.

Measurement noise is real

Infant length is measured lying on a board, ideally by two people, and even in clinic conditions repeat measurements can differ by up to a centimetre — which at 12 months is worth roughly 10–15 percentile points. Weight moves with feeds, wet nappies and which scale was used. Part of any single-visit percentile change is measurement error, which is one more reason clinicians read trends across visits rather than deltas between two points.

Weight and length need reading together

A 20th-percentile weight means something different on a 10th-percentile length (a small, proportionate baby) than on a 90th-percentile length (a tall, lean baby a clinician may want to look at more closely). The calculator reports both percentiles side by side for this reason. This weight-for-stature idea is the infant ancestor of the BMI calculator used from age two onwards — our BMI guide explains where that transition leads.

How to track your baby's growth well

• Measure under the same conditions each time. Same scale, naked or dry-nappy weight, similar time relative to feeds. Consistency matters more than precision — a consistent 50 g bias cancels out of the trend entirely.
• Use completed months. The WHO tables are indexed by completed month, so a baby of 6 months and 3 weeks is “6 months” for chart purposes. The age calculator works this out from the date of birth if you don't want to count.
• Correct for prematurity. If your baby was born early, plot them at their corrected age — age counted from the due date rather than the birth date. Clinics typically correct until at least one year for moderately preterm babies and until two years for those born before 32 weeks.
• Record the numbers, not just the feeling. Keep the actual weights, lengths and dates. Three or four real data points in the percentile calculator tell you far more than a memory that the last visit “seemed fine”.
• Look at both measurements together. A weight percentile only means much alongside the length percentile — proportionate small and proportionate large are both usually fine.

Common mistakes

Treating the 50th percentile as a target. Half of all healthy babies are below the median by definition. A baby tracking steadily along the 15th percentile is growing exactly as they should; percentile envy has no clinical basis.

Comparing numbers across chart systems. A percentile from a CDC-based app and one from the WHO standards are answers to different questions, and can differ noticeably for breastfed babies after three months. If your health service uses WHO-derived charts — as the UK's red book does (RCPCH UK-WHO growth charts) — compare like with like.

Panicking over a single point. One measurement contains growth, feeding timing, and measurement error all at once. The signal is in the trajectory; clinicians look for a sustained trend across two or more visits before reading anything into a change.

Doing arithmetic on percentiles. Percentile points are not equally sized. Around the middle of the chart, a drop from the 50th to the 40th percentile is a quarter of a z-score — barely more than measurement noise. Out at the edge, a drop from the 5th to the 1st is a full standard deviation. The z-score is the honest scale for comparing changes, which is why the calculator reports it alongside the percentile.

When to seek professional advice

The chart is a screening tool, and its thresholds are deliberately simple: a z-score below −2 or above +2 on either measure (roughly outside the 2nd–98th percentile), or a sustained fall across two or more major centile lines, is the conventional cue for a proper assessment. That assessment belongs to your paediatrician, health visitor or family doctor — they will weigh the trend, feeding history, development and family build, none of which a calculator can see. Clinical tools go further than charts, too: paediatric drug doses, for instance, are often scaled by body surface area rather than weight alone. Use the baby percentile calculator between appointments to understand the numbers in the red book — and bring the questions it raises to the people qualified to answer them.