What The
Cubes Must
Prove: IS 456 Cl 16

Codes & standards referenced

From fresh concrete to a verdict

The acceptance logic sits at the end of a chain, and every link has to hold. Sampling follows IS 456 Cl 15.2 with fresh-concrete practice per IS 1199 (Part 1):2018. Each sample is a set of three 150 mm cubes cast and tested per IS 516, and the sample result is the average of those three cubes. The register then forms groups of four non-overlapping consecutive sample results — samples 1–4, 5–8, 9–12 — in sampling order. No cherry-picking, no sliding a weak result into the next group.

Reference: IS 456:2000 Cl 15.2.2. One sample = three cubes; the sample result is their average.

Which σ applies

The mean gate needs a standard deviation. Until the site establishes one from at least 30 results, use the assumed values of Table 8: 3.5 N/mm² for M10–M15, 4.0 N/mm² for M20–M25, and 5.0 N/mm² for M30 and above. The assumed σ is a stand-in, not a right — start building the project data set from the first pour, and switch to the established value as soon as it exists. A tighter established σ raises the bar your group means must clear; a looser one lowers it. Either way the arithmetic stays honest.

The two gates of Table 11

Gate 1, the group mean: the mean of each group of four must equal or exceed the greater of f_ck + 0.825σ — rounded off to the nearest 0.5 N/mm² — and f_ck + 3. Gate 2, the individual floor: every single sample result must be at least f_ck − 3. A brilliant group mean cannot rescue a single failing result, and one strong result cannot rescue a failing group.

Reference: IS 456:2000 Table 11 as amended through Amendment 4; assumed σ per Table 8.

A worked M25 verdict

Four samples from an M25 pour, tested at 28 days, with each sample result already the average of its three cubes. Assumed σ = 4.0 N/mm².

Now the instructive near-miss. Suppose the four results had been 28.4, 26.9, 31.2 and 25.8. The group mean is (28.4 + 26.9 + 31.2 + 25.8) / 4 = 28.075. That clears f_ck + 3 = 28 — and it still fails, because the governing gate is 28.5. Every individual result clears 22 comfortably, but the mean gate does not care. The group is non-conforming, and the concrete it represents goes down the Cl 16.4 investigation route.

Small pours: a different regime

Table 11's Note 2 handles pours up to 30 m³ where the Cl 15.2.2 frequency yields fewer than four samples. There, the mean of all available sample results must reach f_ck + 4, each individual result must be at least f_ck − 2, and if only one sample is taken, that single result must reach f_ck + 4 on its own. For our M25 example, an 18 m³ pour calls for three samples: their mean must be ≥ 29 N/mm² and each result ≥ 23 N/mm². The higher bar is the price of judging a small data set — with few results, a group mean cannot smooth out one bad batch, so each pour carries more of its own proof.

When a group fails

A failed group does not authorise demolition or silent retesting. Cl 16.3 ties each group to the limited quantity of concrete it represents (a group represents a limited quantity — up to 60 m³). Cl 16.4 then routes the affected concrete to structural investigation: the engineer assesses the actual member, typically with non-destructive testing or cores under Cl 17. Under Cl 17.4.3, core results support acceptance when the average equivalent cube strength is at least 85% of f_ck and no single core falls below 75% of f_ck. Quarantine the records first: batch tickets, sampling identity, curing and machine calibration. Half of "failed" groups trace back to testing practice, not to concrete — and you cannot tell which half without the paperwork.

Check your register

New to the vocabulary of characteristic strength and σ? Part 5 of our glossary covers the testing terms, and the design side of the story — why the target mean sits at f_ck + 1.65σ — is in the IS 10262:2019 walkthrough.

Related articles