Why statistics — one cube means nothing
A single cube result tells you almost nothing. Concrete is variable by nature — cement variation, aggregate moisture, batching tolerance, mixing efficiency, placement technique, curing conditions. All of these introduce scatter. A result of 28 MPa on a single cube could be from a genuinely weak batch or from a sampling or testing issue. You can't tell the difference from one number.
IS 456 recognises this. The whole acceptance framework is built on groups of four results, not individual cubes. This isn't arbitrary — it's statistics. The mean of four samples has a tighter distribution than individual results. The standard error of the mean is σ/√n, so the mean of 4 has a standard deviation of σ/2 — half as noisy as a single result.
And underneath all of this sits the normal distribution. If you plot enough cube results from a well-controlled plant, they'll cluster around the mean in a bell curve. IS 456's acceptance criteria are derived from the properties of that curve.
Characteristic strength — what fck actually means
fck is the characteristic compressive strength — the value below which not more than 5% of test results are expected to fall. It's not the mean. It's not the target. It's the lower bound of the distribution at the 5% tail.
Since 5% of a normal distribution sits below mean − 1.65σ, your target mean strength must be:
For M30 with σ = 5.0: target = 30 + 1.65 × 5 = 30 + 8.25 = 38.25 MPa. That's what your mix design must aim for. If you're batching to hit 30 MPa on average, you've already failed — 50% of your results will be below fck.
Standard deviation — Table 8 and your own data
Standard deviation (σ) measures scatter. Higher σ means more inconsistency — your plant is producing wider variation from batch to batch. IS 456 Table 8 gives assumed values when you don't have enough data:
| Grade | Assumed σ (MPa) |
|---|---|
| M10 – M15 | 3.5 |
| M20 – M25 | 4.0 |
| M30 and above | 5.0 |
Reference: IS 456:2000 Table 8 — Assumed Standard Deviation
These are assumed values for when you have fewer than 30 test results. Once you accumulate 30+ results from the same plant with similar mixes, you must compute your actual σ. Most well-run plants in India will see actual σ of 3.0–4.5 MPa for M25–M30, lower than the assumed 5.0. This matters — a lower established σ reduces your target strength and saves cement without compromising quality.
Coefficient of variation — your plant's vital sign
The coefficient of variation (CV) normalises σ against the mean: CV = (σ / mean) × 100. It tells you about process consistency, independent of the grade you're producing.
Excellent. Tight process control. Typical of plants with automated batching and good materials management.
Average. Acceptable but watch for drift. Most well-run Indian RMC plants sit in this band.
Poor. Process is out of control. Investigate materials, batching accuracy, moisture correction, mixing time.
Control charts — catching drift before it fails
A control chart plots your cube results over time and tells you whether the process is stable or drifting. Plot two things on the same time axis: the moving mean of the last 4 samples and the individual results.
Set warning limits at mean ± 2σ and action limits at mean ± 3σ. ACI 214R run rules help you spot trouble before any single point breaches a limit:
- 7 consecutive points on one side of the mean — process has shifted, even if no limit has been breached.
- 7 consecutive points trending up or down — a gradual drift, catch it early.
- Any single point beyond the action limit (±3σ) — stop and investigate.
The point of a control chart isn't to wait for failures — it's to detect when the process starts drifting before it produces a failing result. If your moving mean shifts down by even 1 MPa over three weeks, that's a signal. Find the cause (cement source changed? aggregate moisture sensor drifting? admixture dosing pump clogged?) and fix it.
Worked example — M30, four cubes fail the group mean
M30 grade, σ = 5.0 MPa (assumed, from Table 8). Four 28-day samples come back at 28.5, 31.2, 29.8, 33.1 MPa. Let's run the acceptance check.
Mean = (28.5 + 31.2 + 29.8 + 33.1) / 4 = 122.6 / 4 = 30.65 MPa
fck + 0.825σ = 30 + 0.825 × 5.0 = 34.125 MPa
Rounded to nearest 0.5 → 34.0 MPa
fck + 3 = 33 MPa. Higher value governs: 34.0 MPa
Group mean 30.65 MPa FAILS required 34.0 MPa. That's a 3.35 MPa shortfall. The individual results also need checking: lowest is 28.5 MPa, which is above fck − 3 = 27 MPa, so the individual gate passes. But the group gate fails — this concrete does not comply.
What happens now? Don't just retest. The group of 4 represents a statistically significant sample of the concrete placed. Follow IS 456 Cl 17 for core testing of the in-situ structure. Three cores, uncut strength corrected for size and direction, compared against fck / 0.85 (since core strength is typically 80–85% of cube strength).
Use the concrete acceptance calculator to run these checks on your own data.
What to do when a result fails
A failed cube result is not the end of the world, but it demands investigation — not a blind retest. Here's the protocol I follow on every project:
Verify the test
Was the cube cast and tested correctly?
- · Check the loading rate on the CTM. Was it 14 MPa/min?
- · Check failure pattern — was it a valid conical failure?
- · Check the test certificate for the sample ID against your pour card.
- · Check curing tank temperature logs.
Investigate the mix
Root cause analysis
- · Cement test certificate — was the cement actually the same grade?
- · Batch record — was water corrected for aggregate moisture?
- · Slump at placement — was it within spec or too high?
- · Admixture dose — overdosing superplasticiser increases w/c effectively.
- · Transport time — did the truck wait too long before placement?
Test the structure (if needed)
IS 456 Cl 17 — Core testing
- · Cut 3 cores from the suspect concrete area.
- · Test after 48h conditioning. Correct for core diameter and direction of drilling.
- · Concrete in the structure is acceptable if core strength ≥ 0.85 × fck and average of 3 cores ≥ fck.
- · If cores fail too, escalate to the structural consultant. Load test or retrofit may be needed.
Frequently Asked Questions
Related articles
Amit Haridas
Founder & Proprietor, ConcreteInfo. 25+ years experience in construction QA/QC, concrete technology, and RMC plant operations. NRMCA Certified Trainer (USA) and ISO Lead Auditor.