Statistical process control in your browser

Plot individual measurements and their moving ranges with control limits from MR̄/d₂, and see which run rule flagged every point.

Average and range charts from rational subgroups, with A₂, D₃ and D₄ computed for your subgroup size rather than looked up.

Average and standard-deviation charts for larger subgroups, with A₃, B₃ and B₄ from the exact c₄ rather than a table that stops at 25.

Fraction-defective chart with true stair-step limits when sample sizes vary, and the lower limit clamped honestly at zero.

Number-defective chart for a constant inspection size, with limits np̄ ± 3√(np̄(1−p̄)).

Poisson count chart with the low-count warning NIST prints, and the square-root transforms for when the normal approximation fails.

Defects-per-unit chart with limits recomputed for every inspection size, so the staircase is drawn rather than averaged away.

A run chart about the median with the four runs-test p-values, plus the full Nelson, Western Electric and ISO rule sets side by side.

Descending bars with a cumulative-percentage axis, a configurable "other" bucket and the 80% cut line.

A histogram with four bin rules, a normal probability plot and the Anderson-Darling statistic with its p-value — reported together, never as a bare pass or fail.

Capability from raw subgrouped data, with within and overall sigma computed side by side and the estimator named on the figure.

The log-likelihood profile, λ̂ by maximum likelihood with its confidence interval, and capability recomputed on the transformed scale.

EWMA with exact time-varying limits by default and NIST’s steady-state limits one click away, so both published conventions are reachable.

Tabular CUSUM with both accumulators and a decision interval, from a reference value you can state in sigma or in measurement units, plus the V-mask design arithmetic — explained rather than drawn.

Crossed Gage R&R by both published methods, printing %Contribution, %StudyVar and %Tolerance with their denominators named, and ndc truncated the way the manual says.

Bias against a reference with its t-test, and a linearity regression across the operating range with slope and intercept confidence intervals.

Lot size and AQL to a sampling plan, with the code letter and inspection state that produced it, an OC curve, and the switching rules as a state machine.

Every Shewhart constant, computed from the normal distribution rather than copied from a table, for subgroup sizes from 2 to 100 — with the derivation and the source for each one.