Optimise a process with response surface methodology. Build a central composite or Box-Behnken design, enter your responses, fit a quadratic model, view a contour plot, and find the optimal settings — all client-side.
Set your factors and ranges, then press Generate design. Enter the response for every run and fit the surface.
Watch: Response Surface Methods and Optimisation
You have found the factors that matter — now you want the settings that maximise the result. Enter the lab to map the response surface.
Choose your factors, their low and high settings, and a design (central composite or Box-Behnken). The calculator generates the run sheet — factorial, axial and centre points — in coded and natural units.
Run the experiment and type in one response per run. The calculator fits a curved (not just straight-line) model that captures how the response bends and how factors interact with each other, and reports how well that model fits your data.
The calculator searches across all the possible settings to find the ones that give the best result (highest, lowest, or hit-your-target), with a score showing how good that solution is. A contour plot shows you a map of how the response changes as you move around in the tested region.
Response surface methodology (RSM) is the Improve-phase tool for fine-tuning a process once you know which factors matter. It fits a curved (quadratic) model to designed experimental data so you can map the response across a region and dial in the settings that maximise, minimise, or hit a target — replacing trial-and-error with a clear optimisation map.
RSM combines a designed experiment with a second-order (quadratic) regression model. By adding centre and axial points to a factorial base, the design can estimate curvature — not just straight-line effects. The fitted surface is then searched for the operating point that best meets your goal, making RSM the natural step after screening has revealed which factors are worth optimising.
Response Surface fits a curved model to your experiment, not just a straight one — it includes squared terms that capture curvature and cross terms that capture how factors interact. Fitting it to your data produces an equation for the whole 'surface' of results across the factor settings. The tool then searches that surface for the combination of settings that gives the best predicted result. Because the model is curved, the best point is often an interior sweet spot rather than pushing every factor to its extreme.
A team optimises a coating process by varying spray pressure (3–7 bar) and conveyor speed (10–20 m/min) across a set of designed experiments. The goal: maximum coating thickness with minimum waste.
The response surface model finds the sweet spot at 5.5 bar and 14 m/min. Above or below these values, performance drops off in both directions.
What to do with this: Set the process to 5.5 bar and 14 m/min and put those in the control plan. RSM found the optimum without testing every possible combination.
Most processes have a sweet spot, not a straight line — push a factor too far and the response falls away again. RSM finds that sweet spot in your real units, so you can lock in settings that maximise yield, minimise cost or defects, or land precisely on a target specification with the fewest possible runs.
Use RSM in the Improve phase of DMAIC after a screening experiment has narrowed the field to two or three influential factors. It is the right tool when you suspect curvature — when a centre-point check shows the response is not linear — and you need to set optimal operating conditions rather than just rank factors.
RSM closes the optimisation loop: it turns the vital few factors confirmed in Analyse into a validated, quantified operating window. The predicted optimum and contour map feed straight into your control plan, capability work and standard operating procedures, giving a defensible, data-backed recipe for the process.
Always run one or more confirmation experiments at the predicted optimal settings to verify the response matches the prediction before committing. Document the optimal operating window and the contour map in your control plan, and set up control charts to hold the process at the new settings. If the confirmation run disagrees with the model, widen or shift the region and run a follow-up design — RSM is naturally sequential.
Combine the paired t-test with run charts, control charts, and a DMAIC project structure to verify and sustain improvements.
A paired t-test is used when the same items (machines, patients, stations) are measured twice — before and after, or under two conditions. Because each pair shares variability, the test eliminates that source of noise and is more powerful than a 2-sample test. A 2-sample (independent) t-test is used when the two groups are completely separate, with no natural pairing between observations.
Data is paired when there is a natural one-to-one correspondence between each before and after value — the same machine, the same patient, the same workstation. Examples: cycle time on 10 machines before and after a process change; blood pressure in 20 patients before and after treatment; inspection times for the same parts under two methods. If the items in the two groups are different, use a 2-sample t-test.
The paired t-test assumes: (1) observations are independent of each other (one pair does not influence another); (2) the differences (after minus before) are approximately normally distributed — for small samples use a normality test; (3) there are no extreme outliers in the differences. Violations of normality with small samples suggest using a nonparametric alternative such as the Wilcoxon signed-rank test.
The minimum recommended is about 10–15 pairs, though the test works with fewer if the differences are normally distributed. The required sample size depends on the expected mean difference, the standard deviation of differences, and the desired power (typically 80%). For a kaizen study where you expect a 10% improvement with moderate variability, 20–30 pairs typically gives adequate power. Use a sample size calculator to plan your study.
A significant paired t-test confirms your improvement worked. Structured training shows you how to sustain it and replicate it elsewhere.
More calculators for statistics, quality, and improvement.