Less guesswork. Better experiments.

DOE means Design of Experiments. Several inputs are varied according to a planned structure to investigate their effects and how they interact.

When several controllable inputs may affect an outcome and you need to learn which settings help. You need a clear question, measurable outcomes and controlled experimental conditions.

One-factor-at-a-time trials do not systematically reveal whether the effect of one factor depends on another. A factorial design deliberately tests those combinations.

You should understand your process and measurement method. The guide explains the basics. Complex designs, formal evidence and reliable optimization require additional statistical expertise.

The app uses 2 to the power of k combinations for k factors. Multiply by the runs per combination. Three factors with two runs each give 16 runs. Whether this is sufficient for a specific detection goal needs separate sample size planning.

It includes the first run. A value of 2 means each combination is executed twice independently, not two additional runs after the first.

Yes. Two materials or methods are possible. Enter their names as level − and level +. The signs are codes, not quality ratings.

Far enough to reveal relevant changes while staying within feasible and meaningful limits. Narrow ranges may hide effects; very wide ranges may introduce different mechanisms or unacceptable conditions.

Not in this app. Center points can indicate curvature for quantitative factors. They need an appropriate extended design and analysis.

An independent replicate uses a new experimental unit and re-established conditions. Repeated measurements of the same unit mainly examine measurement variation. Do not treat them as independent process runs.

To reduce systematic alignment of time-related disturbances with factor levels. The app shuffles complete runs before execution. Once results are entered, order is locked.

Choose an appropriate structure such as split-plot. Do not simply rearrange these runs for convenience and then interpret them as a fully randomized experiment.

Analysis waits until every run has a valid numeric response. Empty fields remain missing; zero is a real measurement. Decimal points, decimal commas and scientific notation are accepted, but thousands separators are not.

Each project has one response. For multiple outcomes you can use separate project files with the same design. The app does not automatically optimize trade-offs.

Positive means the response increases on average from level − to level +. Negative means it decreases. Whether this is desirable depends on your objective.

The effect of one factor changes depending on another. More pressure might have little benefit at low temperature but a large benefit at high temperature. Pressure must then be interpreted in context.

Not necessarily. Bars are ranked by absolute effect size without a significance test. Even a large observed effect needs to be assessed against variation, experimental conditions and technical plausibility.

Simplified automatic model selection can create misleading confidence. The app provides descriptive analysis, adding variation and standard error with independent replicates. It does not replace full statistical inference.

It is only the most favorable measured combination, based on its mean. Behavior between levels or outside the tested region may differ. Confirm the choice with new runs.

Check measurement quality, level spacing, variation, selected factors and interactions. An inconclusive result does not prove a factor is unimportant under all conditions.

Yes. Download the plan as CSV. Paste responses from one spreadsheet column in the displayed run order. Use a JSON project file to transfer the entire editable project.

Yes. Planning, analysis, examples, guide, FAQ and exports are free. No registration, subscription or locked paid features. Personal consulting is a separate optional service.

Calculations and project import run in your browser. The DOE app does not send entered experiment data to a server. The privacy policy covers general website services.

The app does not automatically save permanently. Download a JSON project file before closing or reloading. Use “Load project” to continue later. CSV and print views document results; JSON preserves the editable project.

Yes, including workplace projects. Check method suitability, data quality and technical requirements. The app does not approve a process or provide certification evidence.

Fractional factorial, blocked, split-plot, mixture and response surface designs. No automated power planning, p-values, confidence intervals or model validation. The guide describes the free feature set.