Development · Development
“We have too many factors and too little material, and nobody on the team is sure which design fits.”
Design of experiments (DoE): from design to analysis
A well-chosen design saves experiments; a poorly chosen one produces data you cannot learn anything from. CovaSyn picks the right design, checks up front whether the number of runs is enough for your question, and analyses the results statistically. You receive design and analysis as a report that your expert reviews and approves.
When you need this
- A method or process has many possible factors, and the budget only covers a few runs.
- A screening is done, and now the optimum should be found with a response surface (RSM).
- Data from an existing DoE is available, but the model fits poorly or the residuals look odd.
- Before running anything, you want to know whether the planned runs can detect a relevant effect at all.
What it costs when the case stays open
- Experiments without a fitting design use up material you will miss later.
- In the end you have data from which no sound conclusion can be drawn.
- Development loses weeks because experiments have to be repeated.
What you get
Design with justification
A screening or RSM design that fits your factors and material, such as fractional factorial, definitive screening, D-optimal, central composite or Box-Behnken, with its alias structure.
Power analysis before the experiment
How large an effect has to be for the design to detect it at your variability, and whether extra runs are needed.
Statistical analysis
Significant factors and interactions, model selection, residual diagnostics and response surfaces with the optimum.
Next step
Whether a screening can be augmented to an RSM design, which runs are missing and where the design space lies.
From an open case to a approved report
1. Discuss the case
30 minutes on Microsoft Teams about the compound, the question and the data. Afterwards you know whether we take on the case and which data we need.
2. Hand over the data
Raw data and context, such as spectra, chromatograms, stability tables and the specification. The report typically arrives within 5 working days once we have your complete data.
3. Review and release
The report states its data basis, calculation path and limits. Review and release happen on your side under your quality system.
What we evaluate for this
Every number is computed and reproducible, predictions state their uncertainty. Every evaluation is documented with its input data and reasoning so your expert can review and sign it off.
Experimental design: draft to analysis
Design suggestion and creation, augmentation of existing designs, design comparison, power analysis, alias structure, analysis, response surfaces, residual diagnostics, optimisation and design space.
Next experiment from process data
Recommendation for the next design, models from existing process data and multi-objective optimisation when several responses matter at once.
Who it is for
For method development, process development and CMC teams at contract labs, CDMOs, generics manufacturers and specialty chemicals companies who need sound conclusions from few experiments.
What the computation does not replace
- Statistics do not replace process knowledge. You define which factors and ranges make sense; we check and calculate.
- A model only holds within the studied region. A predicted optimum should be checked with confirmation runs in the lab.
Read on
Screening Designs Compared: Fractional, PB and DSD
Screening design comparison: fractional factorial vs Plackett-Burman vs definitive screening, with run counts, resolution and power from real calculations.
Read articleResponse Surface Methodology, End to End
A response surface methodology tutorial with a worked example: 18-run CCD, quadratic fit R2 0.9988, optimum with CIs, design space and residual.
Read articleDoE Power Analysis Before You Run the Experiment
DoE power analysis in plain terms: why a 13-run screening design catches a 1 sigma effect 82% of the time but a 0.5 sigma effect only 31%, and how factor.
Read articleAugment a Screening DoE into a Response Surface
How to augment a DoE design to an RSM: turn a 13-run screening design into a 21-run CCD with 8 new runs instead of starting over with 17.
Read articleDoE Residual Diagnostics: Shapiro-Wilk and Cook's D
DoE residual diagnostics explained: what Shapiro-Wilk, Cook's distance and leverage actually tell you about an RSM fit, on a real central composite design.
Read article
Common questions
Which experimental design fits my question?
For many factors and the question of which ones matter at all, screening designs such as fractional factorials or definitive screening fit. For the optimum of a few factors, RSM designs such as central composite or Box-Behnken fit. CovaSyn compares the options for your factors and material and justifies the choice.
How many runs do I need for a DoE?
It depends on how large an effect you want to detect and how much your measurement varies. A power analysis before the experiment shows whether the design is enough or whether replicates or extra runs are needed. CovaSyn calculates this for your specific design.
Who can analyse our DoE data?
CovaSyn analyses existing designs: effects and interactions, model selection, residual diagnostics and response surfaces. If the model fits poorly, we show whether outliers, missing curvature or a transformation is the cause.
What to do once the screening is finished?
Often the screening can be extended rather than restarted: a full or high-resolution fractional factorial becomes a central composite design with star and centre points; heavily confounded screens need additional runs first. CovaSyn checks which runs are missing and then calculates the optimum and design space.
Who can create and analyse an experimental design for us?
We do. In the initial call we go through goal, factors and constraints and agree on data and scope. The design or analysis typically arrives within 5 working days once we have your complete data.
Related problems we solve
Solve a solubility problem before you run out of material
Our API won't dissolve properly, and we have been trying solvents for weeks.
See solutionCrystallization process design: cooling or antisolvent
Our crystallization yield is too low, and we don't know whether to cool or use an antisolvent.
See solutionSynthesis route planning: retrosynthesis and route scouting
We have the target structure, but no route we want to rely on yet.
See solution
Free PDF
Results you can rely on
Our beginner's guide for chemistry, pharma and biotech: how to tell an estimate from a calculation, seven questions to ask of every result and six questions to ask every provider. 8 pages.
Next step
Tell us about your experimental design.
30 minutes on Microsoft Teams with the CovaSyn team. If it is not a fit, we will say so.

