Is Automation Required to do Experimental Design?

Does DoE require expensive automation to implement? Absolutely not! Expensive automated equipment is not a prerequisite to using experimental design, some tactics for design success are discussed below.

Q: What do I need to get started with DoE?

A: The lab you are currently running your experiments in and a computer with DoE software - it doesn‘t have to be all bells and whistles, there are cheap and even free ones out there, additionally most will do 14-30 day trials.

Q: What can automation help with?

A: Many things! Including dispensing, monitoring and continuous adjustments, sampling to get data from various time points through the reaction, and allowing the experiments to continue without your physical presence meaning you can focus on solving problems rather than implementing lab work. From a DoE point of view, a key benefit of automation is delivering precise control of the changing reaction parameters and recording of parameter variations.

However, a designed experiment can be run simply and in fact automation may even hinder progress by fitting the experiments to the what the automation can do, rather than investigating what the chemistry requires (for example, automated liquid dispensing may limit which solvents are used throughout a study rather than investigating the effect of solvent by changing this variable).

A reason why people may assume automation is a necessity is the perceived complexity of DoE. A typical screening design investigating a chemical reaction may assess 6-10 (or more) different factors which may include: reagent amounts, solvent and solvent volume, temperature, addition rates among others (see our factor selection article for more on this), and these factors are changing in different combinations for between 19-35 experiments. Keeping track of the quantities added and ensuring control of these settings is essential to ensuring the repeatability of the process and therefore understanding the error, and can be prone to fault when done manually. How can we minimize the risk of operator error? Being prepared, organized and meticulous in setting up each run. This can be aided by spreadsheets detailing the reagents, quantities, volumes and any other changing variable in each run. If you’re using plates, the tables can be organized by temperature zone. However you need to do it, whether it's with a electronic lab notebook or an old school print off of the table blue tacked to the fume cupboard, tick off the added reagents as you go helping you to keep track!

If you’re new to design and feel like a lack of equipment is a barrier to you I suggest you start small. Pick a problematic reaction you currently have. Decide on your objective. Select which factors to assess (maybe start with 4 and ensure you hold all other potential factors at a chosen setting) and do a simple factorial design of 11 experiments. Use free software (use a trial) to analyse the data and build a model, assess what the models are telling you, carry out the validation experiments and implement the findings into a hopefully improved process. After carrying out your initial design successfully you will have a blueprint for DoE success and can go on to more involved designs and figure out where automation may help going forwards.

While there’s a range of awesome software you can use, and automation may benefit the practical aspects of design, what you really need to understand before you get started with the process is 1) what question do I want to answer and 2) how am I going to assess that? If you want more tips, help with implementing design or DoE training please get in touch!