Free Design of Experiments Template (Worksheet + Run Tables)
Most design of experiments templates you'll find online are either a blank grid with no guidance, or a spreadsheet full of formulas you're afraid to touch. This one has two parts: a planning worksheet to fill in before you touch a pipette, and two run tables you can copy straight into a spreadsheet. Both are free, no sign-up.
Part 1: The planning worksheet
Most failed experiments fail at the planning stage, not the bench. Fill this in first — it forces the decisions that are much cheaper to make now than after you've run 16 wells.
| Field | What to write | Example |
|---|---|---|
| Objective | One sentence: what decision will this experiment inform? | Find which media components most affect CHO titer |
| Response(s) | What you'll measure, its units, and whether higher or lower is better | Titer (g/L) — maximize |
| Factors | Each variable you'll deliberately change | Glucose, glutamine, temperature |
| Levels | A low and a high setting for each factor | Glucose: 4 g/L / 8 g/L |
| Hard-to-change factors | Anything you can't realistically reset between individual runs | Incubator temperature |
| Run budget | How many runs you can actually afford (time, reagents, plate space) | 8–16 wells |
| Known noise | Typical variation between repeat runs of the same condition | ±0.2 g/L between replicates |
| Nuisance variables | Things that vary but that you aren't studying | Analyst, day, reagent lot |
A few things worth getting right here:
- Pick levels far enough apart to see an effect, but within a range that's safe and realistic for your system. Levels that are too close together make a real effect look like noise.
- Hard-to-change factors matter more than they look. The run tables below assume every run can be done in a fully random order. If one of your factors can't be reset between runs, that assumption breaks — see Hard-to-Change Factors and Split-Plot Designs before you go further.
- Nuisance variables like analyst or day can be handled by blocking rather than ignored — see How to Block an Experiment.
Part 2: Ready-to-use run tables
In both tables, −1 means the factor's low level and +1 means its high level. You'll replace these with your real values from the worksheet.
Std order is just the tidy, textbook way of listing the runs. Run order is the order you actually perform them in — it's been randomized for you as an example. Generate a fresh random order for each experiment you run (step 2 below shows how), so repeated experiments don't all share the same sequence.
3 factors: full factorial (8 runs)
| Std order | A | B | C | Run order | Response |
|---|---|---|---|---|---|
| 1 | −1 | −1 | −1 | 3 | |
| 2 | +1 | −1 | −1 | 7 | |
| 3 | −1 | +1 | −1 | 5 | |
| 4 | +1 | +1 | −1 | 1 | |
| 5 | −1 | −1 | +1 | 4 | |
| 6 | +1 | −1 | +1 | 8 | |
| 7 | −1 | +1 | +1 | 6 | |
| 8 | +1 | +1 | +1 | 2 |
Every combination of low and high for three factors. This lets you estimate every main effect and every interaction with no aliasing at all. More on why in Full Factorial Designs, Explained.
4 factors: half-fraction factorial (8 runs, Resolution IV)
| Std order | A | B | C | D | Run order | Response |
|---|---|---|---|---|---|---|
| 1 | −1 | −1 | −1 | −1 | 8 | |
| 2 | +1 | −1 | −1 | +1 | 4 | |
| 3 | −1 | +1 | −1 | +1 | 1 | |
| 4 | +1 | +1 | −1 | −1 | 2 | |
| 5 | −1 | −1 | +1 | +1 | 3 | |
| 6 | +1 | −1 | +1 | −1 | 7 | |
| 7 | −1 | +1 | +1 | −1 | 5 | |
| 8 | +1 | +1 | +1 | +1 | 6 |
Four factors in the same 8 runs, instead of the 16 a full factorial would need. The trade-off: column D is set to the product of A × B × C, which means main effects are estimated cleanly, but two-factor interactions come out paired up (AB with CD, AC with BD, AD with BC) and can't be told apart. That's usually a good deal for a first screening experiment. The details are in Fractional Factorial Designs and Aliasing, Explained.
Optional: add center points. If all your factors are continuous (concentrations, temperatures — not categories like "supplier A vs. B"), add 2–4 extra runs with every factor set halfway between its low and high level. They give you a check for curvature and an estimate of run-to-run noise for almost no extra cost.
How to use the tables
- Replace −1 and +1 with your real levels. For example, if glucose is factor A, every −1 in column A becomes 4 g/L and every +1 becomes 8 g/L.
- Make a fresh random run order. The tables above include an example, but generate your own for each experiment: in a spreadsheet, add a column of
=RAND(), sort the rows by it, and number them 1–8 in the Run order column. Then do the runs in that order, not top to bottom. Running in standard order lets anything that drifts over time — instrument warm-up, reagent aging, a plate edge effect — get mixed up with whichever factor happens to change at the same pace. - Run and record. Work through the runs in your randomized order and write each result in the Response column.
- Replicate if you can. Even a couple of repeated runs gives you a real estimate of noise, which is what lets you tell a genuine effect from chance.
Where a template stops being enough
These two tables cover a lot of real experiments, but it's worth being honest about where a spreadsheet template runs out:
- More than four or five factors. Choosing the right fraction, and working out what's aliased with what, gets difficult fast. How Many Runs Do You Actually Need? and Which DOE Design Should I Use? cover how to think about it.
- Hard-to-change factors. The tables above assume complete randomization. If you can't actually randomize one factor, analyzing the data as if you did can make effects look real when they aren't.
- Finding an optimum, not just what matters. Two levels per factor can only tell you which direction to go. Pinpointing a peak takes a response surface design — see Screening Design vs. Response Surface.
- The analysis. Calculating effects, running an ANOVA, and deciding which effects are real rather than noise is where most template-based experiments go wrong. Reading Your ANOVA Table and Pareto Chart walks through it.
If you'd rather skip the spreadsheet, BioDOE generates these same designs (and larger ones) from a plain-language goal, randomizes the run order for you, handles hard-to-change factors properly, and runs the analysis when your results are in. The free tier includes 3 designs and 3 analyses a month. Try BioDOE free.