Optimizing a liquid class is a routine, not a talent. Once you have run it a few times it becomes muscle memory, and the point of writing it down as steps is to get you to that muscle memory faster, without the afternoons lost to changing five things at once and not knowing which one helped. What follows is the procedure practitioners actually use: not a solver, not a design of experiments, just a disciplined loop of start close, observe, adjust, and measure. It works on any instrument because it is about the liquid, not the brand.
Step 1: know the liquid before you open the editor
The first step happens away from the screen. Learn how your liquid moves before you ask an instrument to move it. Is it viscous, volatile, cohesive, prone to foaming? Pipette it by hand to get a feel for how it clings, drips, or beads, and pull its density and any hazards from the safety data sheet. Do all of this at your normal lab temperature, because viscosity, density, and vapor pressure each shift by a few degrees, and note that temperature down so a discrepancy weeks later has an explanation instead of a shrug. Everything downstream is easier when you already know what you are dealing with.
Step 2: start from the closest validated class
Do not build from nothing. Almost every liquid resembles one you have handled before, so find its family, aqueous, volatile organic, involatile organic, viscous, or blood product, and clone the validated class for the nearest member. A new master mix starts from a water or PBS class. A new alcohol starts from an ethanol class. The starting class already encodes the family behavior, which means your adjustments will be small and targeted rather than sweeping guesses. Starting close does not skip validation; it shortens the tuning that leads to it.
Step 3: run a real transfer and watch it
Set up one transfer that mimics a single step of your actual method, at a volume you actually run, into the labware you actually use. Then watch it happen. A careful visual inspection catches most problems before any balance does, and it is free. Look for the tell-tale signs.
- Droplets hanging on the tip after aspiration or after dispense.
- Bubbles or foam forming as the liquid leaves the tip.
- An aspiration or dispense height that sits too high or too low in the well.
- Whether the channels are following the liquid level correctly, and whether following should be off at all for this volume.
If it already looks clean, good, you can move to measuring sooner. If it does not, you have your first thing to fix.
Step 4: change one parameter at a time
This is the rule that separates a converging optimization from a random walk. When something looks wrong, resist the urge to adjust several settings together. Change one, re-run, and see what moved. It feels slower, and it is faster overall, because each iteration teaches you what a single knob does for this specific liquid. Let the liquid's properties tell you which knob to reach for first.
- Viscous liquid: lower the flow rate, sometimes to ten percent of the water default, lengthen the settling time so the column can equalize, and prefer a surface dispense over a jet.
- Sticky or high-adhesion liquid: increase the blowout volume so the tip clears fully at the end of the dispense.
- Volatile liquid: increase the blowout to give vapor room and force, and pre-wet the tip so the first transfer is not an outlier against the rest.
- Low surface tension or low cohesion: increase the air transport volume for a bigger trailing buffer against drips, and shorten the settling time so the liquid does not have time to form a drop.
- High surface tension that clings to the source: raise the swap speed so the tip breaks the connection cleanly as it leaves the liquid.
- Low volumes near capillary scale: move to a smaller tip, keep the tip close to the surface, turn following off, and tune the blowout to prevent capillary over-aspiration.
Reach for liquid-level detection early when transfers look inconsistent from well to well; letting the tip sense and follow the surface removes a whole class of height errors at once.
Step 5: measure precision, not just appearance
A transfer that looks clean can still be delivering the wrong volume with a wide spread, and your eye cannot see three percent. Once the visual problems are gone, measure. Dispense a set of replicates across your working range and weigh them on a calibrated balance, or read a dye with a plate reader if a balance is not available; food dye and a reader work comparably well for tuning even where they lack the regulatory standing of a formal kit. Look at the spread first. Precision is what the mechanical parameters buy you, so if the spread is still wider than your method allows, go back to step four with the specific parameter the symptom points at. Do not reach for the correction curve yet.
Step 6: set the correction curve for trueness
With a tight spread in hand, you can finally make the average land on target. The correction curve maps the volume you ask for to the volume the instrument actually commands, cancelling steady over- or under-delivery. One point is not enough, because the offset is rarely the same fraction at 50 microliters as at 300, so add a point wherever accuracy matters across your range. For a stepped workflow like multi-dispensing, correct against the actual volume in the tip at each step, which keeps dropping, rather than the nominal aliquot size. Re-measure after you set the curve to confirm the average moved where you intended.
Step 7: verify, record, and stop
Run one more set of replicates as a clean confirmation, ideally on a fresh setup so you are not just re-measuring the same lucky run. If the precision and trueness sit inside your method's tolerance across the volumes you run, the class is optimized. Write down what you changed and why, the temperature, the tip, the labware, and the measured result, so the class carries its own evidence and the next person adapts from a known-good starting point instead of a blank form. Then stop. Tuning past the tolerance your assay needs is effort you will not get back.
Start close, change one knob at a time, measure precision before you touch trueness, and stop when the numbers meet the method. The discipline is the whole method.