Most protocols are written once, by the person who already knows how to do the thing, on the day they are least likely to notice what they have left out. That is why so many of them work perfectly for their author and fall apart for everyone else.
The failure is almost never a missing step. It is a missing condition: the centrifuge that has to be pre-cooled, the reagent that is only stable for an hour after thawing, the incubation that is thirty minutes in one room and forty in another because the bench by the window runs warm.
Separate the protocol from the run
A protocol is what you intend to do. A run record is what actually happened. Keeping them in one document is the most common documentation mistake in a working lab, because the moment you annotate the protocol with what went wrong on Tuesday, it stops being a protocol and becomes a diary that the next person has to reverse-engineer.
Write the protocol as instructions in the imperative, with no dates, no initials and no results. Write the run record as observations, with all three. When the protocol changes, version it deliberately; when a run deviates, note the deviation on the run, not on the protocol.
The metadata that actually matters
Ask what a reader would need in order to get the same result, and write down only that. In practice it is a short list, and it is almost always the same list:
- Reagent identity, including supplier and catalogue number where a substitute would change the outcome
- Concentration and the units it is expressed in — 5% is not 5 mM, and a protocol that says only "5" has lost the information
- Volumes per reaction, and whether an overage is already included
- Temperature and time for every incubation, with the tolerance that matters ("37 °C ± 1", "5 min, not critical")
- Equipment where the model changes the result: rotor type, plate reader filter set, thermocycler ramp rate
- The order of addition, when it is load-bearing
Record deviations without drowning in notes
The instinct to write everything down produces run records nobody reads. The useful discipline is narrower: record what differed from the protocol, and record what you observed that the protocol did not predict. Everything else is already in the protocol, and repeating it adds nothing.
"Followed protocol" is a complete and honest run record for a run that followed the protocol. "Incubated 45 min instead of 30, phone call" is worth more than three paragraphs describing steps that went exactly as written.
Let the calculations live outside the prose
Protocols that embed arithmetic go stale the moment the scale changes. A protocol written for 10 reactions becomes wrong for 24, and someone will do the multiplication in their head at the bench and get it wrong at least once.
Write the recipe per reaction and let a tool do the scaling. That is what the Master Mix and Medium/Buffer calculators are for: the protocol keeps the concentrations, the tool produces the volumes for the number of reactions you are actually running today, and nobody multiplies anything on the back of a glove box.
Write it for the version of you that has forgotten
The reader you are writing for is not a stranger. It is you, in fourteen months, after two other projects, trying to work out why the numbers from last spring do not match the numbers from this week. That reader has no memory of the room, the reagent lot, or which step you always did slightly differently.
A protocol that survives that reader is a protocol that survives review.