Briefing a CRO
Most disappointing outsourced work traces back to the brief, not the bench. The brief is where the question, the decision it feeds, and the constraints get written down, and where the parts that were assumed rather than stated do their damage.
Lead with the decision
Say what you will do differently depending on the result. "We will advance three candidates into a cell assay" or "we will decide whether to keep the program" changes how an experiment should be designed, how many replicates it needs, and how tight a difference has to be before it counts.
An experiment designed to produce data is a different experiment from one designed to settle a question, and only the second is worth paying for.
Name the acceptance criteria before the work starts
What result would make a candidate pass? What separation between two candidates counts as real? Agreeing this up front protects both sides, because a result that lands in an ambiguous zone otherwise becomes a negotiation rather than an answer.
If nobody can state a criterion, that is a useful thing to discover before running the study rather than after.
Send the material and the history
The molecule's history explains most of the surprises: expression host, purification route, buffer, concentration, freeze-thaw count, and what the material did in your hands. For an antigen, add construct boundaries, tag, labeling method and lot. A vendor who has that information can spot the problem in a day; without it, both sides will be guessing.
Say what has already failed
Prior attempts, including the ones that did not work, save more time than any other part of the brief. It costs nothing to write down that the antibody did not survive a previous freeze-thaw, or that a competitor's reagent gave high background, and it is often the whole answer.
Be specific about the deliverable
Raw data or summaries? Plate maps? A written interpretation or numbers only? In what format, and who owns it? A report that contains a conclusion but not the underlying data cannot be re-analyzed when the program changes direction, which it will.
Agree on how surprises get handled
Experiments go sideways. Decide in advance who gets called, how fast, and with what authority to change the plan. The worst outcome is a vendor who completes a study as written after it became obvious at day two that the study had a problem.
What a good vendor should do with the brief
Push back. A brief that asks for a measurement the assay cannot support, or a comparison that the material quality cannot justify, should come back with questions rather than a quote. That exchange is usually where the study design improves most, and a vendor who never asks anything is not reading carefully.