Research Plan Development

How long does a research project actually take?

Any single answer is guesswork without knowing the audience incidence, the number of languages, the approvals required and how many review rounds the organisation runs. What is reliable is the shape: in most real projects the visible research activities (fieldwork, analysis) occupy substantially less than half the elapsed time. Plan the invisible steps first and the rest of the timeline tells the truth.

What steps do people forget when planning research?

Recruitment and screening lead time driven by the hardest quota cell; translation, back-translation, reconciliation and a cognitive check in each language; every round of internal and client review, counted as rounds with real turnarounds; ethics, legal and data protection approval; instrument programming and full logic testing; the pilot and the decision after it; open-end coding; sign-off on quotes that name a person or organisation; and calendar blackouts.

How do I work out the critical path for a research project?

Write dependencies as relationships rather than an ordering, then trace the longest chain from start to the deliverable date. Every task on it has zero float. Then check the two things a simple chain hides: resource contention, where two parallel tasks need the same person and are therefore sequential in practice, and external dependencies on people outside the project, who have no float and no accountability attached.

A client wants delivery three weeks earlier. What do I say?

Not yes or no, but options with costs. Some things are usually incompressible: recruitment lead time for a low-incidence audience, institutional approval, translation quality, the minimum fieldwork window needed to reach people unavailable on weekdays. Some are compressible at a stated cost: fewer objectives, less analysis depth, fewer review rounds, no pilot, a shorter fieldwork window. Write each cost in what the study will be able to say, not in days.

Why does the analysis stage always get squeezed?

Because it sits last, and it is the only stage with no external party waiting on it. Every upstream slip lands there. The consequence is invisible in the deliverable: the study is on time and reasons badly. Protect the analysis window explicitly and test the plan against a one-week fieldwork slip.

What is the difference between a research plan and a research proposal?

The plan is internal: activities, dependencies, owners, risks and the honest timeline the team will work to. The proposal is external: the approach explained to a buyer, the scope and its exclusions, the assumptions register, the deliverables and the commitment. Build them in that order. A proposal whose timeline was never planned internally is a promise nobody has tested.

What should a research risk register contain?

Six fields per risk: the risk, its likelihood, its impact on the decision rather than on the project, an early warning indicator someone will actually observe, the mitigation, and the trigger point at which the mitigation is enacted. Without a trigger it is a document rather than a control. Include the risk nobody writes down: that the finding will be politically unwelcome.

Can AI build a research timeline?

It can structure the plan, surface the steps that get forgotten, trace the dependencies and hold the whole chart in view. What it must not do is state durations, lead times or availability as known when nobody supplied them, or make a plan fit a deadline by absorbing the compression quietly. Every duration should be labelled historical, supplied or assumed, and every trade-off should stop for a person.