How to build a systematic review project plan: a stage-by-stage timeline, dependencies, milestones, buffer, and the scheduling mistakes that derail reviews.
Knowing that a systematic review takes months is not the same as knowing how to schedule one. The useful question is not "how long," it is "what has to happen in what order, what depends on what, and where will this slip." A review that fails on time almost never fails because someone was slow. It fails because the plan had no buffer, or because a dependency was invisible until it bit.
This is a project management guide, not a methods guide. It assumes you know the stages and want to lay them on a calendar.
Build the plan backwards from your deadline, in stages with explicit dependencies. Allocate roughly: a month for the question and protocol, a month for the search, two to three months for screening, two to three months for extraction and appraisal, one to two months for synthesis, and two to three months for writing and revision. Then add a buffer of at least twenty percent, because you will need it.
The stages overlap in practice, but the dependencies do not disappear. You cannot screen before you search, and you cannot pool before you extract.
What makes a review schedule tricky is that some stages can overlap and some absolutely cannot. Get this wrong and you build a plan that was never achievable.
Hard dependencies (cannot overlap):
Soft dependencies (can overlap):
The soft dependencies are where you claw back time. The hard ones are where plans break.
For a moderate review with a small team, working steadily rather than full time. Adjust the scale to your project.
| Stage | Duration | Notes |
|---|---|---|
| Question, scoping searches, feasibility | 2 to 4 weeks | Includes checking no recent review exists |
| Protocol writing and registration | 2 to 4 weeks | Registration approval adds waiting time |
| Search development and testing | 2 to 4 weeks | Involve a librarian; test before committing |
| Running the search, deduplication | 1 to 2 weeks | Scales with database count |
| Title and abstract screening | 4 to 12 weeks | The biggest variable; scales with record count |
| Full-text retrieval and screening | 3 to 6 weeks | Interlibrary loans add dead time |
| Data extraction | 4 to 10 weeks | Slower in duplicate |
| Risk-of-bias appraisal | 2 to 6 weeks | Can overlap extraction |
| Synthesis or meta-analysis | 3 to 8 weeks | Longer if pooling |
| Writing and internal revision | 6 to 12 weeks | Consistently underestimated |
| Buffer | 20% of total | Not optional |
Add those up and even an efficient project lands past six months. That is the honest floor.
Rather than tracking every task, track a small number of gates. If you hit these, you are on schedule; if you slip one, you know immediately.
Protocol registered. Nothing formal begins until this is done. If this slips, everything slips.
Search run and deduplicated. You now know your record count, which is the single number that determines your screening timeline. Re-plan here with real figures rather than estimates.
Title and abstract screening complete. The largest volume stage is behind you and you know your full-text count.
Included set finalized. You now know exactly what you are extracting from. This is the point where the meta-analysis becomes plannable.
Extraction complete. Synthesis can begin.
First full draft. Everything after this is revision, which has its own long tail.
The record count surprise. You planned for three thousand records and the search returns fourteen thousand. This is the most common and most damaging schedule shock, and it happens because the search was not tested against a realistic scope before the plan was fixed. Mitigate it by running a pilot search early, before you commit to a timeline, and by re-planning the moment you have your real count.
Full-text retrieval. Chasing papers your library does not hold, waiting on interlibrary loans, and hunting down old or obscure studies produces dead time you do not control. Start retrieval as soon as studies pass title/abstract screening rather than waiting for the stage to finish.
Waiting on authors. If you need missing data, you are on someone else's schedule. Contact authors as early as possible and plan for how you will proceed if they do not reply.
The writing tail. Teams treat writing as an afterthought and then spend three months on it. Draft the methods during the review and start the results as soon as the synthesis firms up.
Reviewer availability. Screening stalls when one of your two reviewers is unavailable for a month. Agree availability upfront and build slack around known absences.
Registration approval. Registers check submissions before publishing them, which takes time and may involve a round of clarification. Submit early rather than treating approval as instant.
Start from the deadline and work backwards, but sanity check forwards. Lay out the stages with their hard dependencies fixed and their soft dependencies overlapped. Assign an owner to each stage, because a stage without an owner is a stage that stalls.
Then apply a buffer. Twenty percent is a reasonable minimum and it is not padding; it is the acknowledgement that at least one of the risks above will materialize. A plan with no buffer is a plan that assumes nothing goes wrong, which no review has ever managed.
Finally, re-plan after the search. Your record count is the number the whole screening timeline hangs on, and until you have it, your schedule is a guess. Treat the pre-search plan as provisional and the post-search plan as real.
Sometimes the honest conclusion is that a full systematic review does not fit the time available. There are three legitimate responses.
Narrow the question. A tighter scope cuts the record count and shortens the longest stage. This is usually the best option, because it preserves rigor.
Add reviewers. More screeners relieve the main bottleneck, though coordination overhead limits the gains.
Run a rapid review instead. Streamline the method deliberately and disclose exactly what you compressed. An honest rapid review is respectable. A systematic review with hidden shortcuts is not.
The one response that is not legitimate is keeping the systematic review label while quietly cutting corners to meet the date. That is how the deadline turns into a methodological problem instead of a scheduling one.
How long should I plan for a systematic review? Six months to over a year for a full review, depending on record count, team size, and whether you run a meta-analysis. Always add a buffer.
Which stage takes the longest? Title and abstract screening, which scales directly with your record count. Writing and revision is the other stage people consistently underestimate.
What can overlap? Risk-of-bias appraisal can run alongside extraction, methods writing can happen during the review, and full-text retrieval can start before screening finishes. The protocol-before-search dependency cannot be overlapped.
How much buffer should I add? At least twenty percent. Search results larger than expected, slow full-text retrieval, and unresponsive authors are common enough that a bufferless plan will slip.
What if my deadline is too tight? Narrow the question, add reviewers, or run an explicit rapid review with disclosed shortcuts. Do not quietly cut corners on a review you still call systematic.
Plan a systematic review as a project with dependencies, not as a to-do list. Fix the hard dependencies, overlap the soft ones, assign an owner to each stage, and add a real buffer. Then re-plan the moment your search returns, because the record count is the number that determines everything downstream.
The schedule breaks in predictable places: an unexpectedly large search, slow full-text retrieval, and an underestimated writing tail. Plan for those three and you will be more accurate than most.
Trying to keep a multi-month review on track? Verflux keeps the search, screening, appraisal, and PRISMA counts in one place, so the mechanics do not become the delay.
All analytical features included in the free trial. No credit card, no installation, no R or Python.
Create free account