Most literature reviews do not fail because the researcher read too little. They fail because the researcher read without a stopping rule, and eight months later has four hundred PDFs and no chapter.
A systematic review is not a heavier version of a normal review. It is a lighter one, because every decision is made once, in advance, and then simply followed.
Start with a question narrow enough to answer
"What does the literature say about microfinance?" is not a review question. It is a subject heading. A review question specifies a population, an intervention or phenomenon, and an outcome — and it implies which studies are in and which are out.
Compare: "How has group-lending microfinance affected household savings behaviour in South Asia since 2010?" That question tells you the databases to search, the date cut-off, the geography filter, and — crucially — when to stop.
If your question does not tell you when to stop, it is not yet a question.
Write the protocol before the first search
A protocol is one page. It records:
- The databases you will search, and why those
- Your exact search strings, per database
- Inclusion and exclusion criteria
- The date range and language limits
- How disagreements about inclusion get resolved
Writing this down costs an hour. Not writing it down costs you the ability to describe your method in the viva, and invites the question every examiner eventually asks: how do you know you didn’t just find the studies that agreed with you?
Screen in two passes, never one
The single biggest source of lost momentum is reading full texts too early. Screen in two passes.
Pass one — titles and abstracts only. Decide in under thirty seconds per record. You are answering one question: could this plausibly meet the criteria? Err toward inclusion; a false positive costs you five minutes later, a false negative costs you a source you never learn about.
Pass two — full text. Now apply the criteria strictly, and record the reason for every exclusion. Those reasons become your PRISMA flow diagram, and the diagram is what makes the review auditable.
Extract into a matrix, not into prose
Do not start writing paragraphs while you read. Build a synthesis matrix: one row per study, columns for method, sample, context, key finding, and limitation. Add a column for the theme you think it belongs to, and expect that column to change three times.
The matrix is the review. Once it is full, the themes are visible as clusters of rows, the gaps are visible as empty cells, and the writing becomes description of a structure that already exists.
Organize by theme, never by study
The most common structural failure in a graduate literature review is the annotated bibliography in disguise: Ahmed (2018) found… Khan (2019) found… Rahman (2021) found… Each paragraph is about a study. None of them are about the field.
Organize by claim instead. Each section advances one point about the state of knowledge, and studies appear as evidence for or against that point. Several studies belong in the same paragraph when they agree; the interesting paragraphs are the ones where they don’t.
End at the gap, deliberately
A review is an argument, and its conclusion is your research question. By the final section the reader should already be able to guess what you are about to study, because you have shown them precisely where the scholarship stops.
Name the gap concretely. Not "more research is needed" — which is true of everything — but "no study to date has examined this mechanism in a context where X holds, and the theoretical reason to expect a different result is Y."
That sentence is why your project exists. Everything before it was building toward being able to write it.
We do this work every week — for MPhil and PhD candidates, faculty, and whole departments.
Start a conversation
