When I tell people the first paper of my doctorate is a scoping review, the usual response is a nod and a change of subject. It sounds like the boring bit. It sounds like reading.
It is not reading. It is closer to a stocktake conducted under audit conditions, and it has taught me more about my own thinking than anything else I have done this year.
What it is for
A scoping review answers a different question from a systematic review. A systematic review asks what does the evidence say about this specific effect. A scoping review asks what is here at all: what has been studied, by whom, in what disciplines, using what methods, and where the ground has not been broken.
You do one when the field is scattered rather than settled. Mine is scattered across management and information systems, two disciplines that publish in different journals, use different words for the same thing, and largely do not read each other. That is the condition a scoping review exists for.
You write the rules before you see the data
This is the part that surprises practitioners most, and it is the part I have come to respect.
Before you look at a single paper, you write a protocol: the exact question, the databases you will search, the search strings, the criteria for including and excluding a study, and what you will extract from each one. Then you register it publicly, with a date stamp, before the work starts.
The reason is straightforward and slightly uncomfortable. If you decide what counts as relevant after you have seen what is out there, you will decide in favour of the answer you were hoping for. Every one of us will. Registering the protocol removes the option.
In practice this means that the most consequential decisions in the whole project are made at the point of maximum ignorance, which feels precisely as wrong as it sounds and is nonetheless correct.
The funnel
The shape of the work is a funnel with a brutal taper.
You start with a database search producing thousands of records. You screen every one of them on title and abstract against your criteria, and most of them go. What survives, you retrieve in full text, which is its own small industry of library requests and emails to authors. Then you read each full text against a decision tree and exclude again, recording a reason code for every exclusion so that someone can audit your judgement later.
What comes out the far end is a set of papers small enough to read properly. Thousands in, fewer than a hundred out, and the record of how each one was handled is itself a deliverable.
Then the actual work starts, because a list of papers is not a review. You extract structured data from each one, several dozen fields, and only then can you say anything about the shape of the field.
Three things I did not expect
The excludes are more interesting than the includes. A paper with the perfect title and no relevant content tells you something real about how a field talks to itself. Two of mine looked exactly on target and failed in opposite directions: one had all the right people and almost no AI in it, the other had all the AI and not one senior leader.
A false exclude is the expensive error. Wrongly including a paper costs you a few hours; you notice at full text and drop it. Wrongly excluding one is invisible and permanent, and it silently narrows what you can claim. So the screening is deliberately tuned to over-include, and the humans gate the borderline calls rather than the machine.
The absence of something is not a finding. This one has cost me most. If a search returns nothing, that is a statement about your search, not about the world, until you have proved otherwise. Most of my worst moments this year have been a confident nil that turned out to be a failed query.
Why it takes a year
Because none of the above can be hurried without destroying the thing that makes it worth doing. The value of a scoping review is entirely in its auditability: someone else should be able to follow your protocol and arrive at a comparable set of papers. The moment you take a shortcut you cannot document, you have written an opinion piece with a large bibliography.
I have another few months on mine. I will not say here what it is finding, because that belongs in the paper and the paper is not written. But I can say that the method has done something I did not expect it to do, which is to change what I think the question is.
That, I am told, is normal. It is also the entire point.