Writing a Strong Introduction: The Funnel Approach
Oct 11, 2026
When I triage a new submission, I usually read the last paragraph of the introduction before I read anything else in the manuscript. That includes the abstract.
Most authors assume editors move through a paper in order. Some do. Many of the editors I've worked alongside over twenty years don't, and the reason is practical. That final paragraph is where a paper states what it's actually trying to find out. If I can't locate a clear question there, or if the question doesn't follow from anything that came before it, I already know what kind of review the paper is headed for. Sometimes I already know it won't get one.
So when people ask me how to write an introduction, I don't start with the opening line. I start with the destination. The funnel approach is the most useful way I know to get a reader from a problem they care about to a question they believe in. It's also one of the most misunderstood pieces of writing advice in academic publishing.
What the funnel actually is, and where most advice goes wrong
The textbook version goes like this. Start broad and establish that the topic matters. Narrow to what's known. Narrow again to what isn't known. End with your aim. It's a sound shape. Linguists have described versions of it for decades, most famously John Swales's "create a research space" model, and you can find a good deal of work on how scientific introductions are structured if you want the theory.
The trouble is the word "broad." Researchers read it as "start with the biggest true statement you can find." So I get introductions on a narrow pharmacokinetic question that open with the global burden of cardiovascular disease. I get qualitative studies of nurse handovers that begin with the history of hospital medicine.
My view, and some colleagues disagree with me: the top of the funnel should be much narrower than you think. It should be the broadest statement your specific reader needs to hear, not the broadest statement that's true. A reader of a specialist journal already knows heart failure is common and expensive. Telling them again uses up the attention you need for the part only you can provide.
A quick test for your opening sentence
Ask yourself whether your opening sentence could appear, unchanged, at the start of fifty other papers in the same journal. If it could, it's too wide. Try starting one level down.
Before: Cardiovascular disease is the leading cause of death worldwide and places an enormous burden on health systems.
After: Up to one in four patients hospitalized for heart failure is readmitted within a month, and in rural health systems many of those readmissions follow a missed or delayed first outpatient visit.
The second version is still broad enough to establish stakes, and it already points toward a specific mechanism. The reader can feel where you're heading. (If you use a figure like that "one in four" in a real manuscript, cite the specific source for your population. Don't borrow a number from someone else's introduction.)
The gap is the hinge, and "little is known" isn't a gap
The weakest moment in most introductions I read is the turn from what's known to what isn't. Authors write "little is known about X" or "no studies have examined X in population Y" and move straight on to the aims, as though absence alone justified the work.
It doesn't. Plenty of things haven't been studied because nobody needs to know them. A gap is persuasive only when the reader can see what's wrong, uncertain or stuck because it hasn't been filled.
A grant officer I worked with for years at a mid-sized foundation, Marcus, had a habit that changed how I edit. When he reviewed a proposal, he'd write one question in the margin next to the gap statement: "And so?" He wasn't being glib. He wanted the applicant to finish the thought. If we don't know this, what decision gets made badly? What model stays wrong? What patients get the wrong treatment? He told me once that the proposals he funded almost always answered his margin note before he'd had a chance to write it.
That's the standard I hold manuscripts to.
Before: To our knowledge, no studies have examined automated text-message reminders for post-discharge follow-up in rural heart failure patients.
After: Reminder interventions have been tested mainly in urban academic centers, where missed visits tend to reflect forgetting. In rural settings, missed visits more often reflect transport and scheduling barriers that a reminder alone may not address. Health systems are adopting these programs anyway, without evidence that they work where the barriers are different.
The second version tells me what kind of gap this is. It's a transferability gap with a plausible mechanism for failure and a real-world consequence. A reviewer can argue with it. That's good. A gap you can argue with is a gap with content.
Kinds of gap worth naming
It helps to know which kind of gap you're claiming, because each one sets up a different question:
Conflicting evidence: studies disagree, and you can explain why your design resolves the disagreement.
Transferability: findings exist, but in a population, setting or time where there's reason to expect a different result.
Methodological limitation: prior work is confounded, underpowered or measured the wrong outcome, and your design fixes a specific flaw.
Mechanism: we know that something happens but not why, and the "why" matters for intervention.
Practice outpacing evidence: something is already being done at scale without adequate testing.
If you can't place your gap in one of these, or in some category you can name just as clearly, that's worth sitting with before you submit.
A worked example, paragraph by paragraph
Here's a compact introduction built on the funnel, using the reminder study from above. It's a teaching example, so I've stripped the citations out. A real version would cite each factual claim.
Paragraph 1 (the problem, at the right width): Early readmission after heart failure hospitalization is common and costly, and timely outpatient follow-up is one of the few modifiable factors consistently associated with lower readmission risk. In rural health systems, first follow-up visits are missed more often than in urban ones.
Paragraph 2 (what's known, organized by argument, not by author): Automated reminder programs improve attendance in several outpatient settings. Most trials, however, enrolled patients from urban academic centers, and the effect sizes cluster in populations where forgetting is the main reason for non-attendance.
Paragraph 3 (the gap and why it matters): In rural populations, the barriers documented in qualitative and survey work are largely logistical: distance, transport and inflexible scheduling. A reminder may not address these. Several rural systems have nonetheless adopted text-message reminders, so the practical question is whether those programs change attendance or simply add cost.
Paragraph 4 (the aim, stated as a question you can answer): We therefore conducted a pragmatic cluster-randomized trial in twelve rural clinics to estimate the effect of automated reminders on attendance at the first post-discharge visit within fourteen days. A secondary aim was to test whether the effect differed by patients' travel distance to the clinic.
Notice what paragraph 2 doesn't do. It doesn't go "Smith et al. found... Jones et al. found... Lee et al. found..." That's an annotated bibliography, and it's the most common reason introductions run long. Organize the literature around the claims your argument needs, and let the citations support those claims.
Notice too that the secondary aim in paragraph 4 grows directly out of the mechanism in paragraph 3. When an aim appears that the introduction never set up, reviewers notice, and they often read it as a fishing expedition. Every aim should have a reason somewhere above it.
Mistakes I see in nearly every batch of submissions
The introduction is really a review article. Four or five pages, every relevant paper since 1990. Most journals want an introduction that runs from a few paragraphs to around a page and a half. Check the author guidelines and look at recent papers in the journal.
The question in the introduction doesn't match the question in the methods. The introduction promises effectiveness and the design measures feasibility. Or it builds toward causation and the study is cross-sectional. This mismatch is the single fastest route to a skeptical reviewer.
Results leak into the introduction. "We found that reminders were ineffective" doesn't belong there. Some fields end introductions with a one-line preview of the findings, and that's fine where it's the convention, but the default in most clinical and biomedical journals is to stop at the aim.
Hedged, soft aims. "We aimed to explore the role of..." tells me nothing about what you'll measure. Prefer "estimate," "compare," "test whether" or "describe," with the outcome named.
Novelty claims nobody checked. "This is the first study to..." is a sentence reviewers enjoy disproving. If you use it, make sure your search would hold up, and consider whether the claim adds anything your gap statement hasn't already earned.
Importance asserted rather than shown. Calling a topic "crucial" or "critical" doesn't make it so. A specific consequence does.
What the reporting guidelines ask of an introduction
People think of reporting guidelines as methods checklists, but most of them have explicit items for the introduction, and reviewers increasingly check them.
The ICMJE Recommendations ask authors to give context for the study and to state the specific purpose or research objective, and to cite only directly pertinent references. CONSORT for randomized trials asks for scientific background with an explanation of rationale, then specific objectives or hypotheses. The PRISMA 2020 statement separates rationale (in the context of existing knowledge) from an explicit statement of objectives for systematic reviews. STROBE asks observational studies for background, rationale and specific objectives including any prespecified hypotheses.
Notice the common pattern: rationale, then a specific objective. That's the funnel, codified. If you're not sure which guideline applies to your design, the EQUATOR Network library is the place to start. I'd download the relevant checklist and fill in the introduction items before drafting rather than after. It's a quick way to see whether your rationale and your objectives actually connect.
One more point that matters for trials and prespecified studies: the objectives in your introduction should match what you registered. If your trial registration lists a primary outcome, the aim in your final paragraph should name that same outcome. Editors do compare them.
Write it backward
The best introduction-writer I've worked with was a department chair, Elena, who ran a busy clinical department and still published steadily. Her method looked eccentric when I first saw it. She wrote the final paragraph first, the aims, in two or three sentences. Then she wrote the gap paragraph so that it made those aims feel inevitable. Then she wrote just enough background to make the gap matter. She wrote the opening sentence last.
When I asked her why, she said that if you start at the top, you write toward a destination you haven't chosen yet, and you end up with three pages of throat-clearing. If you start at the bottom, every paragraph has a job: to earn the one below it.
I've taught this to dozens of students since. It works especially well for people who freeze at the blank page, because the aims paragraph is usually the part of the paper they understand best. They wrote it in a protocol or a grant months ago.
A mini template for drafting backward
Aim: In this study, we [estimate / compare / test whether] [specific outcome] in [population and setting] using [design].
Gap: Prior work has [what it did], but [specific limitation or mismatch]. This matters because [concrete consequence for practice, theory or policy].
Known: We know that [two or three claims that the gap depends on], supported by the most direct evidence.
Problem: [The problem stated at the width your journal's readers need, not wider.]
Then reverse the order and smooth the transitions. Your first draft will be stiff. That's fine. Structure first, prose later.
A short checklist before you submit
Could my opening sentence open fifty other papers in this journal? If so, narrow it.
Is the literature organized by claim rather than by author?
Does my gap statement answer "and so?" with a specific consequence?
Does every aim, including secondary ones, have a reason stated above it?
Does the aim name the same outcome, population and design as my methods and my registration?
Have I covered the introduction items in the relevant reporting guideline?
Is it roughly the length of introductions recently published in this journal?
Have I deleted every "to our knowledge, this is the first" that I can't defend?
Where the work really happens
Most weak introductions aren't weak because of the writing. They're weak because the author hasn't yet decided precisely what gap the study fills. That decision happens during reading, long before drafting, when you notice that every trial came from the same kind of setting, or that two literatures never cite each other.
So keep track of those observations as you read. Note the claim each paper supports and the limitation it leaves open, not just the citation. Some researchers do this in a spreadsheet. I've also seen people use PACR, an AI-native research workspace that keeps literature search, reading notes and drafting in one place, so the limitation you flagged while exploring the literature is still sitting next to you when you write the gap paragraph. Whatever tool you use, the habit matters more than the software.
If you have a manuscript on your desk right now, try one thing this week. Cover everything except the final paragraph of your introduction and ask a colleague what the study is testing. If they can tell you in one sentence, you're most of the way there. If they can't, start there and work backward. And if you'd like a single place to keep your reading and drafting connected while you do it, you can set up a PACR workspace and build the next introduction from the bottom up.
