Advice about literature reviews tends to stay abstract. You are told to synthesise rather than summarise, to be critical, to build an argument. All true, and none of it tells you what to actually type.
So this is a literature review example built the other way round. One set of sources, written first the way most students write it, then rewritten the way it should be, with the reasoning between the two versions.
The topic is invented and so are the citations, deliberately, so that nothing here can be lifted into a real submission. The technique transfers to any field.
The material we are working with
Assume you are writing about whether flexible working improves employee productivity, and you have five sources:
- Ashworth (2018) — survey of 400 office staff, self-reported productivity, found a strong positive effect
- Baptiste (2020) — analysis of output data from a call centre, found no significant effect
- Cordeiro and Vale (2021) — interviews with 30 managers, reported mixed views
- Duarte (2022) — survey of 1,200 knowledge workers, self-reported, strong positive effect
- Ekwueme (2023) — observational study in a logistics firm, found a small negative effect
Version one: how it usually gets written
Ashworth (2018) surveyed 400 office staff about flexible working. The study found that employees reported higher productivity when allowed to work flexibly. Ashworth concluded that flexible arrangements benefit organisations.
Baptiste (2020) examined output data in a call centre setting. This study found no significant relationship between flexible working and productivity. Baptiste suggested further research was needed.
Cordeiro and Vale (2021) interviewed 30 managers about their experiences of flexible working. The managers expressed mixed views. Some felt it improved performance while others were more sceptical.
Duarte (2022) surveyed 1,200 knowledge workers and found a strong positive effect on productivity. Ekwueme (2023) conducted an observational study in a logistics firm and found a small negative effect.
Nothing in that is wrong. Every source is described accurately and cited correctly. It would take a marker about ten seconds to identify it as descriptive, and it would sit around a 2:2.
Why it does not work
Every paragraph opens with a surname, which means the structure is dictated by your reading order rather than by any idea.
The five studies never meet. They disagree sharply with one another and the writing does not acknowledge it, so the reader has to do the comparison themselves. Nothing is evaluated, so a 400-person self-report survey carries the same apparent weight as observed output data. And it ends nowhere in particular.
Version two: the same sources, rebuilt
Research on flexible working and productivity reaches strikingly different conclusions, and the pattern of disagreement follows the method used rather than the sector studied.
Studies relying on employee self-report consistently find substantial gains. Ashworth (2018) and Duarte (2022), surveying 400 office staff and 1,200 knowledge workers respectively, both report strong positive effects. Their scale is a genuine strength, and the consistency across two independent samples is not easily dismissed.
Studies using observed output tell a different story. Baptiste (2020), analysing call centre performance data, found no significant effect, and Ekwueme (2023) reported a small negative one in a logistics firm. Neither study depends on how productive employees believe themselves to be.
The divergence is therefore difficult to attribute to context alone, since it holds across office, call centre and logistics settings. A more plausible reading is that self-report captures satisfaction with flexible working and translates it into perceived productivity, an interpretation supported by Cordeiro and Vale (2021), whose managers valued flexibility for retention while remaining uncertain about its effect on output.
What has not been established is whether the gap closes when the work itself is complex. The observational studies examined roles with easily counted outputs, while the self-report studies drew on knowledge work, where output resists simple measurement. No study has yet applied objective measures to complex work, which is the gap this research addresses.
What changed
Roughly the same length. Same five sources. A different chapter.
The opening sentence makes a claim about the field rather than a statement about a study, so the reader knows what the section is arguing before any source appears. Sources are then grouped by method, because that is where the disagreement actually lies — and noticing that is the analytical move the first version never makes.
Evidence is weighed rather than reported. Saying that observational measures do not depend on employee perception is a judgement, and judgement is what separates a first from a 2:1. The contradiction is explained rather than left sitting there. And the paragraph ends by identifying what remains unknown, which hands the reader your research question.
The pattern underneath it
Every strong body paragraph in a review follows roughly the same sequence, and it is worth writing it out until it becomes automatic:
- Claim. Something true about the state of the field.
- Evidence for one position, with several sources grouped together.
- Evidence for the competing position, likewise grouped.
- Your explanation of why they diverge.
- What that leaves unresolved, which carries the reader forward.
If a paragraph of yours does not contain step four, it is a summary with citations attached.
A test you can run on your own draft
Take any paragraph of your review and delete every citation from it.
Read what is left. If there is still an argument standing, the paragraph is doing analytical work and the sources are supporting it. If the paragraph collapses into nothing, the citations were carrying it, and what you have written is a summary.
Apply this to five paragraphs and you will know exactly where your chapter stands, which is usually more useful than reading it through again from the top.
Where to go next
If your paragraphs are sound but the chapter as a whole feels shapeless, the problem is one level up, and our guide on how to structure a literature review deals with arranging the sections themselves. If you are unsure whether yours is the right length, we set out word counts by degree level in how long a literature review should be. And it is worth checking your draft against the mistakes that most often cost marks.
If you would rather have a hand
Rewriting a descriptive review into an analytical one is mechanical once you can see the joins, and genuinely difficult before that.
Our dissertation writing help covers literature reviews at any stage, whether that is planning the sections from your reading or reworking a draft that has come back marked as descriptive. At doctoral level the same applies through our thesis writing service.