The Saunders Research Onion Explained: All Six Layers, With a Worked Example

The Saunders research onion is a diagram of six nested rings, published in Research Methods for Business Students, that arranges the decisions behind a piece of research from the most abstract on the outside to the most practical at the centre. Its appeal is procedural, not theoretical: it tells a student which question to settle first.

Most methodology chapters go wrong in the same way. They open with the questionnaire, because the questionnaire is the part the writer can picture, and then reverse-engineer a philosophy to justify it. The onion inverts that order. You cannot sensibly choose a sampling frame until you know whether you are testing a hypothesis or building one, and you cannot know that until you have settled what you think counts as evidence.

The Saunders research onion explained in one paragraph

Put plainly, the Saunders research onion explained without the diagram is a rule about sequence: settle what counts as evidence, then how theory will be used, then how many methods, then the vehicle, then the time frame, and only then the instrument. Each ring is an answer that removes options from the ring inside it. Students who can state the Saunders research onion explained in that order rarely write a methodology chapter that contradicts itself.

The six research onion layers, from the outside in

Philosophy

The outer ring asks what you believe knowledge is and how it can be obtained. Saunders lists positivism, critical realism, interpretivism, postmodernism and pragmatism. In business research the live contest is usually between the first and the third.

A positivist treats organisational facts as existing independently of the observer, measurable with instruments that other researchers could reuse and get the same numbers. An interpretivist holds that the meaning managers attach to events is the object of study, so a standardised instrument would flatten precisely what matters. Pragmatism, the position a great many dissertations quietly occupy, says the research question decides: use whatever produces a defensible answer.

The layer is not decoration. If you have declared yourself a positivist and your findings chapter is built from twelve unstructured interviews read for themes, the mismatch is visible to any examiner.

Approach to theory development

The second ring is deduction, induction or abduction. Deduction starts from existing theory, derives a hypothesis and tests it against data. Induction starts from the data and works towards a theoretical statement. Abduction moves between the two: a surprising observation prompts a plausible explanation, which is then tested and refined.

Undergraduate work leans deductive because a testable hypothesis is easier to mark. Genuinely inductive work is harder, because the researcher has to resist the temptation to name the theory before the data has earned it.

Methodological choice

Here the onion asks how many methods and of which kind. A mono-method study uses one, quantitative or qualitative. A multi-method study uses several of the same kind: two qualitative techniques, say, without mixing in numbers. Mixed methods combines the two families, either simply or complexly, and it is the term most often used loosely. A survey with one open comment box is not mixed-methods research; it is a quantitative survey with a comment box.

Strategy

The fourth ring names the vehicle: experiment, survey, archival research, case study, ethnography, action research, grounded theory or narrative inquiry. Strategies are not exclusive, and the better dissertations combine two deliberately — a survey to establish the distribution of a phenomenon, then a case study to explain the outlying organisation.

The strategy is where feasibility usually bites. An experiment needs control the researcher rarely has. Ethnography needs access measured in months. A case study needs one organisation willing to talk, which is why so many are written about the student's own employer.

Time horizon

Cross-sectional research takes a snapshot; longitudinal research follows the same subjects over time. Almost every taught-programme dissertation is cross-sectional, for the simple reason that the submission date arrives before the second wave of data would. Saying so plainly is better than implying a longitudinal design the timetable never permitted.

Techniques and procedures

The centre of the onion holds the operational detail: sampling frame and size, the instrument itself, how the data will be collected, how it will be analysed, and the ethical approvals required. This is the layer most students draft first and the layer the onion places last, precisely because every choice in it is constrained by the five rings outside.

How the research onion layers constrain one another

The value of the diagram is not the list of research onion layers. It is the claim that the rings are nested, so a decision on the outside narrows what remains available on the inside.

Declare positivism and you have largely committed to deduction, to a mono-method quantitative design, to a survey or experiment, and to a probability sample with a stated confidence level. Declare interpretivism and the same chain leads towards induction, qualitative technique, a case study or ethnography, and purposive sampling where the defence is the richness of the case, not its representativeness.

The chains are tendencies, not rules, and a good methodology chapter names the point where it departs from the expected path and says why. What is not defensible is arriving at the centre by a route that never passed through the outer rings at all.

A worked example of peeling the research onion

Take a question a supervisor might actually approve: does the shift to hybrid working change how first-line managers in UK retail banking assess the performance of their teams?

Philosophy. The object of study is what managers count as evidence of performance, which is a matter of interpretation rather than an external fact. Interpretivism, with a note that the researcher's own experience of hybrid management is a source of bias to be declared.

Approach. Inductive. There is theory on performance management and theory on remote work, but the intersection in this sector is thin enough that starting from a hypothesis would import assumptions the data has not supported.

Methodological choice. Multi-method qualitative: semi-structured interviews plus documentary analysis of the appraisal forms actually in use. Two qualitative techniques, no numbers, so not mixed methods.

Strategy. Case study of two banks, chosen because one moved to a fixed three-day office pattern and the other left it to team discretion. The contrast is the analytical lever.

Time horizon. Cross-sectional, and the limitation stated: the study captures attitudes eighteen months after the policies were introduced and cannot show how they shifted during that period.

Techniques and procedures. Fourteen interviews, seven in each bank, purposively sampled to cover branch and telephone-channel managers. Thematic analysis following Braun and Clarke. Appraisal documentation requested under a confidentiality undertaking. Ethical approval before any contact, with consent withdrawable up to the point of transcription.

Written out this way the chapter takes perhaps eight hundred words and every sentence follows from the one above it. That is what the diagram is for.

What the model does not decide

Two things sit outside the onion and are regularly mistaken for part of it.

Access is the first. No ring asks whether the organisation will let you in, yet access determines more dissertations than philosophy does. A design that survives every layer and then cannot recruit participants is not a design. Sound practice is to confirm access in principle before the strategy ring is settled, and to keep a second case in reserve.

Analysis technique is the second. The centre names data collection and analysis together, which encourages students to write "thematic analysis" and stop. Which variant, whose coding frame, how disagreement between coders was resolved, at what point coding stopped — none of that is prescribed by the model, and all of it is marked.

Where the model is criticised

Three objections recur, and acknowledging them tends to earn marks rather than cost them.

The first is that the onion presents philosophy as a choice the researcher makes, when in practice most researchers inherit a position from their discipline and their supervisor and then find the label that fits. The diagram tidies up a process that is rarely so deliberate.

The second is that the layers are not as tightly nested as the picture suggests. Mixed-methods work sits awkwardly on a single philosophical ring, and pragmatism was arguably added to the outer layer to accommodate designs the earlier versions could not place.

The third is practical. Because the diagram is easy to reproduce, it is frequently included as a figure and then ignored, with the chapter continuing exactly as it would have done without it. A figure that does not constrain the text underneath it is ornament.

Using it without padding the chapter

Reproducing the diagram and labelling the ring you have chosen at each level is worth a paragraph, not a section. The marks are in the justification: why this philosophy for this question, and what it ruled out. Examiners see the diagram several hundred times a year and stop reading it; what they have not seen is a candidate who says, in one clause, which alternative was rejected and on what ground. A sentence such as "a survey was rejected because the construct under study is how managers interpret an outcome, which a fixed scale would presuppose" does more work than a full page restating the model.

Two habits separate a methodology chapter that uses the Saunders research onion from one that merely displays it, and both are about the research onion layers instead of the picture. The first is naming the alternative at each layer and giving the reason it was rejected, in one sentence. The second is carrying the consequences forward, so that the limitations section refers back to the time horizon chosen in ring five, not listing generic constraints.

A worked example of the onion in use

A short example makes the layering concrete. A student wants to know why a retailer's staff turnover rose after a bonus scheme changed. Philosophy: the question is about how people understood the change, so interpretivism. Approach: no theory is being tested, so inductive. Strategy: a case study of that retailer. Choice: qualitative only. Time horizon: cross-sectional, because the change has already happened. Technique: semi-structured interviews with leavers and stayers.

Every layer in that example constrains the next, which is the whole claim the diagram makes. Swap the philosophy to positivism and the same question produces a survey, a hypothesis about bonus structure and turnover, and a sample large enough to test it — a different project reaching a different kind of answer.