# Qualitative Content Analysis: A Worked Coding Example

A worked qualitative content-analysis grid: define categories, pilot them on responses, and describe the material without overclaiming.

- Canonical URL: https://www.harmate.com/en/blog/qualitative-content-analysis-coding-example
- Author: Harmate Team
- Published: 2026-10-25
- Updated: 2026-10-25T01:14:40.367665+00:00
- Language: en

## Content

# Qualitative Content Analysis: A Worked Coding Example

A coding grid applies defined categories to a specific body of material. It can help describe what appears in interview or survey responses when each category has a bounded meaning, an example, and a rule for using it. The table does not decide what people’s words mean on the analyst’s behalf.

The example below shows qualitative content analysis using a few fictional responses. It is neither a universal method nor an automated classifier. A person defines categories from a question and context, then may revise them when the material does not fit the grid cleanly.

## What this grid is meant to describe

Imagine a training provider wants to understand which conditions prevented some participants from using a new procedure at work. The question is bounded: “In situations you encountered after training, what helped or prevented you from using the procedure?” It is not intended to evaluate people or, on its own, explain a workplace result.

Content analysis can fit when the question and corpus are bounded enough to develop and revise categories. Philipp Mayring discusses choosing units of analysis, coding rules, developing or revising categories, and anchoring categories in examples from the material. He also notes that this approach is less suitable when the inquiry is very open and the relevant categories have not yet been delimited. See [Qualitative Content Analysis (Mayring, 2000)](https://www.qualitative-research.net/index.php/fqs/article/view/1089/2385), especially paragraphs 7, 12, 15, 17, and 27.

This focus differs from reflexive thematic analysis, which develops an interpretation of patterns of meaning; themes are not simply sorting categories. For that approach, see [codes, themes, and interpretation](/en/blog/thematic-analysis-codes-themes-and-interpretation). If the question concerns how meaning is built through turns in an exchange, [discourse analysis](/en/blog/discourse-analysis-read-an-exchange-without-guessing-intentions) addresses a different object.

## Separate the coding unit from its context

Before writing categories, specify three things:

- **The corpus:** which responses are included, for which question, period, and filters?
- **The context unit:** what portion will you reread to understand a passage? Here, the respondent’s full answer to the question.
- **The coding unit:** what passage receives a category? Here, a sentence or a few sentences describing a situation related to using the procedure.

The full response can remain the context even when only one passage carries the coded information. If an answer describes two different situations, retain both passages and their context. Do not reduce the source to a keyword.

## Fictional example: revise the grid after a pilot

Suppose the first three answers in a pilot are:

| Fictional response | Passage selected as the coding unit | Initial reading |
|---|---|---|
| R-01 | “The template is in a space my account cannot open.” | An access barrier may be described. |
| R-02 | “The paper guide still showed the old order of steps.” | The support may no longer match the procedure. |
| R-03 | “I did not handle any case like that this week.” | The respondent does not describe a failed attempt; there was no relevant opportunity. |

A single category called “barrier to use” would be too broad to distinguish R-01 from R-02 and would not fit R-03. After reading the pilot, the analyst revises the grid, makes its criteria explicit, and recodes all three responses:

| Category | Include when… | Exclude when… | Fictional anchor example |
|---|---|---|---|
| Resource access | The passage reports a resource-access obstacle during the period, even if it was later resolved. | The document is accessible but its content is old or contradictory. | R-01: “my account cannot open” |
| Version or instruction to check | The passage says the material or instructions consulted did not match the procedure in force. | The respondent simply had no situation in which to use it. | R-02: “the old order of steps” |
| No relevant opportunity reported | The passage describes no work situation in which the procedure could be tried during the period. | An opportunity existed, but an access or instruction issue is reported. | R-03: “no case like that this week” |

Definitions, inclusion and exclusion rules, and examples anchored in responses make the grid open to discussion and revision. They do not turn it into a general dictionary of words. If a later answer fits none of the categories, mark it as uncovered, reread its context, and decide as an analyst whether to revise a category or leave the passage outside scope. After a revision, revisit responses already coded so the same version of the grid is applied throughout.

This build-and-pilot sequence is consistent with Mayring’s approach: categories can be developed from the material and revised, with definitions grounded in examples and explicit rules.

## Describe the result without confusing respondents and mentions

Add a fourth fictional answer: “I could open the template only after a colleague gave me the right access.” In this small corpus, two of the four fictional respondents report at least one resource-access obstacle. The counting unit is the respondent, not each passage. R-04 reports an obstacle later resolved; R-01 does not report whether it was resolved. Record that status separately rather than treating the situations as equivalent. The denominator is these four available responses, not people who did not answer. That count describes only these four fictional responses. It does not measure prevalence among everyone invited, and it does not establish that access permissions caused a general problem.

Also define what you count. If the result unit is the respondent, count each category at most once per respondent, even when it is mentioned several times. If you count every coded passage or mention, report that separately: one person may contribute more than once. Do not compare these denominators as if they answered the same question. Missing or partial answers are neither an absence of difficulty nor a “none” category; report their number and the available scope.

Your working table can include the available response identifier, question, coding unit, context unit to reread, category, excerpt, and a limitation note. Keep the source text intact and review ambiguous cases. Do not turn a descriptive category into a judgment about a respondent’s personality, intent, or competence.

To prepare multiple exports and check their provenance before coding, see the [guide to preparing survey responses before analysis](/en/blog/prepare-survey-responses-before-analysis-a-practical-method-and-example). For a larger open-response corpus that connects codes, frequency, and a decision, the [seven-step open-response method](/en/blog/analyzing-200-open-responses-without-bias-a-method-for-actionable-decisions) addresses the broader task.

## Key takeaways

A useful coding grid names its corpus, context unit, coding unit, categories, and rules. It is piloted, open to revision, and keeps responses that do not fit visible. Categories describe the selected material; they do not prove a cause or a rate for an entire population.

If you collect or import open responses, [Harmate Open Questions](https://www.harmate.com/en/product/open-questions) can help organize responses and return to source excerpts associated with themes. The analyst remains responsible for choosing the method, grid, and interpretation.