Inductive Categorisation: A Complete Worked Example
Four consecutive slides reproduce two book tables and say five words about them. Those five words are the whole of the teaching commentary — so the rest of this page is the tables themselves, the arithmetic that checks them, and the point at which the example stops.
Four slides, and five words of commentary
The worked example occupies slides 11 to 14, all four of them images of book pages rather than slide text. Slide 11 is titled “Participant responses” and slide 12 “Participant responses continued...”. Slides 13 and 14 carry no text at all — not a title, not a caption, not a line.
Five words, then, of teaching commentary. The deck never says the example is a categorisation, never tells you what to notice, and never draws a conclusion. Everything below is either the reproduced tables or this library's reading of them, kept apart.
Table 17.2: the thirty-one responses as collected
The first table is a single column headed “Participant Responses”. Its title states the question that produced them and states that the list is unordered. The responses are reproduced exactly as printed, including spelling conventions that are not Australian; correcting them would misrepresent a reproduction.
TABLE 17.2 — PARTICIPANT RESPONSES, VERBATIM AND IN SOURCE ORDER
| # | Participant response |
|---|---|
| 1 | There is not enough space for everyone. |
| 2 | Our office furniture is dated and needs replacing. |
| 3 | We need a better cleaning service for the office. |
| 4 | We need more objective recruitment and hiring standards. |
| 5 | We need objective performance appraisal and reward systems. |
| 6 | We need consistent application of policy. |
| 7 | There are leadership problems. |
| 8 | Nonproductive staff members should not be retained. |
| 9 | Each department has stereotypes of the other departments. |
| 10 | Decisions are often based on inaccurate information. |
| 11 | We need more opportunities for advancement here. |
| 12 | Our product is not consistent because there are too many styles. |
| 13 | There is too much gossiping and criticizing. |
| 14 | Responsibilities at various levels are unclear. |
| 15 | We need a suggestion box. |
| 16 | We need more computer terminals. |
| 17 | There is a lot of “us and them” sentiment here. |
| 18 | There is a lack of attention to individual needs. |
| 19 | There is favoritism and preferential treatment of staff. |
| 20 | More training is needed at all levels. |
| 21 | There needs to be better assessment of employee ability and performance so that promotions can be more objectively based. |
| 22 | Training is needed for new employees. |
| 23 | Many employees are carrying the weight of other untrained employees. |
| 24 | This office is “turf” oriented. |
| 25 | There is a pecking order at every level and within every level. |
| 26 | Communication needs improving. |
| 27 | Certain departments are put on a pedestal. |
| 28 | There are too many review levels for our product. |
| 29 | Too many signatures are required. |
| 30 | There is a lot of overlap and redundancy. |
| 31 | The components of our office work against one another rather than as a team. |
The numbering is this library's; the source numbers nothing. The slide break falls after 16.
Table 17.3: the same responses in six inductive categories
The second table has two columns, headed “Inductive Categories” and “Participant Responses”. The word “Inductive” is the entire method statement for the example. The two right-hand columns below are added by this library and are what makes the second table readable against the first: the position each response holds in Table 17.2, and the size of each category.
TABLE 17.3 — CATEGORISATION OF THE RESPONSES, VERBATIM AND IN SOURCE ORDER
| Inductive category | Participant responses, in the order this table prints them | Positions in Table 17.2 | Size |
|---|---|---|---|
| Management issues | There are leadership problems. · We need a suggestion box. · There is a lack of attention to individual needs. · There is favoritism and preferential treatment of staff. · Decisions are often based on inaccurate information. · We need consistent application of policy. | 7, 15, 18, 19, 10, 6 | 6 |
| Physical environment | We need a better cleaning service for the office. · Our office furniture is dated and needs replacing. · We need more computer terminals. · There is not enough space for everyone. | 3, 2, 16, 1 | 4 |
| Personnel practices | We need more objective recruitment and hiring standards. · We need objective performance appraisal and reward systems. · Nonproductive staff members should not be retained. · There needs to be better assessment of employee ability and performance so that promotions can be more objectively based. | 4, 5, 8, 21 | 4 |
| Employee development | More training is needed at all levels. · Training is needed for new employees. · Many employees are carrying the weight of other untrained employees. · We need more opportunities for advancement here. | 20, 22, 23, 11 | 4 |
| Intergroup and interpersonal relations | This office is “turf” oriented. · There is a lot of “us and them” sentiment here. · There is a pecking order at every level and within every level. · Communication needs improving. · There is too much gossiping and criticizing. · Certain departments are put on a pedestal. · Each department has stereotypes of the other departments. | 24, 17, 25, 26, 13, 27, 9 | 7 |
| Work structure | There are too many review levels for our product. · Too many signatures are required. · Responsibilities at various levels are unclear · The components of our office work against one another rather than as a team. · There is a lot of overlap and redundancy. · Our product is not consistent because there are too many styles. | 28, 29, 14, 31, 30, 12 | 6 |
Columns 1 and 2 are the source's, responses separated by a middle dot. Columns 3 and 4 are this library's mapping; the source numbers nothing and states no size.
Reading the second table against the first
Run your eye down the third column. Every position from 1 to 31 appears exactly once across the six rows — the completeness and double-counting check, done, and the check the source leaves to you.
Two further things fall out. Source order is not preserved inside the categories — Management issues runs 7, 15, 18, 19, 10, 6 — so something reordered them and the source does not say what. And the last four responses all land in Work structure, where elsewhere neighbours scatter, against a title calling the list unordered. Both are checkable above; neither is adjudicated here.
Three checks to run on any categorisation, including this one
- Completeness — does every raw item appear in the categorised table? All thirty-one do.
- No double counting — does any item appear twice? None does.
- Wording — has any item been altered? One full stop, nothing else.
- All three are performed above by this library; the source performs and reports none.
What the example actually establishes
Read honestly, it demonstrates four things. Being strict matters: the temptation with the only worked example in a body of teaching material is to make it carry more than it holds.
Four things this example shows, and no more
What an open-ended question yields
Thirty-one short, unstructured, first-person statements about one organisation, in no stated order — the concrete form of an abstraction the notes only describe.
Categorisation reorders without rewriting
Every response appears in the second table in the same words. At this level it is a sorting operation on unaltered data, and the example shows that cleanly.
Categories come out uneven, and that is left alone
The six run from four responses to seven, the largest nearly twice the smallest. The imbalance is neither corrected nor mentioned.
Similar-looking responses land apart
Several could plausibly have gone elsewhere, and no reason is given. Two verifiable cases are below.
Equipment filed as premises
- “We need more computer terminals.” goes to Physical environment, with cleaning and furniture.
- Not to Work structure with the review levels and signatures, where a reading centred on how the work gets done would put it.
- Terminals are equipment; the category treats them as premises. The principle is not stated.
The finest boundary in the scheme
- “We need objective performance appraisal and reward systems.” goes to Personnel practices.
- “We need more opportunities for advancement here.” goes to Employee development.
- Both concern an employee's career. That boundary is the finest drawn and is nowhere stated.
The lesson is not that the example got these wrong. It is that a categorisation is a series of judgements of exactly this kind, that the judgements are the analytical work, and that one which does not record them cannot be audited by anyone — the analyst included.
The three things the example does not show
The second absence costs a reader most. You can see that the categories were arrived at; you cannot see how, and how is the part you need in order to do it yourself. The reason is structural rather than editorial — the tables are reproduced from a chapter that is not in the batch, and the derivation would have sat in the text around them.
What it is filed under, and where it stops
The example sits in the week's Part I, between a slide defining coding and a slide about analysis software. It does not code: no code, symbol, name or word is applied to any segment of any response. It categorises — and the week's vocabulary separates the two, since the notes make categories the input to coding while the deck defines coding as marking segments of data with symbols or names. A categorisation is filed under coding without comment.
Nor does it use the week's own process models. A four-step model sits on the slides before it and a three-stage diagram on the slide after; neither is mentioned on any of the four example slides. Those models, and the five accounts the week gives of what qualitative analysis consists of, are at the steps in qualitative data analysis.
And it stops early. No count per category is reported, nothing prioritised, no finding stated, no recommendation made, and the categorisation is never written up. The week's Part II explains how to write a results section and the subject's only dataset is never carried into one — see writing the results section.
Doing the same thing to your own data
The source supplies no procedure and this page does not invent one. What follows is about record-keeping rather than method — what to write down so your categorisation can be checked. It stops short of how to form the categories, because the material does not say.
What to record while you categorise (this library's guidance, not the source's)
Number the raw list before you sort anything
As Table 17.2 is numbered above. Without stable positions you cannot build the third column or show anyone what you did.
Log each category as you name it
The example shows six finished names and no history. The order yours appear in, and what prompted each, is the derivation this one is missing.
Record every response that could have gone two ways
The computer terminals response is one. Note the alternative you rejected and why — it is the part a reader will question.
Publish both counts, and say who checked
Sum your category sizes against your item count. Then report what a second coder agreed and disagreed with, or say plainly that nobody checked. An unchecked categorisation is still usable; one presented as though checked is not.
What to carry forward
- Thirty-one responses sorted into six inductive categories — both tables above in full, the only complete worked analysis in six weeks of teaching material.
- The counts are this library's, derived and checked: 6 + 4 + 4 + 4 + 7 + 6 = 31. The source states neither.
- Categorisation here reorders without rewriting: every response keeps its own words, one full stop aside.
- No rule for deriving the categories is stated, and nothing establishes that a second analyst would have produced the same six.
- Record the derivation in your own work — the example that would have taught it is missing its explanation.
Frequently asked questions
How many responses and how many categories are there?
Thirty-one responses and six categories, with 6, 4, 4, 4, 7 and 6 responses in them. Both figures are derived here by counting the reproduced tables; the source states no count anywhere, on either slide or in either table title. The category sizes sum to thirty-one, so every response is used once and none is used twice.
What method was used to derive the six categories?
The source does not say. The column heading reads “Inductive Categories” and that word is the whole of the method statement. No procedure, criterion, target number or stopping rule is given anywhere in the week, and this page does not attach a named method to the example, because doing so would put words in the source's mouth.
Why can I see the categories but not how they were reached?
The tables are numbered 17.2 and 17.3, which means they are reproduced from chapter 17 of a book that is not in the supplied material. In a textbook the derivation is explained in the surrounding chapter text, and that text was not supplied with the slides. The gap is in what was provided, not in the original analysis.
Did anyone check the categorisation?
Not in anything supplied. No second coder, no percentage agreement, no audit, no member check and no peer debrief is reported or mentioned. Nothing in the material establishes that another analyst working from the same thirty-one responses would have produced these six categories.
Can I reuse these six categories on my own data?
No. They are the categories one analysis produced from one organisation's answers to one question, reproduced here as an example of the operation rather than as a scheme. Inductive categories are derived from the data in front of you; borrowing someone else's makes the exercise deductive and changes what your results can claim.
Does the example show what to do after categorising?
No. It reports no count per category, prioritises nothing, states no finding and makes no recommendation, and the categorisation is never written up. The week's second half explains how to write a results section and never carries this dataset into one.
References and source attribution
- Supplied teaching source, Week 12: Analysing qualitative data and presenting results — slide deck, slides 11–14. Four images of book pages reproducing Tables 17.2 and 17.3; the deck's own text across all four slides is “Participant responses” and “Participant responses continued...”.
- Tables 17.2 and 17.3 of an unnamed book chapter 17, reproduced as slide images. The chapter title, author, publisher and year are not given anywhere in the supplied material and no Table 17.1 appears, so a complete reference cannot be constructed from what was provided.
- Supplied teaching source, Week 12 — slide 9 (the eleven-term glossary, which defines inductive codes as codes developed whilst data is examined) and slide 10 (the third of the week's three definitions of coding), the slides immediately before the worked example.
- Supplied teaching source, Week 12 — study notes, section 3, “Establishing Trustworthiness”: the auditability, dependability and confirmability criteria against which this page measures the example.
- O'Leary, Z. 2017, The Essential Guide to Doing Your Research Project, 3rd ed., Sage Publications, London. The supplied source also cites the 2010 edition, Doing your research project, as the origin of the week's three-stage analysis diagram.
Suggested questions for Ask KEVOS
- Show me the thirty-one responses and which category each one ended up in.
- What would I have to write down for my own categorisation to be auditable?
- Which responses in this example could defensibly have gone into a different category?
- What does the supplied material say about deriving inductive categories, and what does it leave out?
- How does this worked example compare with the four-step and three-stage models in the same week?
- What should I report about coder agreement if I am the only person analysing my data?