calendar
Edited on October 13, 2026

Knirk and Gustafson Model: The 3 Stages, With an Example

The Knirk and Gustafson model from the 1986 book: problem determination, design and development, worked through one course, with a real case and templates.

Carolina Martin, Learning Experience Manager at Coursebox AI
Carolina Martin
Customer Success Manager
Key takeaways

The Knirk and Gustafson model is a three-stage instructional design process from Frederick Knirk and Kent Gustafson's 1986 textbook: problem determination, design and development, with a check on whether instruction is a solution in the first stage and tryout and revision in the last.

  • The book's figure and its part titles use slightly different names for the same three stages.
  • A Singapore health campus used the model's design stage to write objectives for three quality improvement courses.
  • The book teaches objectives in ABCD form, which names the audience, the behaviour, the condition and the degree.
  • In course requests typed into Coursebox over 90 days, 0 of 32 workplace requests said what was going wrong on the job.
What is missing from your course request?

Five questions, about a minute. Answer for one course you have been asked to build for staff, such as an induction, a procedure or a safety task.

1
2
3
4
5
Have you written down what is going wrong on the job, and how you know?
Instructional Designers

Go from storyboard to an interactive course in minutes

Take a hypothetical family-owned winery in South Australia. In the first three weeks of last vintage, its routine hygiene swabs failed on 9 of 40 hose and pump checks, all on gear cleaned on night shifts, and one tank of white juice needed extra treatment. The cellar supervisor's first request was "a hygiene course for the casuals". When someone looked closely, two of the causes had nothing to do with what the casuals knew.

I'm an instructional designer at Coursebox, where we make instructional design software for training teams, so I'll say where I tested our own product. Frederick Knirk and Kent Gustafson's 1986 model starts where that supervisor should have started: with the problem, and with whether a course is the fix at all. I set out the three stages as the book's own figure labels them, walk the winery course through each one, and finish with a product test and two free templates.

What is the Knirk and Gustafson model?

The Knirk and Gustafson model is a three-stage instructional design process from Frederick Knirk and Kent Gustafson's 1986 textbook: problem determination, design and development, with a check on whether instruction is a solution in the first stage and tryout and revision in the last.

Knirk and Gustafson’s model groups work into problem determination, design and development. Problem determination includes the problem, entry skills, goals and organisation; design connects objectives, strategies and media; development includes materials, implementation, result analysis and revision. The high-level groups do not put evaluation outside the model.

View the full-size diagram

It comes from their book Instructional Technology: A Systematic Approach to Education, published by Holt, Rinehart and Winston in New York. The authors call it their instructional development model, and they use it to organise the whole book. Their description of the first stage is plain: "The first step of the instructional development model used in this text involves identifying the problem."

The book is a lending-only scan on the Internet Archive, so I read it through the archive's search of its scanned text. That gave me the table of contents, the figure captions and the labels inside the authors' model figure, but not full pages. What I say about the stages comes from those labels and short passages of the text, and I say where I am interpreting.

Three points set the scope:

  • It is a textbook model. The authors assume their reader has "little background", and the book's examples include schools as well as training. The model is the frame for the chapters, not a research finding.
  • It has three stages and a management strand. The figure labels the stages problem determination, design and development. The authors call each block a "subsystem", and they add a management subsystem for organising, evaluating and reporting on the project.
  • It has fewer stages, and it came later. It is a three-stage alternative to ADDIE and the Dick and Carey model, and it is younger than both: the process later called ADDIE dates from 1975, and Dick and Carey was first published in 1978.

Here is the model at a glance.

StageThe book's part titleWhat the authors' figure includesThe question it asks of your course
1. Problem determinationProblem definitionIdentifying the problem, needs assessment, task analysis, entering behaviours, broad instructional goals, deciding whether instruction is a solution, and organising the projectWhat is going wrong, how do you know, and will a course fix it?
2. DesignInstructional designWriting objectives, then selecting strategies and mediaWhat will learners be able to do, and how will they learn it?
3. DevelopmentInstructional developmentSearching for existing materials, developing new ones, testing prototypes, conducting a tryout, analysing results, revising and implementingDoes it work with real learners, and what has to change before release?
note:Two sets of names for the same three stages

The figure labels the stages problem determination, design and development. The book's three parts are called problem definition, instructional design and instructional development. Both sets refer to the same structure, so cite the figure labels when you name the stages and the part titles when you point to chapters.

For choosing between this and other models, the guide to instructional design models compares them side by side.

Who were Knirk and Gustafson, and which source to read

Frederick G. Knirk was at the University of Southern California when the book came out. Kent L. Gustafson was a professor of instructional technology at the University of Georgia by 1991, and he was the lead author of the Survey of Instructional Development Models, a catalogue of design models that ran to four editions between 1981 and 2002.

  • Read the 1986 book first. It is the only edition I found. Part One covers problem identification, needs assessment, task analysis and project management; Part Two covers objectives, strategies and media; Part Three covers materials, tryout and revision.
  • Read Gustafson and Branch's survey for context, not for this model. The 2002 edition explains how design models differ by scale. It does not include the Knirk and Gustafson model, and neither do the 1981, 1991 or 1997 editions.

The 3 stages, worked through one course

Here is the example I'll use from here on. Its figures are illustrative. The winery crushes about 1,500 tonnes a year and hires 8 casual cellar hands for each vintage, from February to May.

  • Learners: new casual cellar hands. Some have worked a vintage elsewhere, most have no winery experience, several hold a forklift licence and some speak English as an additional language. None of that shows they can clean a hose to this winery's standard.
  • People: the cellar supervisor owns the procedure and the roster, a senior cellar hand demonstrates and signs off, and the winemaker owns wine quality.
  • Task: cleaning, sanitising, draining, capping and racking transfer hoses and pumps after every wine movement, to the winery's five-step hose hygiene card: flush, sanitise, drain, cap both ends, rack off the floor.
  • Intended outcome: every new cellar hand cleans and stores hoses and pumps to the card from their first shift. Winemaking, chemical handling and tank entry are out of scope.

The example suits the model because its first stage changes the brief. Two of the causes are not training problems, the objective comes straight from the task, and development includes a tryout with a measurable check, the next hygiene swab. The five-step card is an invented teaching summary, not a complete cleaning or chemical procedure. Real work uses the current approved procedure and authorisation. The course uses the winery’s own card and the chemical supplier's instructions, and states no sanitiser concentrations or temperatures.

1. Problem determination

The first stage defines the problem before anything is designed. The authors' figure puts several jobs in this block: identify the problem, gather needs assessment data and analyse the task, identify learners' entering behaviours and broad instructional goals, organise the management of the project, and "Determine if instruction is a solution". The box "Determine if instruction is a solution" decides whether a course should exist at all.

  • What you decide: what is going wrong, how you know, which causes training can fix, who owns the rest, what learners can already do and the broad goal.
  • In the winery course: the senior cellar hand watched three night-shift clean-ups and talked to the casuals. Four causes came out, and only some of them belong in a course.
Cause foundCan training fix it?Who owns the fix
New hands drained hoses but did not cap both ends, and did not know why capping countsYesThe course, then the senior cellar hand's sign-off
Hands skipped the card when the shift was busy, and most had never been shown itPartly: they need to see the card and practise using it under time pressureThe course, with the card hung at the hose rack
The sanitiser station ran empty on night shifts because refills were rostered on day shift onlyNoThe cellar supervisor, through the roster
The hose rack was moved in a shed reorganisation, so hoses were coiled on the floorNoThe cellar supervisor, by putting the rack back

Entering behaviours are what learners can already do before the course. Most of these casuals have no winery hygiene experience, can follow a picture card and do better with plain English and photos. They start the day before vintage, which leaves about an hour of training before their first shift. The broad goal follows: every new cellar hand cleans and stores hoses and pumps to the card, every time, from their first shift.

Action mapping asks the same question in more detail, for each thing people need to do on the job. The Knirk and Gustafson model puts it in a single box, which makes it easy to skip.

2. Design

Design turns the goal into objectives, then chooses how people will learn. A chapter title in the book calls objectives "the cornerstone", and the book teaches the ABCD method for writing them: the audience, the behaviour, the condition and the degree. The strategies chapter then turns to how to "select strategies and media which maximize opportunities for learning".

  • What you decide: an objective you could watch someone meet, then the strategy, media and an assessment that checks that objective.
  • In the winery course: "Given a used transfer hose and pump at the end of a wine transfer and the hose hygiene card, the new cellar hand flushes, sanitises, drains, caps both ends and racks the hose off the floor, in the order on the card, without prompting, with a pass on the next hygiene swab."

That one sentence is four parts. The new cellar hand is the audience, the five steps are the behaviour, the used hose and the card are the condition, and the order without prompting plus a passing swab are the degree. For more on writing objectives this way, see measurable learning outcomes.

The rest of the design follows from the objective:

  • Strategy: a short phone-friendly module before the first shift, with last vintage's failed swabs as the story and the card step by step with photos; then a demonstration by the senior cellar hand, supervised clean-ups until the approved readiness criteria are met, then an observed sign-off.
  • Media: the photo card at the hose rack as the job aid, a short video of a correct clean-up shot in the winery, and the online module.
  • Assessment: an observed sign-off against the five steps plus the swab result, with a short check in the module of why capping and racking count. A recall quiz on step numbers would not show that anyone can do the job.

The design stage is also the part of the model I found used in a published workplace programme.

Two nurses in blue scrubs talk with a doctor in a white coat at a hospital nurses' station, with a whiteboard on the wall behind them
RDNE Stock project / Pexels

Case study: Quality improvement courses at a Singapore health campus

JurongHealth Campus is a regional health system in Singapore, made up of a 700-bed general hospital and a 400-bed community hospital. It wanted more of its staff able to run quality improvement (QI) projects, set a target of 25% of staff trained, and needed courses for staff at different levels.

  • What they did: The campus's quality team wrote the training objectives "following the design stage of the Knirk and Gustafson Instructional Design Model", in ABCD form with Bloom's taxonomy, and built the content with Gagné's nine events. Three instructor-led courses followed: a half-day QI beginners course, a full-day QI fundamentals course and a half-day QI leadership course. The curriculum was piloted with groups of staff from August to October 2020, adjusted, given trainer notes and rolled out in 2021.
  • Evidence and result: In Ng and colleagues' 2024 paper in BMJ Open Quality, the share of staff trained in at least one QI course rose from 11.3% (470 staff) in January 2020 to 22.0% (909 staff) in March 2024, short of the 25% target. Staff rated the beginners and fundamentals courses 4.58 out of 5 on average (the paper does not say how many responded). Of 840 staff trained over three years, 439 still worked there.
  • Limitation: Only the design stage is credited to the model, and the courses also drew on Bloom and Gagné. The authors say their evaluation captures Kirkpatrick levels 1 and 2 only, reaction and learning, and that the link to the organisation's performance indicators has not been established. The clinical improvements in the paper's abstract came from projects run in 2017 to 2019, before these courses existed.
  • Lesson: The design stage can be used on its own: it gave the team one method for writing objectives for three courses at three staff levels. The team also piloted the courses with staff before rollout, a step the paper does not attribute to the model. To claim more than reach and ratings, plan a measure of what trained staff do differently from the start.

3. Development

Development builds the course and tests it before release. The figure's development block covers sources of material (search for existing materials, revise them or develop what is missing) and the boxes that do the quality work: "Test prototypes", "Conduct tryout", "Analyze results", "Revise materials", "Modify based on data", and "Stop revision when objective is met". The text is direct about it: Their text calls for trying out and revising the materials.

  • What you decide: what already exists, who will try the course before release, what you will record, what you will change and when you stop.
  • In the winery course: the card and the supervisor's photos already exist, so only the module and the video are new. The week before vintage, two returning casuals and two new hands brought in early do the module and one supervised clean-up each. The senior cellar hand notes where they hesitate, and the gear is swabbed.

In this illustrative tryout, both new hands cap one end only, because the card's photo shows a single cap. The fix is a new photo and the words "both ends" on the card and in the module, then a re-check. After release, the swab results from weeks 1 to 3 of vintage are compared with last vintage's 9 of 40.

warning:The model does not leave out evaluation

The model does include evaluation. The book's development stage contains a tryout, analysis of results and revision until the objective is met, and the authors call for formative evaluation while materials are developed and summative evaluation at the end. Plan the tryout before you build, or development turns into production with nothing to revise against.

How the three stages fit together

The book reprints its model figure at the start of chapters with one subsystem highlighted, so the three stages read as parts of one model rather than separate checklists. The authors place the analysis and evaluation inside the development subsystem block, and they write of "revision cycles".

I read the box labels, but not the arrows or the order of every box, so I don't treat the steps inside each stage as a fixed sequence. In practice the stages do not stay sealed. A tryout can show that an objective was wrong, which sends you back to design, and it can expose a cause that training cannot fix, which sends you back to problem determination. In the winery course, a hose left unsanitised because the sanitiser station ran empty is not a course problem, however many times the course is revised.

How to use the Knirk and Gustafson model to design a course

This is the order I work in. It follows the three stages, but the numbering is mine, not the book's.

  1. Write down what is going wrong on the job and how you know, with a number if you have one.
  2. List the causes and mark which ones training can fix. Give every other cause an owner and keep it out of the course.
  3. Note what learners can already do, and set the broad goal.
  4. Write one objective in ABCD form for each thing learners must be able to do.
  5. Choose the strategy, the media and an assessment that checks the objective as written.
  6. Look for existing materials before building anything new.
  7. Try the course with a few learners, record where they struggle, revise and try again until the objective is met.
  8. Release it, and keep checking the result you started with.

The three-stage plan for the winery course

StageDecisionWinery course
Problem determinationThe problem and the evidence9 of 40 swabs failed in the first three weeks of last vintage, all on night-shift gear
Problem determinationCauses training can fixHands not capping both ends; hands skipping the card when busy or never shown it
Problem determinationCauses it cannot fixSanitiser station empty on night shifts; hose rack moved. Both go to the cellar supervisor
Problem determinationEntering behavioursMostly no winery hygiene experience; can follow a picture card; about an hour before the first shift
DesignObjectiveFive steps in the card's order, without prompting, with a pass on the next swab
DesignStrategy and mediaPhone module, demonstration, supervised practice to the readiness criteria, photo card at the rack, short video
DesignAssessmentObserved sign-off against the five steps, plus the swab result
DevelopmentExisting materialsThe card and the supervisor's photos; only the module and video are new
DevelopmentTryoutTwo returning and two new casuals the week before vintage; note hesitations; swab the gear
DevelopmentRevision and stop ruleFix the one-cap photo and add "both ends"; stop when the objective is met
Knirk and Gustafson three-stage planning worksheet

Plan a course in the model's three stages: the problem and its evidence, which causes training can and cannot fix with an owner for each, entering behaviours and the goal, an ABCD objective builder, strategy, media and assessment, and a development plan with the tryout and stop rule. Filled in for the winery course, with a blank version for your own. Editable Word file, no email required.

What Coursebox course requests leave out

In the course requests I looked at, the first stage is the one that went missing. I looked at course requests typed into Coursebox in the 90 days to 7 October 2026: one per person, from English-language browsers, with no uploaded file. Of 266 prompts, 104 were course requests and the combined classifications flagged 32 as for workplace training. Two independent AI readers classified every request against written definitions, I re-read every positive for the problem and tryout questions, and the counts combine requests flagged by either reader. They are classification estimates, not statistical upper bounds.

0 of 32
workplace course requests said what was going wrong on the job, gave evidence of a problem or asked whether training was the fix; across all 104 requests, the combined classifications flagged 1 as naming a specific gap
Source: Coursebox course requests, PostHog, 90 days to 7 October 2026

The requests that went beyond a topic started at the design stage. The combined classifications flagged 33 of the 104 named teaching strategies or activities and 30 asked for assessments. None planned a tryout with learners: the closest, a clinic onboarding brief, flagged two policy items for leadership sign-off.

The sample is small and covers what people typed, not finished courses, and many people may analyse the problem before they open a course builder and never type it. Still, a builder only knows what the brief tells it. A brief that says only "a hygiene course for the casuals" gives the builder no way to know that two of the causes belong to the supervisor, not the course.

tip:Put the problem in the brief

Before you ask anyone, or any tool, to build a course, write three things into the brief: what is going wrong and how you know, the causes the course must address, and the causes it must leave alone because someone else owns them. Name the owners too, so nobody turns a roster problem into a quiz question.

Knirk and Gustafson vs ADDIE and Dick and Carey

All three cover analysis, design, development and some form of evaluation. They differ in how much they spell out.

FeatureKnirk and GustafsonADDIEDick and Carey
OriginKnirk and Gustafson, 1986 textbook1975 process for the US armed services; the ADDIE name spread from the late 1980sDick and Carey, 1978; 9th edition 2021
Stages3: problem determination, design, development5: analysis, design, development, implementation, evaluation10 steps, from the goal to summative evaluation
Distinctive stepDeciding whether instruction is a solution, inside the first stageImplementation as its own phaseFormative evaluation with learners as its own step
Where evaluation sitsTryout, analysis of results and revision inside developmentEvaluation is named as a phase and informs decisions throughoutSeparate formative and summative steps
Best forOne designer or a small team that wants a short, defensible sequence with the training check up frontProjects where stakeholders expect a familiar plan with sign-offsHigh-stakes skills where step-level detail and formal tryouts are worth the time
Weak spotShort summaries omit much of the model; the original book is hard to accessToo broad to guide decisions on its ownHeavy for small projects
Use it whenThe first question is whether a course is the fixYou need a shared project structureGetting the skill exactly right justifies more analysis

Pick ADDIE when the project needs a shared structure with named implementation and evaluation phases. Pick Dick and Carey when formal tryouts with learners and step-level analysis justify the time. Pick Knirk and Gustafson when you work alone or in a small team, need a short plan you can explain, and the first real question is whether training is the answer. That fit is my judgement from the model's three stages; the book does not say it.

Where it sits among design models

Gustafson's survey sorts models by what they produce: classroom models for one or a few hours of instruction, product models for packages others will deliver, and system models for whole courses aimed at an organisation's problems. The taxonomy began with four categories in 1981 and settled on three in later editions. The Gerlach and Ely model and the Kemp model are both classroom models in the survey, and Dick and Carey is a system model.

None of the four editions of Gustafson's survey places his own co-authored model, so any placement is a reading, and here is mine. Its needs assessment, its check on whether instruction is a solution, its management subsystem and its tryout-and-revise development stage are features the survey links with system models. Its school examples and its readers, educators who may work alone, look like the classroom column. I'd put it between the two.

Limitations of the Knirk and Gustafson model

  • Thin in print. The three labels carry little guidance on their own, and the book had one edition. The step that decides whether instruction is a solution sits inside a single box, so it is easy to drop when the model is summarised.
  • No tested evidence. I found no study that tested the model as a whole. The published uses I found apply part of it, or report one group with no comparison: a 2011 university course in Thailand built through a Thai textbook's version of the three stages and reported posttest scores above pretest scores for 54 first-year students, with no control group.
  • Designers carry out goals set by others. The book says an instructional technologist "is primarily responsible for implementing policy decisions", as quoted in a 1989 paper by Denis Hlynka. In a workplace, that can sit awkwardly with deciding whether instruction is a solution, so agree up front who can say "this is not a training problem".
  • Little on what happens after release. The development boxes I read evaluate the materials against the objective, their effectiveness and their costs. I found nothing that plans a check of on-the-job behaviour months later, so pair the model with the Kirkpatrick model if you need that.

A programme that stops at reaction and learning shows the same gap. JurongHealth Campus, which used only the model's design stage, reported staff’s self-ratings of skills and knowledge before and after the beginners course:

3.14 to 4.29
QI beginners’ self-ratings of skills and knowledge before and after the course, out of 5, at JurongHealth Campus; the paper gives no respondent count
Source: Ng and colleagues, BMJ Open Quality, 2024

The paper's authors say their evaluation reaches Kirkpatrick levels 1 and 2 only. These are self-ratings, not an objective skill test or evidence that QI projects improved.

Using the Knirk and Gustafson model with an AI course tool

On 12 October 2026 I gave Coursebox, the course builder we make, a brief for the winery course written stage by stage. Coursebox is our product, so weigh my account with that in mind. The brief started with the problem determination: the 9 of 40 failed swabs, what the cellar team found, and the two supervisor fixes that must stay out of the course. Then came the ABCD objective word for word, the design, and a development section with a trainer page for the tryout. I asked for seven sections in a set order and no final quiz.

After one revision asking for exactly seven sections, the second outline had them in order, and the rendered pages kept most of what the first stage had settled:

  • "Last Vintage Warning" used the 9 of 40 failed swabs, the night shifts and the extra treatment, and said what the team had noticed.
  • The empty sanitiser station and the moved rack did not appear as learner tasks.
  • The five steps came in the card's order, with no sanitiser concentration; step 2 sent learners to the winery card and the chemical supplier.
  • "Observed Sign-Off" was a five-step tick list plus a swab, not a quiz.
  • "Trainer Tryout Notes" was a separate page for the senior cellar hand: two returning casuals and two new hands, one supervised clean-up each, hesitations noted, the gear swabbed. That is the development stage made visible.

One page fell short. I had asked for three busy night-shift situations, such as the next transfer starting in five minutes. "Busy Night-Shift Choices" had one scenario, a hose with one end capped, plus three generic rule cards. The time pressure that the problem determination had found behind last vintage's failures was not practised at all.

I fixed it in the page editor. I rewrote two of the rule cards as the missing situations, each with what happens next and what to do instead: "The next transfer starts in five minutes", where leaving the caps and rack for later means the hose sits open on the floor through the next transfer, and "It looks clean already", where skipping the flush and sanitise leaves residue for the next swab to find. After a reload, both situations and their outcomes showed in the learner view, and the generic "Use the card" card was gone.

I checked the outline and the rendered pages. Nobody has taken the course, so the test shows what the pages ask learners to do, not whether anyone cleans a hose better. That is what a tryout and the first weeks of swab results would show.

Tryout and revision log for the development stage

Plan a small tryout before release and record it: who tried the course, what they did, where they hesitated, the check result, the change you made and the re-check, with a stop rule tied to the objective. Filled in for the winery course with illustrative entries, with blank pages for your own. It gives no score. Editable Word file, no email required.

Putting the Knirk and Gustafson model to work

The Knirk and Gustafson model's value lies in two parts that are easy to skip. The first stage asks whether instruction is a solution, and the last stage tries the course with learners and revises it until the objective is met. Use it when you need a short plan you can defend, and add the Kirkpatrick levels if you have to show what changed on the job.

Your next step: take one course you have been asked to build and write down what is going wrong, how you know and which causes a course cannot fix, with an owner for each. If that list leaves nothing for training, you have saved yourself a course.

Frequently Asked Questions

The Knirk and Gustafson model is a three-stage instructional design process from Frederick Knirk and Kent Gustafson's 1986 textbook: problem determination, design and development, with a check on whether instruction is a solution in the first stage and tryout and revision in the last.

Problem determination, design and development. Problem determination identifies the problem, assesses needs, analyses the task, notes learners' entering behaviours, sets broad goals and decides whether instruction is a solution. Design writes objectives and selects strategies and media. Development finds or builds materials, tries them out, analyses the results and revises them until the objective is met.

ADDIE has five phases and treats implementation and evaluation as phases of their own. The Knirk and Gustafson model has three stages, puts tryout and revision inside development, and adds a check on whether instruction is a solution in the first stage. The process later called ADDIE dates from 1975, so it is older than the 1986 model.

No. The model does not appear in any of the four editions of the Survey of Instructional Development Models (1981, 1991, 1997 and 2002), even though Kent Gustafson is an author of both the model and the survey. The survey sorts models into classroom, product and system categories.

Yes. The development stage includes testing prototypes, a tryout, analysis of results and revision until the objective is met, and the book calls for formative evaluation during development and summative evaluation at the end. It says less about checking job performance after release.

Carolina Martin, Learning Experience Manager at Coursebox AI

Carolina Martin

Customer Success Manager

Carolina is a highly experienced learning designer with tech start-up experience. She is based in Sydney and previously designed courses alongside Coursebox founder Travis in a previous job. She is passionate about delivering engaging learning experiences and brings the same perspective to customer experience too. She manages our customer support documentation and product testing processes.