eduKateSG Learning Node Series · 0069

How Performance Support Works | Put the Right Help at the Point of Need

A learner can know what to do and still fail at the exact moment the knowledge is needed.

The problem may not be ignorance. It may be that the task is rare, the sequence is long, the stakes are high, the environment is distracting, the rules change, or remembering every detail is an expensive use of human attention.

That is where performance support enters.

Performance support is guidance embedded close enough to the real task that a person can use it while deciding or acting.

A checklist beside a machine. A worked example beside a difficult spreadsheet. A decision tree inside a service desk. A rubric beside a student’s draft. A formula sheet during a complex engineering calculation. A prompt that appears only when a user reaches the relevant step.

The goal is not to replace learning. The goal is to stop demanding memory where memory adds little value, so that attention can remain available for judgement, interpretation and exception handling.

The 50-Second Read

• Training prepares capability before the task; performance support helps capability during the task.
• Useful support appears at the point of need, contains only what the user needs, and is easier to use than guessing.
• Checklists, job aids, decision tables, templates, examples, prompts and embedded help can all be performance-support tools.
• Support is especially useful for infrequent, high-consequence, rapidly changing or memory-heavy tasks.
• Not every task should be supported externally. Foundational knowledge and fluent skills still need to live inside the learner.
• Poor support can create dependence, interruption, clutter or false confidence.
• The design problem is to decide what must be memorised, what can be looked up, what should be automated and what still requires human judgement.
• Good support reduces avoidable cognitive load without removing the thinking that defines competent performance.

Canonical Owner Boundary

This page owns performance support: guidance intentionally positioned inside or immediately beside real performance. How Cognitive Offloading Works owns the broader use of external tools to reduce mental load. How Training Transfer Climate Works owns whether the receiving environment permits trained capability to be used. How Scaffolding Works owns temporary instructional support designed to fade as competence develops. Performance support asks a narrower operational question: what help should still be available when the person is doing the real work?

1. Training and Performance Support Solve Different Time Problems

Training is usually separated from the moment of performance. You learn first, then use later.

Performance support moves part of the information back toward the moment of use.

That distinction matters because human memory is not a perfect warehouse. Even a competent person may forget a rarely used sequence, confuse two similar rules, or remember the principle but not the current threshold value.

A system that insists everything must be recalled unaided may be testing memory rather than protecting performance.

2. The Point of Need Changes the Design

A forty-page manual can contain the correct answer and still fail as performance support.

At the point of need, the user has a live task. Search time, ambiguity and interruption matter. The support must therefore be findable, brief enough to navigate and shaped around the decision being made.

This is why a one-page decision table can outperform a comprehensive handbook during a time-sensitive task even though the handbook contains more knowledge.

3. Job Aids Are External Memory With an Interface

A job aid does more than store information. It organises information around action.

• A checklist answers: what must not be omitted?
• A flowchart answers: which branch should I take?
• A lookup table answers: what value applies in this case?
• A template answers: what structure should this output have?
• An annotated example answers: what does acceptable performance look like?
• A troubleshooting guide answers: what should I inspect next?

The information architecture should match the cognitive decision.

4. Performance Support Is Not the Same as Giving the Answer

A support tool can preserve judgement.

A pilot checklist does not fly the aircraft. A laboratory protocol does not interpret an anomalous result. A marking rubric does not decide whether an argument is persuasive. A surgical safety checklist does not replace surgical expertise.

The best support often removes low-value recall demands while leaving the high-value reasoning with the human.

5. Rossett and Schafer: Guidance That Informs Planning and Action

Allison Rossett and Lisa Schafer helped establish the modern performance-support vocabulary around repositories and tools that inform and guide planning and action. The Association for Talent Development summarises this tradition by distinguishing performance support from training and listing job aids, checklists, documentation, decision tools, templates and application-specific help among common forms.

Source: Association for Talent Development, Science of Learning 101: When to Build Performance Support.

6. Rare Tasks Are Natural Candidates

If a procedure is performed once every two years, maintaining every detail in fluent memory may be unrealistic and inefficient.

For rare tasks, the learning target may be partly procedural and partly navigational: know the principles, recognise the situation, know where the trusted procedure lives, and use it correctly.

This is different from a multiplication fact or common spelling pattern that should become rapidly retrievable because it is used constantly.

7. High-Consequence Tasks Need Omission Protection

Competence does not make people immune to omission.

When a process contains several necessary checks and skipping one can produce disproportionate harm, a checklist can act as a guardrail against memory failure.

The checklist should not become theatre. It must correspond to real failure points, be short enough to use and sit at a meaningful pause in the workflow.

8. Rapidly Changing Information Should Not Be Forced Into Long-Term Memory

Some facts expire faster than people can reasonably relearn them.

Current prices, policy thresholds, software menu locations, regulatory details and product specifications may be better retrieved from an authoritative live source than memorised.

The capability then shifts from raw recall toward source selection, interpretation and correct application.

9. Embedded Support Reduces the Search Gap

The farther the user must travel from the task to find help, the more likely support is abandoned.

Embedded support can appear inside software, beside equipment, on a form, within a workflow or through a context-sensitive prompt.

The design objective is not maximal interruption. It is minimal useful intervention.

10. Mobile Performance Support Extends Help Into Dynamic Work

Mobile systems can move support into locations where a desktop manual is unavailable. The Association for Talent Development reported findings from Yao Huang’s doctoral research on mobile performance support, including support for complex tasks in dynamic contexts and perceived benefits for confidence, efficiency and knowledge sharing. The evidence base is still smaller than for many mature instructional methods, so strong universal claims would be premature.

Source: ATD, Benefits of Mobile Performance Support Systems.

11. The Memory Question Comes First

Before building support, ask what truly needs to be in memory.

• Does the learner need this knowledge instantly?
• Will looking it up interrupt performance dangerously?
• Is the knowledge foundational for understanding later material?
• Does expertise depend on recognising patterns that require a large internal knowledge base?
• Or is the information mostly arbitrary, infrequent or changing?

Performance support is strongest when this boundary is deliberate.

12. Students Already Use Performance Support

A student planning an essay with a paragraph frame, checking a laboratory report against a rubric, using a formula sheet, or following a revision checklist is already operating with external performance support.

The educational question is whether that support helps the student perform the target capability or quietly performs too much of it.

13. The Examination Changes the Boundary

If the final assessment is closed-book, support must eventually be removed for capabilities the learner must produce independently.

If the authentic profession permits reference materials, calculators, standards, code libraries or checklists, training should also teach competent tool use rather than pretending expertise is unaided recall.

The correct boundary depends on the destination.

14. A Rubric Can Be Performance Support—or Noise

A rubric helps only if the learner can translate its language into decisions.

“Shows insight” is not very supportive if the student cannot recognise what insight looks like. A strong rubric connects criteria to observable features, examples and revision moves.

Support must be actionable, not merely official.

15. Search Cost Is Part of Cognitive Cost

If a learner spends four minutes finding the relevant paragraph in a support document, the system has imposed a new load.

Information retrieval therefore belongs inside performance-support design: naming, indexing, search, visual hierarchy, chunking, examples and placement all matter.

A correct answer that cannot be found in time is operationally close to no answer.

16. Support Should Match the Decision Granularity

A beginner may need a sequence. An intermediate learner may need a checkpoint. An expert may need only an exception rule or current parameter.

One support surface for every user can create either overload for experts or dangerous underspecification for novices.

Good systems can layer the support: quick cue first, explanation second, full reference third.

17. Performance Support Can Become a Dependency

If a learner consults a prompt for a fact that should become fluent, the prompt may prevent retrieval practice.

If software supplies every next step, the user may learn button-following without understanding the process.

This is the core danger: the support can protect immediate output while quietly weakening future independence.

18. Design for Failure Modes, Not for Decoration

Performance support should begin with error analysis.

• What is commonly forgotten?
• What gets confused?
• Where do novices take the wrong branch?
• Which step has a high consequence if omitted?
• What information changes too quickly to memorise?
• Where does the user stop and ask someone else?

The support should occupy the failure point.

19. CivDJ Cross-Domain Comparison: Cockpits, Kitchens and Code

A cockpit makes the principle visible. Pilots learn extensively, yet important procedures are still supported by checklists because the cost of omission is high and the operating environment is complex.

A professional kitchen uses labelled stations, mise en place, recipe specifications and tickets. The chef still needs judgement; the environment carries part of the sequence.

Software development uses documentation, code completion, tests and version history. Skilled developers do not memorise every API call. They retain conceptual models and retrieve changing detail when required.

Across domains, the same architecture appears: internalise what must be fast and generative; externalise what is arbitrary, volatile or omission-prone; preserve human judgement where context matters.

20. Rainbolt Missing-Node Scan: Where Support Usually Fails

• The support exists but is not where the task occurs.
• The document is technically complete but too long to navigate.
• The user cannot tell which version is current.
• The checklist contains so many items that every item loses salience.
• The tool supplies answers without teaching when those answers apply.
• The support uses expert vocabulary that novices cannot decode.
• The learner depends on prompts for knowledge that should become fluent.
• The support does not cover edge cases.
• The organisation trains one workflow but the live support reflects another.
• Nobody removes obsolete guidance after the process changes.

21. Build the Minimum Useful Support

More information is not automatically more support.

1. Identify the real task.
2. Find the recurring failure point.
3. Decide what must be recalled and what may be externalised.
4. Choose the smallest useful form: cue, checklist, table, flowchart, example or template.
5. Place it at the point of need.
6. Test whether a real user can find and use it under realistic conditions.
7. Measure whether errors, delay or unnecessary help-seeking fall.
8. Remove or revise support when the workflow changes.

22. Performance Support and AI

AI can make performance support conversational and context-sensitive, but it changes the risk profile.

A static approved checklist has a known boundary. A generative system may produce fluent but incorrect guidance, omit a constraint or overstate confidence.

For consequential tasks, AI support needs source grounding, visible uncertainty, escalation routes and clear separation between recommendation and authorised decision.

The design rule remains unchanged: support should increase reliable performance, not merely increase the amount of text available.

23. Measure the Real Outcome

A support tool is not successful because people opened it.

Useful measures can include task accuracy, completion time, omission rate, escalation rate, rework, confidence calibration, quality variation and whether users can still perform appropriately when the support is unavailable.

For learning contexts, also check whether support is helping later independence or merely improving the supported attempt.

24. Evidence and Limits

The performance-support field draws from human performance technology, cognitive psychology, instructional design, ergonomics and workplace learning. Evidence is heterogeneous because tools range from simple checklists to context-sensitive digital systems. Some claims are supported primarily by applied studies rather than large mature meta-analytic literatures.

That means the strongest conclusion is architectural rather than universal: external guidance can improve performance when memory is not the real target, the support is well matched to the task, and the system preserves the judgement and knowledge that users still need internally.

25. The Return Path

The worker reaches the rare procedure. The student reaches the complicated proof. The technician reaches the unfamiliar fault. The teacher reaches the decision that happens only twice a year.

Competence does not require pretending that the environment is empty.

Sometimes expertise means knowing. Sometimes it means recognising. Sometimes it means retrieving the right trusted aid and applying it correctly.

The best performance support does not do the thinking for you. It makes sure your thinking reaches the part of the task where thinking matters most.