Plan a course with backward design

Backward design plans a course in three stages: first the results you want, then the evidence that will show learners reached them, and only then the pages that teach. Learning Design has an optional mode for it, with an academic preset for school and university units and a performance preset for customer education and employee training.

The mode is per course and optional. Courses that keep Standard work exactly as before.

Turn the mode on

1. Choose Learning Design in the editor toolbar and open any tab.
2. In the workspace header, set Design approach to Backward design: academic or Backward design: performance.
3. For performance, set Learners to Customers (product adoption) or Employees. Customers is the default. This changes wording only, such as Adoption outcome or Business goal.

The tabs change to 1 · Results, 2 · Evidence (2 · Practice & evidence for performance), 3 · Learning plan and Alignment. The page addresses stay the same, so a link such as ?design=goals still opens the first tab. ?design=evidence opens the evidence tab; with the mode off it opens Goals & objectives instead.

Choosing Standard again hides the extra tabs and fields but keeps everything you entered. Switching back shows it again. Undo reverses any change of approach.

Start from Draft with AI

From Studio home, choose Draft with AI and set Design approach to Academic or Performance. Standard leaves this mode off. Plan course shows Results, Evidence, and Pages before anything is written. Edit the plan, then choose Build course. One build writes the results, the evidence, and every page. Performance plans the course for customers. After the course exists, set Learners to Employees when the course is for employees. See Draft from a prompt or document.

Stage 1: Results

1 · Results shows the results panels above the course goal and objectives.

Academic courses have four lists:

• Established goals: short references to the standards or program goals the unit serves, such as a standard code.
• Transfer goals: what learners can do on their own in new situations.
• Big ideas (understandings): what learners should come to understand. This is different from the Bloom’s level Understand.
• Essential questions: open questions that frame the unit.

Performance courses have three panels:

• Adoption outcome (or Business goal): a type, a one-sentence statement, a Metric with Measure, Baseline, Target and By when, and an optional Product signal such as automation_created. A product signal is a name only. Nothing is connected or collected.
• User roles (or Job roles): who the course serves, with jobs to be done, one per line.
• In-product tasks (or On-the-job tasks): the observable actions that drive the outcome. Each task can name its roles, a job to be done and a product signal.

Write performance objectives as tasks in the product. The objective field shows an example in that form.

What each objective supports links each objective to the transfer goals and big ideas (academic) or tasks (performance) it serves. Choose a chip to link or unlink it.

Is training the answer?

Training fixes only a gap in knowledge or skill. Before you build practice for a performance task, open Is training the answer? on the task and answer each question Yes, a problem, No or Not sure. The questions cover information, resources, incentives, knowledge and skill, capacity and motives, environment first.

Studio shows a Recommended remedy: Training, Documentation, In-product guidance, Product change or Incentive or process. A knowledge gap recommends training. Otherwise the first problem in the environment decides it: a resources problem recommends a product change, an information problem recommends in-product guidance for customers or documentation for employees, and incentives, capacity or motives recommend an incentive or process change. Use Your choice to override the recommendation.

A task whose remedy isn’t training shows Not a training problem and stays in the design. Copy “Not a training problem” list copies those tasks and their remedies, one per line, for your product or docs team. The check is advice only. It never blocks building or publishing.

Stage 2: Evidence

The evidence tab plans what will show that learners reached the results, before any page is written. Choose Add performance task, Add practice (performance only) or Add other evidence. A course can plan up to 16 evidence items.

For each item you can set:

• the kind and a format, such as In-product task, Scenario, Simulation, Quiz, Observation or Usage signal;
• a title, and for a performance task an optional Task brief;
• Success criteria, one per line, up to 8;
• the objectives it shows, and the transfer goals, big ideas or tasks it targets;
• an evaluation level (Reaction, Learning, Behavior or Results) and Measured after the course;
• Built as: the page, and optionally the question, that carries it. Items show Planned until they are built, then Built.

Performance courses also show a Measurement plan: the outcome’s result metric and each task’s behavior signal, labeled Measured after the course. The course itself doesn’t collect these.

If you delete an objective or a page that evidence points to, the evidence item drops that link. An item whose page is gone shows Planned again.

To remove an evidence item, choose its trash button (announced as Remove evidence followed by its number). This removes the planning item; it does not delete a course page or quiz.

Stage 3: Learning plan

3 · Learning plan is the blueprint. With the mode on, each page in the plan can also name:

• the evidence it Prepares for;
• the Roles it serves (performance). A page with no role serves All roles;
• a Focus of Transfer, Meaning or Acquisition (academic).

In performance courses, Show pages for filters the plan to one role. Pages for All roles always stay in the list. Learners still see every page; role paths for learners aren’t available yet.

Design checks

With the mode on, Alignment opens with a Design checks section. Each check says what it found and Why this matters, links to this section with Learn more, and offers a button that opens the tab where you fix it. Checks marked Look at are worth acting on; checks marked Note are for information, such as evidence that isn’t built yet.

Design checks are advice. They never count as alignment gaps, never appear in Review errors, and never block building, publishing or quizzes.

Check	When it appears
No evidence planned	An objective has no evidence item. If a page already teaches it, the check asks you to plan the evidence first.
Not built yet	An evidence item has no Built as page and isn’t Measured after the course.
Questions on other objectives	Evidence is built as questions that aren’t linked to the objectives it measures.
Doesn’t support a goal or task	An objective isn’t linked to any transfer goal, big idea or task in What each objective supports.
Never practiced	An objective is taught and assessed, but nothing practices it. This check also runs without the mode (see Coverage checks).
Teaching block not linked (note)	A teaching block on a page that serves objectives has its links switched off, or links only to objectives the page doesn’t serve. Open the page goes to it. Also runs without the mode.
Only indirect evidence	All of an objective’s evidence is indirect (a Usage signal, or the Reaction level), and no scored question or rubric criterion assesses it.
One kind of evidence (note)	Only one evidence format shows the objective. Scored quiz questions count as one format, Quiz. Add evidence in another format, such as a scenario.
No performance task (academic)	A transfer goal has no performance task in the evidence.
No transfer or meaning focus (academic)	No page in the learning plan has a Transfer or Meaning focus.
No outcome metric (performance)	The outcome or its Measure is missing.
No practice planned (performance)	A task whose remedy is training has no practice or performance evidence, and no practiced objective supports it.
Not a training problem (performance)	The quick check points to a remedy other than training.
Not checked yet (performance)	A task has no Is training the answer? answers.
Page without a task (performance)	A planned page isn’t linked to any task through its objectives.
Role without pages (performance)	A role has tasks but no planned page serves it.
No behavior measure (performance)	No task has a product signal and no evidence is at the Behavior level.

Two checks can ask for a suggestion. Save pending changes first. On No evidence planned, choose Suggest evidence. On No practice planned, choose Suggest practice or Draft this task as a page. While a request runs the button reads Suggesting… or Drafting…, and the buttons stay off while a suggestion is open.

Suggest evidence and Suggest practice open an AI suggestion in Design checks. Read Why this suggestion, then choose Use or Discard. Nothing in the design changes until Use. Discard leaves it as it was.

Draft this task as a page adds an empty page titled with the task (up to 200 characters, or Practice when the task has no wording). When an objective supports that task, the page is linked to it. The AI suggestion for its contents then appears in the gap list, and Why this suggestion is shown before Use. Use fills that page. Discard leaves the page empty.

When the list is long, choose Show all to see the remaining checks. Show fewer returns to the shorter list. This changes only how much of the list is visible; it does not dismiss a check.

Choose Dismiss for this course to hide a check that doesn’t apply. Show dismissed lists hidden checks, and Restore brings one back. A dismissed check doesn’t offer these suggestions. Dismissals are saved with the course and can be undone.

The tabs 1 · Results, 2 · Evidence and 3 · Learning plan show a small step mark: an empty ring when the stage has nothing in it yet, a half ring while it has checks to look at, and a check mark when it is done. The marks are advice too; every tab can always be opened.

Agents get the same checks as designChecks from the alignment endpoint and get_course_alignment, only for courses with the mode on.

Plan with an agent

Agents connected over MCP or the HTTP API can write the same plan with the course.design.update operation and read design checks from the alignment results. See Connect through MCP.

Sources

The stages follow the backward design approach of Wiggins and McTighe (Understanding by Design®, ASCD). The performance preset draws on Cathy Moore’s action mapping, performance analysis by Mager and Pipe, Thomas Gilbert’s Behavior Engineering Model and the evaluation levels of the New World Kirkpatrick® Model. Studio’s questions, labels and examples are our own wording. Understanding by Design® and Kirkpatrick® are registered trademarks of their owners, who don’t endorse Studio.