Backward Design in Instructional Design: Start With the Assessment

Backward design (often written “backwards design”) is a way of planning a course from the end. You start by deciding what learners should be able to do when they finish. Next you decide what evidence would convince you they can do it: the assessment. Only then do you plan the lessons and practice that get them there. Most people plan the other way round, starting with the content they know and adding a quiz at the end, which is how courses end up long, thorough and oddly unhelpful.

The ranking pages on backward design are mostly short glossary entries written for school teachers working from state standards. This guide takes the same three stages and runs them on a self-paced online course, where there’s no teacher in the room to notice confusion. It’s also honest about something most pages skip: how much evidence there is that backward design works.

Where backward design comes from

Grant Wiggins and Jay McTighe gave the approach its name in Understanding by Design, first published in 1998. In the expanded 2005 edition they set out three stages: “identify desired results”, “determine acceptable evidence” and “plan learning experiences and instruction” (Wiggins & McTighe, 2005, pp. 17–18).

They don’t present it as new. They write that the idea is “hardly radical or new”, and credit Ralph Tyler, who “described the logic of backward design clearly and succinctly” in 1949 (p. 20). What they add is a diagnosis of what goes wrong without it. They name two “twin sins” of traditional planning. The first is coverage: marching through a textbook or a set of lecture notes “in a valiant attempt to traverse all the factual material within a prescribed time” (p. 16). The second is the activity that’s fun but aimless.

Both sins are common in online courses. Coverage looks like a module for every chapter of the expert’s knowledge. Aimless activity looks like a drag-and-drop game or a discussion prompt that keeps learners busy but practises nothing they’ll be assessed on or need to do afterwards.

What the evidence says

The research shows the logic is sound; the direct evidence is thin. It’s worth separating the two.

The logic. Backward design is one version of what educational researchers call alignment: objectives, teaching activities and assessment all pointing at the same performance. Describing a unit planned this way, John Biggs writes that “all components in the system became aligned to the objectives” (Biggs, 1996, p. 360). Biggs also explains why the assessment matters so much: it shapes what students actually do to prepare. If the assessment asks only for low-level recall, he argues, “the system will be driven by backwash from testing, not by the curriculum” (p. 350). Designing the assessment first is a way to make that backwash work for you rather than against you.

The direct evidence. Studies that test backward design itself are mostly case reports and single-course comparisons. A typical example is a nonmajors’ biology course at a two-year technical college in the US, redesigned with backward design (Long et al., 2020). Comparing one spring semester before the redesign (102 students graded) with one after (99 students), the share of students who received a D, an F or withdrew fell from 51.0% to 35.4%. That’s an encouraging result, but it comes from one college, two semesters and no control group, and the instructors also moved lecture time towards discussion and activities, so the redesign can’t be separated from other changes.

So, calibrated: there is some evidence, mostly from small, uncontrolled studies, that backward-designed courses produce better results, and good reason from the alignment literature to expect them to. We’d use it on every course, but not because a trial has proved it.

How to design an online course backwards

To keep it concrete, we’ll follow one invented but typical example: a short self-paced course that teaches independent café owners to cost a menu item and set its price.

Step 1: Decide what learners must be able to do

Stage 1 is about priorities. Wiggins and McTighe stress that it “calls for clarity about priorities” (p. 18), because there’s always more you could teach than learners need. Write the result as something you could watch a learner do, not something they “understand”. Our guide to writing learning objectives covers the wording in detail.

For the café course, the expert could teach food cost theory, supplier negotiation, menu psychology and accounting software. The analysis says owners mostly guess their prices. So the desired result is narrow: given a recipe and current supplier prices, cost one plate and set a menu price that hits a target margin. Everything else is either supporting knowledge or a different course.

What it should produce: two to five observable outcomes for the whole course, and a short list of what you’re deliberately leaving out.

Step 2: Decide what evidence will show they can do it

This is the stage most course creators skip, and the one that gives the method its name.

The question is: what would a learner have to produce for you to be confident they can do this for real? For the café course, a multiple-choice quiz on food cost formulas would be weak evidence. Strong evidence is the task itself: a costing worksheet for a dish from the learner’s own menu, with a price and a margin.

Evidence isn’t only the final task. Wiggins and McTighe list “traditional quizzes and tests, performance tasks and projects, observations and dialogues, as well as students’ self-assessments” (p. 19). Plan the small checks too: a quick quiz after the lesson on yield (how much of a raw ingredient survives trimming and cooking), and a practice costing with a worked answer to compare against. Our library entry on formative assessment covers these checks in more depth.

In a self-paced course the evidence has to work without a teacher watching, which leaves three practical options:

• Auto-marked items for anything with a single right answer (the cost of 180 g of beef at a given price per kilo). Well-written multiple-choice and numeric questions do more than people expect; see our guide to writing multiple-choice questions.
• A rubric with model answers for tasks learners check themselves. Show a completed costing worksheet and a short checklist of what a correct one includes.
• A submission someone reviews, if you can staff it: you, a tutor or a peer. This is the strongest evidence and the most expensive. Keep it for the final task.

What it should produce: the final task, written out in full, plus a list of the smaller checks and where each one sits in the course.

Step 3: Plan only the lessons the evidence needs

Now plan the teaching. Work backwards from the final task: what does a learner need to know, and what do they need to practise, to complete it well? For the café course:

1. Reading a recipe as a list of quantities, and converting units.
2. Finding the cost per gram or per millilitre from a supplier invoice.
3. Allowing for yield and waste.
4. Adding up a plate cost and setting a price from a target food cost percentage.
5. A full practice costing, then the final task.

Each lesson ends with practice that looks like a piece of the final task. Each piece of content earns its place by helping with that practice. Material that doesn’t, such as the history of menu engineering, either goes into an optional resource or goes. This is also where cognitive load matters: a lesson that tries to cover conversion, yield and pricing at once will overload a beginner, so the order above adds one new idea at a time. Our entry on cognitive load theory explains why.

What it should produce: a course map listing modules and lessons, with the practice activity in each and the check that follows it.

Most great online learning starts at an outcome. Whether that’s a learning outcome in the higher education space, a skill in the VET sector, or a job skill in the workplace. Once you know what a person needs to achieve to become better at, you can effectively work backwards to ensure there are adequate lessons, activities and reinforcement of the key concepts to build those competencies.

Step 4: Check the alignment on one page

Backward design fails quietly when the three stages drift apart during the build. Before you record anything, put the plan in a single table and read across each row:

Gaps show up fast. An outcome with no evidence won’t be checked. Evidence with no practice tests something learners never tried. Content with no row is a candidate for cutting. If you’ve already mapped your objectives against Bloom’s taxonomy, this table is the same idea applied to the whole course.

Is backward design just teaching to the test?

It’s the most common objection, and it’s fair in one case: when the assessment is a poor proxy for the goal. If the outcome is “cost a plate” and the test is a recall quiz on definitions, then drilling learners for the quiz is teaching to the test, and they’ll leave unable to do the job.

When the assessment is the performance, the objection falls away. Preparing a café owner to cost a dish from their own menu is simply teaching them to cost dishes. This is Biggs’s backwash point turned the right way round: if learners will study whatever the assessment rewards, make sure it rewards the real skill.

Backward design and ADDIE

Backward design isn’t a rival to the ADDIE model. ADDIE lists the kinds of work a course needs, from analysis to evaluation. Backward design is a rule about the order of decisions inside ADDIE’s design phase: outcomes, then assessment, then content. It depends on good analysis too. Step 1 is only as good as what you know about the learners and the gap you’re closing.

Four common mistakes

1. Writing the assessment last “to save time”. It’s the step that decides what the content is for. Leaving it until last is how courses end up with a quiz that tests the easy parts.
2. Choosing evidence your course can’t collect. A rich project with no one to review it gives learners no feedback. In a self-paced course, plan the rubric and the model answer as carefully as the task.
3. Keeping content because it’s good. Good content that serves no outcome still costs learners time and attention. Move it to an optional resources page.
4. Planning backwards once, then building forwards. Recheck the alignment table after the pilot module. That’s where drift shows up.

What to do next

Pick the course you’re planning now, and write its final assessment in full, instructions and checklist included, before you touch the lesson plan. Then build the four-column alignment table from it. If a module doesn’t appear in any row, ask what it’s for.