Formative Assessment Examples for Self-Paced Online Courses

Search for formative assessment examples and you’ll find long lists: mini whiteboards, thumbs up or down, exit tickets, think-pair-share, a traffic-light card on every desk. They’re good ideas, and nearly all of them assume the same thing: a teacher in the room who looks at the answers and changes the lesson on the spot.

A self-paced online course doesn’t have that teacher. Nobody sees the thumbs. If a learner gets a check-in question wrong at 10 pm on a train, the decision about what happens next was made months earlier, by you, when you set up the quiz. This entry covers what formative assessment actually is, what the research does and doesn’t support (the famous numbers are shakier than most lists admit), and nine examples built for courses where the “teacher” is the design.

What the research says

Formative is about use, not format

The definition most researchers now work from comes from Paul Black and Dylan Wiliam. In their 2009 paper they wrote that practice in a classroom is formative “to the extent that evidence about student achievement is elicited, interpreted, and used by teachers, learners, or their peers, to make decisions about the next steps in instruction that are likely to be better, or better founded, than the decisions they would have taken in the absence of the evidence that was elicited” (Black & Wiliam, 2009, as quoted in Brookhart & DePascale, 2025).

Two things in that sentence matter for course builders. First, nothing in it is about the type of activity. A quiz, a forum post or a draft essay can all be formative, and none of them is formative by default. What makes it formative is that the evidence gets used to decide what happens next. Second, the people who can use the evidence include learners and their peers, not only teachers. In a self-paced course, that’s most of the opportunity.

The same chapter lists five strategies, attributed to Dylan Wiliam and Marnie Thompson, that make up formative practice: clarifying learning targets and criteria for success; engineering discussions that elicit evidence of learning; providing feedback that helps improve learning; activating students as resources for one another; and activating students as the owners of their own learning (Brookhart & DePascale, 2025). Every example later in this entry maps to one or more of those five.

The famous numbers don’t hold up

You’ll often read that formative assessment produces effect sizes of 0.4 to 0.7, which would make it one of the most powerful things a teacher can do. That range traces back to the popular summary of Black and Wiliam’s 1998 review, not the review itself, according to Derek Briggs and colleagues (Briggs et al., 2012).

When Neal Kingston and Brooke Nash pooled the studies they could find that met their criteria, they estimated an average effect of about 0.20 standard deviations, with considerable variation between studies. Briggs and colleagues then argued that methodological problems made even that figure “somewhat equivocal”, and concluded that “considerable uncertainty remains” about the effect of formative assessment practices on achievement (Briggs et al., 2012).

Randy Bennett’s critical review reached a similar view from a different direction. He argued that the term covers so many different practices that effects should be expected to vary widely from one implementation to the next, and was blunt about the headline figures:

None of this says formative assessment doesn’t work. Bennett notes that research suggests its general practices can facilitate learning (Bennett, 2011). It says the average is modest, the range is wide, and the details of what you actually do decide which end of the range you land at.

The most recent overview: positive, varied, uncertain

A 2024 umbrella review by Andrew Sortwell and colleagues pulled together 13 meta-analyses of formative assessment in schools. Formative assessment produced “trivial to large positive effects on student learning, with no negative effects identified”, and the size of the effect varied with the type of formative assessment. Three of the meta-analyses found small effects for computer-based formative assessment, between 0.21 and 0.34 (Sortwell et al., 2024).

The authors are careful about how far this goes. Most of the meta-analyses were of moderate quality, but the certainty of the evidence was rated very low in nine of the 13 and low in three, which they say “precludes robust recommendations” about the best strategies. The review covers school students, not adults in self-paced courses.

Self-assessment works better with something to check against

Self-assessment, where learners judge their own work, matters more in a self-paced course than in a classroom, because it’s often the only judgement available in the moment. Zi Yan and colleagues pooled 26 studies (98 effect sizes) of self-assessment interventions in a 2023 meta-analysis and found a significant overall effect on performance, g = .455 (Yan et al., 2023). The important finding was what made it work:

No other moderator they tested was significant. In practice, asking learners “how do you think you did?” helps a little. Asking them to compare their work with a model answer, a rubric or someone else’s feedback helps much more.

What the online research adds

The research on formative assessment in online courses is thinner and mostly descriptive. A systematic review by Jane Gikandi, Donna Morrow and Niki Davis considered 91 peer-reviewed articles and focused on 18 key empirical studies in higher education (Gikandi et al., 2011). They identified “ongoing authentic assessment activities and interactive formative feedback” as the important characteristics, and found that asynchronous discussions let students rethink and assess their own understanding.

One of the studies they summarise, a blended course using online self-test quizzes, points to the catch in all of this:

That’s a case study, not an experiment, and stronger students may simply be the ones who read feedback. But it fits the definition: a check only helps if someone uses what it shows. A self-paced course can put the evidence in front of learners; it can’t make them look at it unless the design asks them to.

How strong is the evidence?

Moderate, which is why we’ve rated this entry that way. The core idea, that learning improves when evidence about it is used to adjust what happens next, is well supported and sits underneath the stronger evidence on effective feedback and retrieval practice. The claims for “formative assessment” as a package are much less certain: effects are modest on average, vary widely, and come mostly from schools. We’d treat the specific practices with their own evidence (feedback that explains, self-assessment against a standard, low-stakes practice testing) as the reliable parts, and the label as a useful way of organising them.

How to apply it in a course

The test for any formative check comes straight from the definition: when you write it, can you say what will happen next for each likely result? If the answer is “nothing”, it’s a score, not formative assessment.

In a self-paced course, two people can act on the evidence. The learner acts on it straight away, through the feedback, branching and next steps you’ve written in advance. You act on it later, when you look at the course data and revise a lesson. Good examples serve one or both.

We’ll use one invented example throughout: a self-paced course teaching small-business owners to read a profit and loss statement. It’s the kind of practical skill where learners make predictable mistakes, such as treating revenue as profit or forgetting that their own wages are a cost.

Nine formative assessment examples for self-paced courses

1. A pre-check that routes learners. Before the module on gross margin, three questions test whether learners can tell revenue from profit. Those who get all three skip a five-minute primer; those who don’t are sent to it. The evidence decides where each learner starts. This also stops experienced learners drowning in material they don’t need, a point cognitive load theory makes about novices and experts.

2. A check-in quiz with an explanation for every wrong answer. Three questions at the end of a lesson, each wrong option with two sentences on why it’s wrong and what rule would have got the learner there. This is the most common formative check in online courses, and the explanation is what makes it formative: a green tick or a red cross gives the learner almost nothing to act on. Our entry on effective feedback covers how to write those explanations, and the guide to writing multiple-choice questions covers how to make wrong options reveal specific mistakes.

3. A confidence rating alongside the answer. After each answer, learners mark how sure they were. A confident wrong answer is the most useful thing a quiz can surface, because it shows a misconception rather than a guess. In the profit and loss course, a confident “yes” to “Is your own drawing a business expense?” triggers a short extra explanation and a second question on the same idea.

4. Self-assessment against a model answer. Learners read a simple profit and loss statement and write three sentences on what it tells them. Then they see a model answer and tick which of its four points they covered. This is the kind of self-assessment Yan and colleagues found worked best: judgement against something explicit, not a bare “how did you go?”

5. “Explain why” prompts. Instead of asking for the answer, ask learners to explain why it’s right (“Why did gross margin fall even though sales went up?”), then show an expert’s explanation to compare with. Explaining exposes gaps that recognising the right option hides. Our entry on elaborative interrogation covers the research on why-questions.

6. A draft that gets feedback before the final. The course’s main task is for learners to annotate their own business’s last quarterly statement. They submit a draft of the first section, get comments from a tutor or against a rubric, then finish. Feedback on a draft is formative because there’s still something to change; the same comments on the final submission would be summative.

7. Peer review in a forum. Learners post their annotated statement (with the figures disguised if they prefer) and review two others against the same rubric. Reviewing someone else’s work against criteria is practice in the judgement they need for their own, which is the “students as resources for one another” strategy in action.

8. A one-question reflection at the end of a lesson. “Which part of this lesson are you least sure about?” with four options and a free-text field. Learners who pick an option get a link to the matching section. The answers also tell you, across the whole cohort, which part of the lesson to fix first.

9. Question data you use to revise the course. Every few weeks, look at the per-question results. If most learners miss the same question, the question or the lesson is the problem, not the learners. Rewrite one or the other. This is formative evaluation of the course itself, the evaluation phase of the ADDIE model, and it’s the part of formative assessment where you, not the learner, are the one making the next-step decision.

Make the checks low stakes and the next step obvious

Two design rules run through all nine examples. First, keep formative checks low stakes: several or unlimited retakes, no effect on the certificate, and wording that makes clear the point is to find out what isn’t working yet. Learners won’t show you their misconceptions if every check counts against them. Keep the summative assessment, the one that decides the certificate, separate and clearly labelled.

Second, never leave a check without a next step. Each result needs a planned response: an explanation, a link, a retry with a different question, a branch, or a note to you that the lesson needs work. The research on online courses suggests many learners won’t go looking for feedback, so put the next step in the path rather than off to the side.

Plan these checks when you write your learning objectives, not after the content is built: each objective should have at least one formative check that tells you, and the learner, whether it’s been met.

How the platforms support it

Most formative assessment lives in what you write: the questions, the explanations, the model answers, the next steps. The platform decides how much of that you can build in and what data you get back. We haven’t yet published our reviews of Thinkific, LearnWorlds, Teachable and Kajabi, and assessment features change often, so check these against your platform’s current documentation, on the plan you’re paying for:

• Feedback per answer option. Can you write a different explanation for each wrong answer, or only one message for right and one for wrong?
• Ungraded quizzes and retakes. Can a quiz be set as practice that doesn’t count towards completion or a certificate, with retakes allowed and different questions drawn each time?
• Branching or conditional content. Can a quiz result unlock, skip or recommend a lesson, so a pre-check can route learners?
• Open responses and assignments. Can learners submit written work or files, and can you or a tutor reply with comments and ask for a resubmission?
• Per-question reporting. Can you see, for each question, how many learners chose each option, and export it? Without this, example 9 is guesswork.
• Discussion or peer review. Is there a place for learners to post work and comment on each other’s, inside the course or alongside it?

If your platform can’t do one of these, there’s usually a workaround: a linked form for confidence ratings, a model answer in a downloadable file, a spreadsheet export for question data. The checks matter more than the tool.

Further reading

• Bennett (2011), “Formative assessment: A critical review”. The clearest account of what the term does and doesn’t mean, and why the headline numbers are suspect.
• Briggs et al. (2012). A short commentary on the Kingston and Nash meta-analysis that explains where the 0.4 to 0.7 figures came from and why pooling these studies is hard.
• Yan, Wang, Boud & Lao (2023). The meta-analysis behind the advice to give self-assessment something explicit to check against.
• Gikandi, Morrow & Davis (2011). Older, but still the most cited review of formative assessment in online higher education, with useful detail on discussions, quizzes and e-portfolios.
• Sortwell et al. (2024). The most recent overview of the meta-analyses, with an honest rating of how certain the evidence is. Open access.

For the rest of the principles behind good course design, browse our research library.