Hypothesis Lab: make a forecast that can lose
Define a learning outcome, expose the assumptions in a forecast, and decide what evidence would change your mind.
Write a claim with an exit
Julia Senta proposes a shorter mission introduction because more learners might reach the notebook task. That is a hypothesis, not a result. Write the mechanism you expect: clearer instructions could reduce confusion before the first action. Then name an alternative explanation, such as prior familiarity with the activity. Choose one observable outcome and a review point before collecting evidence. If every possible result counts as success, the claim cannot teach the crew anything.
Define the funnel before the number
Define a visitor, a start, and a completion so the same event is counted consistently. For this proposed activity, a start means opening the first task; completion means submitting a notebook with a claim and a source. Reopening a task should not create another learner in the same analysis. Record the counting window, exclusions, and missing records. Use the minimum necessary information, avoiding public learner identities. An observed figure needs a source, observation time, and unit; absent data remains absent rather than becoming zero.
Show assumptions without disguising them
DEMO forecast, entirely illustrative: 100 visitors × assumed 20% starts × assumed 40% completion gives 8 completions. None of those values is an observed audience figure. With the same assumed visitor count, start assumptions of 10–30% and completion assumptions of 20–50% produce a scenario range of 2–15 completions. This is an assumption range, not a statistical confidence interval. Rates outside those bounds falsify the chosen rate assumptions for that observation window. A changed audience or task requires a fresh forecast.
Plan the next discriminating observation
A future comparison should keep the task and completion rule stable while varying the introduction. Decide in advance how to handle incomplete records and what evidence is sufficient; a tiny informal trial cannot establish broad efficacy. Look for where confusion occurs, not just which version has a larger total. Keep outcomes that challenge the idea in the notebook. The actual next action is to write a hypothesis card with its mechanism, outcome definition, assumed range, and a specific observation that would force revision.
Try it yourself
DEMO planning exercise: assume 100 visitors and a 20% start rate. Compare completion assumptions of 20%, 40%, and 50%. Acceptance: calculate 4, 8, and 10 completions, label every output illustrative, and name a measured completion rate outside 20–50% as a reason to revise that assumption range.
Check your result
- Completion rates use starters as their denominator.
- A forecast is not evidence that anyone visited or learned.
- The card states a revision trigger before results are available.
This draft stays on your device until you download it. Use the TVIJO discussion for server-stored feedback.