Phases model

The seven phases a quiz session runs through, and which of them belong to us.

Decision: ⚪ idea

Context

A session is not one screen repeated a hundred times. It is a fixed sequence of phases, each with a single job, each able to be skipped, measured and dropped on its own.

Phase	Job	Can be skipped
Category select	Pick the topics that matter, so a long quiz can be cut to them	Optional, and only if the quiz defines categories
Questions	Collect the answers - see answer model	No, this is the quiz
Checkpoints	Pay the taker back mid-quiz - see checkpoints	Turned off from any card
Demographics	The four fields - see demographics	Yes
E-mail capture	Save the result and ask for consent - see results saving and marketing	Yes
Results calculation	Compute the result while the wait sells the product - see engaging loader	No
Short results	The first payoff, before the full result - see short results card	No

How the sequence behaves:

• The order is fixed, the membership is not - category select appears only when a quiz has categories, checkpoints stop appearing once a taker turns them off, and the rest are always present.
• Checkpoints are the recurring phase - they interrupt questions repeatedly rather than following them once.
• Everything asked for comes after the answers - by the time we ask for demographics or an e-mail, the data we actually need is already collected, so a taker who leaves costs us the extras and not the result.
• The end announces itself - the header switches to "almost done" for the closing phases, which is pacing doing its job - see progress and pacing.
• The last two phases are the results module's - calculation and short results are owned there and merely happen while the taker is still inside the questionnaire flow. The phase list is the contract between the two.

Opportunity

• One job per phase - drop-off can be attributed to a specific phase instead of to "the quiz", which is the only way to know what to fix.
• The data is safe early - putting every ask after the questions means abandoning at demographics still leaves a complete set of answers.
• Category select rescues long quizzes - a taker who would never finish a hundred questions can finish a subset of them.
• Staged payoff - a short result before the full one gives the taker something immediately and gives the full screen a reason to be opened.
• Shared phases, one implementation - the questionnaire does not rebuild what the results module already owns.

Risk

• The closing wall - demographics, then an e-mail, then a wait, all stacked at the exact moment the taker wants their result.
• A decision before the quiz starts - category select sits at the highest drop-off point in the funnel and asks for thought before anything has been earned.
• Skippable data is missing for a reason - the people who skip demographics are not a random sample of takers, and the gap is invisible in the dataset.
• The loader is a deliberate delay - if the computation is fast, the wait is theatre; if it is slow, the theatre is the only thing keeping people there.
• A shared contract cuts both ways - changing the phase list means changing something the results module depends on.