The two default question types, and why they fall flat
Yes/no questions ("Do you think generosity matters?") can be answered in one word, and usually are. Once someone says "yes," the question has technically been answered — there's nowhere left for the discussion to go unless the leader asks a second, better question anyway.
Quiz questions ("What did Paul say in verse 3?") test memory and attention, not understanding. The discussion ends the instant someone recalls the right word — which rewards whoever was paying closest attention, not whoever might have the most to say.
A three-layer framework
Every good discussion question sits at one of three layers, and knowing which layer you're writing for makes the question easier to build on purpose rather than improvise under pressure.
Observation: what does the text actually say — who, what, where, when. Interpretation: what did it mean to its first hearers, and what does it mean now. Application: what does it ask of me, specifically, this week. This is the same three-step shape behind Formation's own inductive method — the layers that make personal study rigorous are the same ones that make a written question land.
Five techniques
1. Write the application question first, then work backward
Decide what you actually want the group to walk away considering, then build the observation and interpretation questions that lead there. Writing in the opposite order — starting from the text and hoping application shows up by the end — is how sessions run out of time before the question that matters most.
2. Replace "what do you think" with "where have you seen this"
"What do you think about forgiveness" invites an opinion with no obligation to explain it. "Where have you seen forgiveness cost someone something real" requires an actual memory — and memories reliably produce better discussion than opinions do.
3. Add "for you" to turn a factual question personal
"What does grace mean" is a definition question with a textbook answer. "What does grace mean for you, in what you're dealing with this year" is the same topic, aimed at an actual person instead of a concept.
4. Prepare one uncomfortable question per session, even if you don't use it
Groups tend to gravitate toward the safest reading of a text on their own. Having one harder question ready — even if you decide mid-session not to ask it — keeps you from defaulting to only the comfortable ones out of habit.
5. Test every question by answering it yourself in one word
"Yes." "Paul." "Forgiveness." If your own honest answer fits in a single word, rewrite the question before the meeting — the group's answer will fit in one word too.
Worked example: the Good Samaritan
A lawyer, testing Jesus, asks a definitional question: who counts as my neighbor (Luke 10:29)? Notice first how Jesus opens the whole exchange a few verses earlier — not with an answer, but with a question that hands the reasoning back to the lawyer himself:
"What is written in the Law?" Jesus replied. "How do you read it?"
Luke 10:26, BSB.
Then, after telling the parable, Jesus never actually answers the lawyer's original question. He replaces it with a different one entirely:
"Which of these three do you think was a neighbor to the man who fell into the hands of robbers?"
Luke 10:36, BSB.
The quiz version: "What happened in this parable?" — answerable by retelling the plot, testing whether anyone was listening rather than asking anything of them.
The version Jesus actually asks: not "who fits the category of neighbor" (a definition, which is exactly what the lawyer wanted — a rule with clear edges) but "which of these three was a neighbor" — a question that forces him to identify with a role in the story rather than sort people into a category. That's the whole technique in one line: a factual "who counts" question converted into a personal "which one are you" question, by someone who never once answered the version he was originally asked.
Practicing this in Formation
The questions already written into Formation's group study guides are a working example of this framework in practice — every one of them was built application-first, then checked against the one-word test before publishing.