Key takeaways
- →Workshop output is raw material, not a result: sixty ideas on a wall commit nobody to anything until a named person has chosen between them.
- →Reserve the last third of the session for converging. If you are still generating at the two-thirds mark, you have already decided not to decide.
- →State the decision rule before the vote, not after: consensus, majority, or one named decider advised by the group.
- →Cut to five candidates before you decide anything. Groups cannot meaningfully choose between twenty options in a room.
- →Every decision needs four fields: what was decided, who owns it, by when, and what would make us revisit it.
- →Send the recap within 48 hours. After that the memory of the session has faded and the record becomes contested rather than shared.
Workshop output is not a decision. A wall of sticky notes, a filled-in board, a shared doc with forty bullet points: all of that is raw material, and the number of workshops that end with material and no decision is close to the number of workshops that get run. The gap is almost always structural rather than a failure of will in the room.
Three things cause it. There was no convergence phase, so the group generated right up to the closing slide. The decision rule was never stated, so the discussion drifted toward a vague agreement nobody is bound by. And the record went out days later in a form nobody could act on.
This piece is about the last third of a session and the two days after it: how to cut a long list down to a real shortlist, how to make an actual call in the room, what to write down, and what to do when the group genuinely cannot decide.
Why output dies between the room and the calendar
The first cause is that generating feels productive and choosing feels risky. A room adding ideas is animated and pleasant; a room choosing between them has to disappoint someone. Given the option, groups will extend the comfortable phase until the clock removes the choice, and then agree to 'take it offline'.
The second is false consensus. Nobody objects at five to five, so the facilitator records agreement, and three people leave privately believing the decision was wrong and unbinding. This is the Abilene paradox in its most common workplace form: a decision everyone in the room privately doubts and nobody contests, which quietly fails to happen over the following fortnight.
The third is that the output is unusable. Photographs of a whiteboard, twenty-eight anonymous stickies in someone's handwriting, a doc with no owners against any line. Even a motivated attendee cannot turn that into work on Monday, and after a week nobody can remember which cluster the argument was actually about.
The fourth is momentum decay. An unfinished decision stays mentally live for a few days and then stops nagging at anyone. If the follow-up lands while the session is still an open loop, it gets acted on; if it lands the following week, it reads as an email about a meeting that has already ended.
The four states of workshop output
It helps to be blunt with a group about which of these you have at any point in the session, because people routinely mistake the second state for the fourth.
Raw material is everything generated: ideas, complaints, options, observations. It has no owner and commits nobody. Clustered material is the same content grouped into themes, which feels like enormous progress and is actually just tidier raw material. A shortlist is three to five named candidates that the group has agreed are the only ones still in play, which is the first state that has cost anybody anything. A decision is one option, one owner, one date and one condition for revisiting it.
Most sessions stop between clustered material and a shortlist, then describe themselves as having decided. When you name the four states out loud at the start of the convergence phase, groups become noticeably more willing to cut, because they can see they are still two steps from finished.
Protect the last third for converging
Decide the convergence block before you design the activities, and treat it as immovable. In a two-hour session that is forty minutes; in a full day it is the last ninety. Whatever the generating phase produces by that point is what you work with.
Announce the switch explicitly. 'We have twenty-two ideas. We are not adding any more; from here we are cutting.' Saying it out loud matters more than it should, because otherwise every convergence exercise gets interrupted by three more ideas and the group never psychologically leaves the divergent phase.
Put a visible countdown on the convergence itself. A timer running on the shortlist exercise changes the quality of discussion immediately: people stop arguing about definitions and start arguing about which option they want, which is the argument you need.
Expect resistance around the cut, and expect it from whoever generated the most. The useful line is that nothing is being deleted, only deprioritised, and the full list goes into the recap. That is true, cheap to promise, and it removes most of the objection.
Cutting sixty items down to five
This is the mechanical part and it takes about twenty-five minutes for a group of ten to fifteen. Run it in order and do not skip the silent step.
De-duplicate in the open, fast
Read the list aloud and merge obvious repeats with the group's agreement. Three minutes, no debate about wording. Sixty items usually becomes forty here without anyone losing anything.
Silent individual selection
Everyone privately picks their top three, with no discussion first. Doing this silently before anyone speaks is what stops the room anchoring on whichever option the most senior person mentioned in the last five minutes.
Reveal the distribution at once
Show all the selections together rather than one at a time. Sequential reveals produce a bandwagon: later voters drift toward whatever is already leading, and you lose the independent judgement you just collected.
Take the top five and stop
Not the top eight because two were close. If there is a genuine tie at the boundary, take six and cut it in the discussion, but do not let the shortlist grow past that.
Give each candidate ninety seconds of challenge
One person argues why it should not be on the list. Ninety seconds each, timed. This is where the weak candidate usually collapses, and where the group finds the objection that would otherwise have surfaced two weeks later.
Apply the decision rule
Whatever you stated at the start: a final vote, or the named decider choosing with the shortlist and the challenges in front of them. Do it in the room, not afterwards.
State the decision rule before the vote
The most common cause of a decision that does not stick is that nobody agreed how it would be made. People participate differently depending on whether they are voting, advising or being consulted, and finding out afterwards that their vote was advisory is how you lose their engagement for the next three sessions.
There are three rules worth using. Majority vote is fast and legitimate when the options are genuinely comparable and the group has the authority. Consent, meaning nobody has a reasoned objection, works well when you need everyone to actively implement the outcome. Decider-advised, where one named person chooses having heard the group, is the honest choice for most workplace decisions and the one facilitators most often avoid stating because it feels less democratic.
Say the rule in one sentence before the shortlist exercise starts, and name the decider if there is one. Groups are entirely comfortable advising a decision-maker; what they resent is being invited to vote on something that had already been decided.
Whichever rule you use, collect the individual positions privately first. A visible early tally biases everything that follows, and the anchoring effect of the first two spoken opinions in a room is strong enough to reshape a whole discussion.
| Decision rule | Use when | Time cost | Failure mode |
|---|---|---|---|
| Majority vote | Comparable options, group has authority | 5–10 min | A large, quiet minority who will not implement it |
| Consent (no reasoned objection) | Everyone must actively carry it out | 15–30 min | One holdout blocks a decision the rest need |
| Decider-advised | Accountability sits with one person | 10–15 min | Feels like theatre if the decider had already chosen |
| Deferred with a date | Genuinely missing information | 3 min | Becomes the default escape from every hard call |
What a decision record must contain
Write it in the room, on screen, while people can see it. A record drafted afterwards from memory always drifts toward what the writer thought was agreed, and by the time anyone notices, the two versions have been acting on different assumptions for a fortnight.
Four fields, per decision. What was decided, in one sentence a person outside the session could understand. Who owns it, which is one named individual rather than a team or a function. By when, which is a date and not 'next quarter'. And what would make us revisit this, which is the field almost everyone skips and the one that stops the decision being reopened informally in six weeks.
Add a fifth line for the discarded options, with one clause each on why they were dropped. It takes two minutes and it prevents the most tedious meeting in corporate life, which is the one where somebody proposes the option you already rejected and nobody can remember the reason.
Read the record back to the room before anyone leaves. Not a summary of the session, just the decisions, out loud. This is where the person who quietly disagreed either speaks or genuinely consents, and it costs three minutes.
Decision
One sentence, plain language, understandable by someone who was not in the room.
Owner
One named person. Not a team, not a job title, not two co-owners.
By when
A date. If the date is more than six weeks out, add a checkpoint date as well.
Revisit trigger
The condition that would legitimately reopen this: new data, a failed pilot, a budget change.
Discarded options
The rejected candidates with one clause each on why, so the debate does not restart in six weeks.
The 48 hours after the session
An unfinished session stays mentally active for a short while and then stops. Send the record inside two days, while people still feel the loop is open, and put the decisions at the top rather than at the bottom of a chronological account of the workshop.
The format that gets read is short: the decisions with their owners and dates first, then the shortlist that did not win, then a link to the full raw output for anyone who wants it. Nobody reads a narrative recap of a session they attended. A tool that captures contributions live helps here, because with Session Flo the ideas, votes and rankings are already recorded in the participants' own words and the recap becomes an editing job rather than a writing job.
Send it to a named list rather than a channel, and tag each owner on their own line. A decision addressed to everyone is a decision addressed to nobody, and that is where most workshop output quietly expires.
Book the check-in before you send the recap. A fifteen-minute review two weeks out, already in the calendar, does more for follow-through than any amount of wording in the record itself.
When the group genuinely cannot decide
Sometimes the honest answer is that the decision cannot be made today, and dressing that up as agreement is worse than admitting it. The test is simple: is the blocker missing information, missing authority, or unresolved disagreement? Each has a different exit.
If it is missing information, name the specific thing you need, who is getting it, and the date the decision happens. That is a deferral with a shape, and it is a legitimate output. 'We need more data' with no owner is not.
If it is missing authority, stop the discussion and say so. Groups will spend forty minutes negotiating something none of them can approve, largely because nobody wants to be the person who points this out. The output becomes a recommendation with the group's reasoning attached, sent to whoever can actually decide.
If it is real disagreement, do not resolve it with a vote at five to five. Capture both positions in the record with the strongest version of each argument, name the smallest experiment that would settle it, and give that experiment an owner and a date. A live disagreement with a test attached is a better outcome than a vote that half the room does not accept.
Checking whether the decisions survived
Top Tips
- Two weeks after the session, ask the owners one question each: has this started, yes or no. Not a status update, a binary. The pattern across sessions tells you whether your decisions are real.
- If more than about a third of decisions have not started, the problem is usually ownership. Look for records where the owner was a team rather than a person.
- Re-read the discarded options list before the next session on the same topic, and open by naming them. It saves the first twenty minutes reliably.
- Track how many decisions each session produces, not how many ideas. A workshop that generated ninety ideas and one decision is worse than one that generated twenty and made four.
- Watch for decisions that keep being remade. That means the revisit trigger was never written down, so people reopen the topic whenever they feel like it rather than when something has actually changed.
- Ask attendees a fortnight later whether they can name what was decided without checking the recap. If they cannot, the convergence phase was too rushed, whatever the record says.
Say the state out loud
Tell the group which of the four states you are in: raw, clustered, shortlisted, decided. Most rooms think clustering counts as deciding.
Never vote before silence
Collect individual choices privately before any discussion of preferences, so the first two spoken opinions do not set the answer.
Write it while they watch
Draft the decision record on screen in the room. Anything written afterwards drifts toward the author's version of the conversation.
One owner, always
If you cannot name a single person, you do not have a decision yet; you have an intention with a team attached.
Frequently asked questions
How much of a workshop should be spent converging?
About a third, and it should be at the end and protected in the plan before you design anything else. Forty minutes of a two-hour session, ninety minutes of a full day. The common mistake is treating convergence as a summary rather than as work: cutting a list of thirty to a shortlist of five, challenging each candidate, and applying a decision rule takes real time and is the part of the session that produces anything durable.
What if the most senior person in the room disagrees with the group's vote?
That is a sign the decision rule was wrong rather than that the vote was wrong. If the senior person holds the accountability, the honest rule is decider-advised, stated at the start: the group shortlists and argues, they choose. Running a majority vote and then overriding it is the fastest way to make people cynical about the next workshop, because it reveals that the participation was decorative.
How do I stop the group from generating more ideas during convergence?
Announce the switch explicitly and give the new ideas somewhere to go. 'We are not adding to the list; anything new goes in the parking area and appears in the recap' takes ten seconds and removes most of the pressure. The reason people keep generating is that adding is comfortable and choosing is not, so the transition needs to be a stated moment rather than something you hope happens.
Should voting be anonymous?
Collect the individual choices privately and reveal the distribution all at once. Private collection stops people converging on whichever option the room appears to favour, and simultaneous reveal stops the bandwagon that sequential voting produces. Whether the individual names are attached afterwards depends on your culture, but the collection itself should always happen before anyone sees anyone else's answer.
What is a reasonable number of decisions to expect from one workshop?
Three to five for a half-day session with a group of ten to fifteen. More than that usually means the decisions were small enough that they did not need a workshop, and fewer usually means the convergence phase got squeezed. Count decisions rather than ideas when you assess whether a session was worth the collective hours it consumed.
How do I follow up without becoming the project manager for everything?
Put the check-in in the calendar before you send the recap, and make it fifteen minutes with one binary question per owner: has this started. You are not tracking the work, you are testing whether the decision was real. If the same owners repeatedly answer no, escalate the pattern rather than the individual items, because it means the workshop is producing decisions the organisation is not resourcing.
Keep reading
- closing a session so the decisions stick — The final ten minutes that turn a shortlist into something people act on.
- running dot voting to reach a group decision — A mechanical way to cut a long list to a shortlist in under fifteen minutes.
- two-phase activities that collect then vote — Separating generation from selection so the room does not do both at once.
- session recaps people actually read — Getting the decision record out inside the 48-hour window.
- capture ideas, votes and rankings with Session Flo — Recording contributions live so the recap is an editing job rather than a writing one.