Use it

Retros Against Humanity

The commonest way a remote retro fails is the round-robin: everyone reads their list out in turn, the last person has nothing left to say, and the same point gets discussed three times. This fixes it by having everybody answer the same question at the same moment — and by never putting a name on an answer at all.

What it is

The shape of Cards Against Humanity, doing an honest job

A prompt is read out. Everyone writes one real retro point that answers it, on a blank card, at the same moment. The answers are read with no names on them, and then the room talks about them until actions fall out. The names never go on.

It is not a party game and there is no scoring in it. There is no judge, no points, and no winner — the borrowed part is the shape: a prompt, an answer from everybody, and everybody playing at the same moment.

Everybody answers at the same moment

This is the part that does the work. You write your card with the prompt in front of you and nobody else's answer in your ear — so what you put down is the true thing the prompt brought to mind, not a version of it quietly edited to match what the person before you just said. The game used to ask for a pile of cards up front, before any prompt was read, and the answer was picked out of the pile. The team that plays it asked for that to go: one card, written at the prompt, is the whole game now.

The cards carry no names, ever

Every answer is read without anybody knowing whose it is, and it stays that way — through the discussion, and in the record that leaves the room. So the room reacts to what was said rather than to who said it, and the uncomfortable point gets its hearing on the same terms as the safe one. Anybody who wants to claim their card can simply say so; the room never does it for them.

Nobody wins

There is nothing to score and no round to lose, which is deliberate. The moment there is a winner, people start playing the funny card instead of the true one — and the true one is the entire reason everybody gave up an hour.

The rounds

Six prompts by default, and three moves in each of them

A round is three moves, and the order of them is the whole design.

A prompt lands on every screen at once and everybody answers it — writes one card there and then, or passes. While that is going the room shows who has answered and nothing else. The facilitator reveals the cards once the room is in, with no names on them, so the room reads the answers without knowing whose they are. The discussion opens when the room has read them, and the actions get captured against the prompt while people are still talking about it. No move in the round puts a name on a card.

There is no writing stage before the first prompt. There used to be — everybody wrote a hand of cards first and played from it, with writing a new one as the way out when nothing in the hand fitted — and the team that plays it asked for the hand to go. The way out is the game.

Six prompts is the default, drawn from one shuffled deck of thirty-three, and the facilitator can change the number in the lobby while the room is still gathering, or draw extra cards later.

There are two packs to draw them from, and the facilitator picks in the lobby alongside the counts. House is Box UK's original deck: warm, daft, and safe to read out in front of anybody. After Hours is thirty-three prompts written for this site — sharper, and pointed at the process rather than at anybody in the room. Both are readable in full from the Play page without opening a room.

Why the pack is settled before the first prompt

The deck is laid out once, when the pack is chosen, and the rounds deal off the front of it — which is what makes "no prompt is read out twice" a property of the data rather than a check somebody has to remember. Swapping mid-retro would deal from both packs into one session and lose track of which prompts had already been used, so the room refuses it rather than merging the two.

Why the names never go on

Reading an answer changes when you already know who wrote it. The room used to hold the names back for one move and then attach them when the discussion opened, on the reasoning that giving them up afterwards is what stops a session being anonymous feedback. Running it with a real team is what changed it: the discussion did not need them. People claim their own card when they want to be asked about it, which is the same outcome by consent — and the ones who do not claim it are the ones the anonymity was for.

A finished round

One prompt, five answers, and what the team did about them

_____ keeps me up at night.
  1. The release script nobody understands

  2. Only one person can approve a deploy

  3. Nobody trusts the tests

  4. Friday afternoon releases

  5. On-call is undefined

  • Pair on the release script on Tuesday and write down what it does — Dev and Brian
  • Second approver for deploys, starting this sprint — Ada to nominate

Five people, five different answers, and not one of them a joke. That is what answering at the same moment buys: nobody heard anybody else's answer before writing their own, so nobody shaped theirs to fit the room.

Now look at what the round actually found. Two of those answers are the same problem — one person is a single point of failure, and a script only one person can read. Neither would have come out of a round-robin: whoever spoke first would have said theirs, and the second, hearing it, would have said “same as that” and moved on. They played at the same moment, so both are on the table.

And “on-call is undefined” is the one that matters most and would have been said last or not at all. It is the sort of thing that sounds like a complaint when you say it out loud with your name attached, and reads like a gap in the process when the room meets it without one. Nobody had to own it to get it discussed — the room agreed it was a real problem while it was still anybody's, and whoever wrote it got agreement rather than a defence to mount.

Two action points out of one prompt, and those DO have names against them — an action nobody owns is not an action. That is a retro doing its job, and it took about six minutes.

What leaves the room

The only durable thing a session produces

Rooms are deleted 48 hours after the last thing anybody does in them, and nothing is stored anywhere else. So the closing screen is not a summary page you can come back to — it is the record, and it has to be taken out of the room while the room still exists.

Every participant gets it, not the facilitator alone. It appears when the facilitator finishes the retro, and closing the room is a second press afterwards — that gap is the room's chance to take it away, and it is worth not rushing. Retros overrun, and a session that simply stops leaves with nothing.

It copies as Markdown, which is chosen for where retro actions actually go: a Jira description, a Slack message and a document all render it, and anywhere that does not still shows something a person can read. The action points are collected at the top, above the session they came out of, because the list at the top is the part that gets pasted into a ticket on Monday.

That is the retro

6 prompts, 5 people. The room is deleted 48 hours after the last thing anybody does in it and nothing is kept anywhere else, so take this with you now.

Action points

  1. Pair on the release script on Tuesday and write down what it does Dev · prompt 3
  2. Second approver for deploys, starting this sprint Ada · prompt 3

The session

Prompt 3

_____ keeps me up at night.

  • The release script nobody understands

  • Only one person can approve a deploy

  • Nobody trusts the tests

  • Friday afternoon releases

  • On-call is undefined

  • Pair on the release script on Tuesday and write down what it does Dev
  • Second approver for deploys, starting this sprint Ada
Prompt 5

The best thing about this sprint was _____.

  • No weekends worked

  • Passed

  • Passed

  • Passed

  • Passed

Retros Against Humanity was created by Box UK (https://www.boxuk.com/) and published as a free agile game under Creative Commons BY-NC-SA (https://creativecommons.org/licenses/by-nc-sa/4.0/). It is a variation of Cards Against Humanity (https://www.cardsagainsthumanity.com/), published under the same licence. This is an unofficial implementation and is not affiliated with either.

A pass is written down as a pass, on screen and in the file — prompt 5 above is one. A prompt four people had nothing to say to is a finding about the sprint, and rendering it as a blank row throws that finding away.

The same two rounds, copied out. This is the whole of the format — a heading, a quote and two lists:

## Action points

1. Pair on the release script on Tuesday and write down what it does — Ada (prompt 3)
2. Second approver for deploys, starting this sprint — Ada (prompt 3)

## The session

### Prompt 3

> _____ keeps me up at night.

- The release script nobody understands
- Only one person can approve a deploy
- Nobody trusts the tests
- Friday afternoon releases
- On-call is undefined

### Prompt 5

> The best thing about this sprint was _____.

- No weekends worked
- _Passed_
- _Passed_
- _Passed_
- _Passed_

The attribution above travels with the text rather than staying on the page. The deck belongs to Box UK and is published under a licence that requires it, so an export pasted into a ticket carries the credit the same way this page does.