Is Mad, Sad, Glad too touchy-feely for an engineering team?

It depends on the team and the moment, not on the discipline. Engineering teams that have just lived through a stressful release benefit as much as any other team. The reverse is also true: a healthy team that has just shipped well doesn't need Mad, Sad, Glad and would rightly find it forced. The format doesn't care about your job title, it cares about whether you have something to surface.

What if someone shares something that needs HR or 1:1 follow-up?

Name it in the room, briefly, and take the rest offline. 'That sounds like something we should talk about properly after this. Can we set time for it?' Don't try to resolve a serious issue inside a group retro. The format wasn't built for it, and the rest of the team will spend the meeting waiting for the floor to settle.

How often should we run it?

Once a quarter is the most common cadence. Every sprint is too often. The format loses its weight, and the team starts performing the expected feelings instead of surfacing real ones, which is worse than not running it at all.

Should the manager attend?

Usually yes, but the manager should commit to listening and not speaking until the discussion phase is well underway. The manager attending changes what the team will say. The manager skipping signals the meeting is unimportant. Pick attending, then sit on your hands.

What if nobody has anything in the Mad column?

Believe them. If the team is genuinely fine, Mad, Sad, Glad isn't the format you need this sprint. Switch to 4Ls or KALM. An empty Mad column is data, not a facilitation failure, and pushing for content people don't have is how you end up performing a fight that wasn't there.

The toolkit · Mad, Sad, Glad

Mad, Sad, Glad retrospective

An emotional check-in retro. Use it sparingly, when the team needs to talk about how the sprint felt and not just what shipped.

30 prompts · 3 columns · free

😡

Mad

What frustrated you this sprint?

10 prompts in this bucket

😔

Sad

What disappointed you this sprint?

10 prompts in this bucket

😊

Glad

What made you happy this sprint?

10 prompts in this bucket

Try another framework

Frameworks that pair well with Mad, Sad, Glad.

4Ls

Liked, Learned, Lacked, Longed for. More texture than 3-column.

Rose, Bud, Thorn

What's blooming, what's growing, what's prickly. Garden metaphor, real insight.

Glad, Sad, Mad, Wondering

Mad/Sad/Glad with a curiosity dimension. Surfaces uncertainty, not just emotion.

Or see all 10 retro frameworks.

About the Mad, Sad, Glad retrospective

Mad, Sad, Glad is the retro for the sprint where the work is fine but something is off. Tension between two teammates that everyone can feel and nobody will name. A decision that landed badly. A pattern the team has been quietly tolerating because addressing it sounds like extra work. The format puts feelings on the table on purpose: what the data can't see, the team has to say. It's a high-trust format and it takes a confident facilitator. Run it when the team has been together long enough to be honest, when there's an unnamed thing in the room, or after an incident where the formal post-mortem covered the facts and missed everything that actually mattered. Use it sparingly. Run Mad, Sad, Glad every sprint and the team will start performing the expected emotions instead of surfacing real ones. Once a quarter, or after a specific event, is the right cadence.

When Mad, Sad, Glad is the right call

After conflict. After a difficult sprint the numbers will say went fine. After a layoff or re-org. When a teammate has been unusually quiet. When the same Mad-shaped item keeps appearing as a Lacked in your 4Ls. Run it when you suspect there's a layer the team isn't talking about, and when you have the trust to surface it. If trust isn't there yet, build it with a few rounds of Rose, Bud, Thorn first. Trying to run Mad, Sad, Glad on a brand-new team is like trying to do family therapy on the second date.

Facilitator notes for Mad, Sad, Glad

The biggest mistake is rushing past Mad. The Mad column is where the team needs to feel heard, not solved. Resist the urge to jump to action on each item. Acknowledge, name it, move on. The second mistake is treating Glad as a closing pleasantry, the equivalent of 'thanks team, see you tomorrow'. Spend real time there. The Glad column is where the team rebuilds the ground that Mad and Sad have churned. Names. Specific moments. Not 'I liked the team this sprint' but 'when Maya stayed late on Wednesday to help me ship the migration, that mattered.'

Mad, Sad, Glad versus 4Ls

The 4Ls includes feelings (Liked) but balances them with cognitive content (Learned) and concrete asks (Lacked, Longed for). Mad, Sad, Glad is feelings only, no buffer. Use 4Ls when you want feelings included in a normal retro. Use Mad, Sad, Glad when feelings are the whole point and a normal retro would skip past them at speed.

How to run Mad, Sad, Glad async

Carefully, or not at all. The format depends on facial expressions and tone of voice to land safely. Async, you lose both. If you must run async, do it as written cards in a doc, give people 24 hours to add, and follow up with a synchronous call to discuss. Do not try to run Mad, Sad, Glad in Slack threads. That's how a Mad becomes a war.

Frequently asked

Is Mad, Sad, Glad too touchy-feely for an engineering team?: It depends on the team and the moment, not on the discipline. Engineering teams that have just lived through a stressful release benefit as much as any other team. The reverse is also true: a healthy team that has just shipped well doesn't need Mad, Sad, Glad and would rightly find it forced. The format doesn't care about your job title, it cares about whether you have something to surface.
What if someone shares something that needs HR or 1:1 follow-up?: Name it in the room, briefly, and take the rest offline. 'That sounds like something we should talk about properly after this. Can we set time for it?' Don't try to resolve a serious issue inside a group retro. The format wasn't built for it, and the rest of the team will spend the meeting waiting for the floor to settle.
How often should we run it?: Once a quarter is the most common cadence. Every sprint is too often. The format loses its weight, and the team starts performing the expected feelings instead of surfacing real ones, which is worse than not running it at all.
Should the manager attend?: Usually yes, but the manager should commit to listening and not speaking until the discussion phase is well underway. The manager attending changes what the team will say. The manager skipping signals the meeting is unimportant. Pick attending, then sit on your hands.
What if nobody has anything in the Mad column?: Believe them. If the team is genuinely fine, Mad, Sad, Glad isn't the format you need this sprint. Switch to 4Ls or KALM. An empty Mad column is data, not a facilitation failure, and pushing for content people don't have is how you end up performing a fight that wasn't there.

When not to use it

Newly-formed teams without enough trust yet. The format depends on the team feeling safe enough to name uncomfortable things, and that safety doesn't show up in week two. Run Rose, Bud, Thorn for a few sprints first to build the ground.

How to run a Mad, Sad, Glad retro

1Block 45 minutes. Mad, Sad, Glad needs space to breathe. Don't try to rush it.
2Open with explicit psychological safety language. 'Nothing leaves this room. We're naming feelings, not assigning blame.'
3Show all three prompts. Six minutes of silent writing per column, in the order Mad, Sad, Glad. The order matters.
4Discuss Mad first. Loudest voices last. Acknowledge each item without trying to fix it. The job here is to be heard, not to solve.
5Move to Sad. Slower pace. Some items will overlap with Mad, that's fine. Sit with each one before moving on.
6Close with Glad, deliberately. Spend a third of the meeting here. Names matter. Specific moments matter. Pull the team back to ground.
7Identify one thing the team will change as a result, not three. Mad, Sad, Glad outputs are heavier than other formats and the team can only hold one.