The post-peak retro that actually changes next year

Why retros do not change anything

The usual pattern: a meeting in mid-January, a list of problems, general agreement that things should be better, and a document that is opened again in November when everyone has forgotten it exists and the season has already started.

Three things cause this, and all three are fixable.

It happens too late. By mid-January the detail has gone. People remember that fulfilment was difficult; they do not remember that the difficulty was a specific courier cut-off on two particular days.

The output is a list of problems. Problems do not survive eleven months. Decisions do.

Nothing connects it to next year. The document exists and nothing causes anyone to read it at the moment it would help.

Run it in the first quiet week

Not in January. In the first week where the volume has dropped and people can think, which for most stores is early to mid December or immediately after the returns wave, depending on the shape of the season.

The detail decays fast. A retro run within two weeks produces specifics; one run six weeks later produces impressions, and impressions are not actionable.

Ninety minutes is enough. Longer produces more discussion and not more findings.

The four questions

1. What did we find out too late? The most valuable question and the one most retros omit. Something that was true at eleven in the morning and reached the right person at four in the afternoon is a routing problem, and routing problems are cheap to fix and highly repeatable.

2. What did we do that was not planned? Work that appeared and had to be absorbed. Some of it was genuinely unpredictable; most of it happens every year and simply is not in the plan.

3. Where did we spend time we did not expect to? Ask everyone individually rather than as a group, because the answers are specific and slightly embarrassing, and people volunteer them less readily in a room.

4. What nearly went wrong? Near misses are more useful than failures, because they are more numerous and nobody is defensive about them. Anything that was caught by luck rather than by process will eventually not be caught.

Turning answers into decisions

Every finding leaves the room as one of three things, and anything that does not is dropped deliberately rather than accidentally.

A change to make now, while it is fresh and nobody is busy. Most valuable, and the category people skip because it feels premature in December.

A change to make before next peak, with a specific month attached. Not "before next year", which is how things arrive undone.

An accepted cost. Some things are not worth fixing, and saying so explicitly stops them being rediscovered and re-discussed every year.

The output should fit on one page. A longer document is one nobody reads, which returns you to the problem you were solving.

Making it survive eleven months

The habit that separates retros that change things from retros that do not.

Put the second category, the changes due before next peak, into whatever you actually use to track work, with dates. Not in the retro document. A finding that lives only in a document depends on somebody remembering to open it, and they will not.

Then create one calendar entry, dated about six weeks before next peak, that says: read the retro. That single reminder is what connects a December conversation to an October decision, and it costs one minute to create.

A retro finding that does not end up in a task with a date is a note about a feeling.

Who should be in the room

Everyone who worked the season, including seasonal staff if any are still around, and especially them.

Temporary staff see the process without the assumptions that come from having built it, and they are about to leave, so they have nothing to lose by saying what was confusing. That combination makes them the most honest source in the room, and their observations expire the day they finish.

If seasonal people have already left, a five-minute exit conversation at the time is the substitute, and it is worth doing routinely for exactly this reason.

Running it so people say something

A retro where everyone agrees things went reasonably well has failed, and it usually fails for structural reasons rather than because the season was fine.

Ask for written answers first. Five minutes in silence before anyone speaks. Without it, the first person to talk sets the tone and the range, and everyone calibrates to them.

Start with the near misses. Nobody is defensive about something that did not actually go wrong, so it warms the room up in a way that starting with failures does not.

Have the most senior person speak last, or ideally not at all until the end. An owner who opens with their view of the season has closed most of the useful conversation before it started.

Ask the quiet people directly. By name, about something specific they worked on. General invitations to contribute reliably produce nothing from the people whose view you most need.

What to do with disagreement

Two people will remember the same event differently, and the instinct is to resolve it. Usually you should not.

A disagreement about what happened is often a disagreement about what people could see, and that is the more useful finding. If fulfilment thought stock was fine and marketing thought it was critical, the interesting question is not who was right but why they had different pictures.

That is almost always an information routing problem, and routing problems are cheap to fix and highly repeatable. Chase the discrepancy rather than the verdict.

The half-page that goes to next year

Whatever the retro produces, compress it before storing it.

Three headings. Changed already, which is the list you can stop thinking about. To do before next peak, each with a month attached and each already in your task tracking. Accepted, the things not worth fixing, recorded so nobody rediscovers and re-debates them in eleven months.

That third heading does more work than it appears to. Without it, the same three issues get raised every year, discussed at length, and quietly dropped again, which wastes an hour annually and slightly erodes faith in the exercise.

If you only ask one question

Ask what people found out too late.

It surfaces routing problems, which are the cheapest category to fix and the most likely to recur. It is unthreatening, because it is about information flow rather than anybody's performance. And it produces specifics rather than impressions, because people remember the moment they learned something they wished they had known earlier.

Everything else in a retro is useful. That one question is the one that reliably changes something.

Doing it alone

Plenty of stores are one or two people, and a retro with yourself sounds pointless. It is not, though the failure mode is different.

Alone, you will skip it because you already know what happened. You know the events; you do not know the pattern, because you were inside it for six weeks. Writing the four answers down is what converts a season you lived through into a season you can see.

Do it in writing rather than by thinking about it, and do it in the same place you will look next year. The whole value is in the record existing eleven months from now, at which point your memory of December will have compressed into something much less useful than a page written at the time.

The habit that compounds

A single retro produces a handful of improvements. Three consecutive years of retros produces something more useful: a record of which problems recur.

Anything appearing in three consecutive years is structural rather than situational, and it has survived two rounds of good intentions. That is the one worth solving properly, even if it costs more than it appeared to when it was a single year's annoyance.

Keeping the pages together, in one place, is what makes that visible. Separate documents in separate folders never get compared.

Common questions