People / Strategy / Scale
How to run an annual company retrospective before you plan next year
Three questions, one honest hour per team, and a one-page Keep / Fix / Stop that makes your strategy offsite better. Free toolkit attached.
Every Q4 I watch the same thing happen. Leadership books the offsite, opens a blank slide called "Priorities for next year", and starts with the market and the growth target. The year that just happened barely gets a mention.
That's a shame, because your teams already know which handoffs broke, which decision had no owner, and which "top priority" quietly lost to whatever was loudest that week. A retrospective gets that into the room before you plan anything.
It's step one of the annual strategy process I run with scaling companies as a Fractional Chief People Officer, and the step most often skipped. Below: how to run it, whether to use a survey or a live call, and how to turn the answers into decisions.
Free download: Annual Retrospective Toolkit
Invitation message, facilitator script, question bank, AI prompts and the Keep / Fix / Stop canvas.
Download the PDF PDF, 10 pages
Why look back before you plan forward
Strategy built on opinion tends to repeat last year's mistakes with nicer slides. A retrospective gives you evidence from the people closest to the work: what helped, what got in the way, and what everyone called important without ever making time for it. It also tells the team that their year counts as an input.
The method in four steps
- Run it team by team. Use the weekly meeting you already have and block 60 minutes. Aim for groups of 6 to 10 people. Merge teams of two or three, and split anything above about 12.
- Ask three questions. What went well this year? What could have gone better? What should we start doing, or do differently, next year? With eight to ten people these alone fill the hour. Push for examples. "Process issues" gives you nothing to fix; "handoffs from Product to Engineering were unclear when requirements changed after planning" does.
- Capture raw notes. Don't summarize live. Use a note-taker or recording tool and keep people's own wording. Tag as you go: [Worked], [Didn't work], [Change], [Example], [Disagreement]. If you record or use AI transcription, get people's consent, follow your company's privacy rules, and review sensitive notes before they go into an AI tool.
- Synthesize, then check. Let AI clean each team's notes and pull themes across teams, then have a human check the result against the raw notes. Tell the AI not to treat the most repeated point as the most important, and not to invent causes the team never mentioned. A rarely raised issue can still be the expensive one.
When the team lead shouldn't be in the room
Sometimes the leader's presence changes the answers: people rarely challenge them, they tend to steer the discussion, or part of the issue is how they work. Then run the session skip-level. The direct leader doesn't join, and the consolidated themes are shared afterwards. The toolkit has a note telling them it isn't a review of them.
You don't need this for every team. It's a judgment call about trust and dynamics, not a sign that something is wrong with management.
Survey or call? The honest pros and cons
Both work, with different material.
| Live call | Survey (Tally) | |
|---|---|---|
| Pros |
|
|
| Cons |
|
|
My default: a call for teams of about ten or fewer with reasonable trust, because the follow-up question is where the good material lives. Reach for the survey when the team is spread out, trust is shaky, or as a warm-up.
Setting it up in Tally
Tally has a free plan with unlimited forms and responses, and its built-in AI builds a form from a plain-language prompt. Paste in the instructions below and you can have a working form in about a minute. Then check it says what you meant.
- Name the form something human, like "Looking back at 2026". In the intro, say why you're asking and that answers will be synthesized, not quoted back.
- Add the three questions as long-answer fields, each with the line "A specific example helps most."
- Don't ask for name or email. Add an optional team dropdown. If a team is only a handful of people, say their answers might be recognizable instead of promising anonymity you can't keep.
- Set a closing date about a week out and send the invitation (the toolkit has the copy).
- Collect the answers in a Google Sheet or Notion through Tally's integrations, or export a CSV. Summarize per team with AI, then do the human check.
Already using an HR platform like Taito? You may not need another tool. Taito offers anonymous surveys with results by theme and automated follow-ups, so check whether you can run the retrospective there. Keep the same three open-ended questions: you want examples, not another set of scores. Three good questions beat forty that nobody wants to answer.
Running a live retrospective
The biggest advantage of a live conversation is that you can go beyond the first answers. Someone says collaboration between teams didn't work well. But what does that actually mean? Were responsibilities unclear? Did decisions take too long? Were priorities constantly changing? That's where the interesting part starts.
Give everyone space to contribute, ask for specific examples, and don't rush to resolve disagreements. Different people can experience the same situation very differently, and understanding why is often worth more than a quick consensus.
Most importantly, resist the urge to solve problems as they come up. The purpose is to understand what happened and what needs attention. Deciding what to do about it comes later, as part of the strategy process. The toolkit has a facilitator script, a 60-minute agenda and prompts for tense moments.
What you end up with
One page with three headings: Keep, Fix, Stop, plus what next year's strategy has to address and two questions I'd never skip. What did we keep saying was important but never prioritize? And if we change nothing, what is most likely to hurt us over the next 12 months?
Say three teams tell you priorities kept changing. It's tempting to write "better communication" on next year's list and move on. But maybe nobody owns prioritization. Or leadership keeps adding projects without stopping anything. Or the company has committed to more than it has the capacity to deliver.
Those are different problems and they need different decisions: an owner, a limit on concurrent initiatives, a stop list. The retrospective won't make those decisions for you, but it makes them hard to ignore when leadership sits down to plan. That page goes straight into your offsite.
Take the toolkit with you
Not sure where your company stands right now? The Scaling Snapshot is a free first step. And if you'd like a hand running the retrospective or the strategy process behind it, see how a Fractional CPO works or get in touch.