Hansei, the reflection that should open every hoshin planning cycle, is the stage most often skipped. Not because people don’t value it, but because the evidence it needs is scattered. Here is what a useful reflection covers and how to run it in a week.
In this guide
- Reflection is not a status report
- Question 1: which bets held?
- Question 2: which targets moved, and why?
- Question 3: what capacity did the plan really use?
- Running it in a week
- A worked example: a reflection summary
- What the reflection needs from the software
- What the missing record costs
- Reflection from the record in Profit.co
- Questions this article answers
Key Takeaways
- →A verdict on every primary correlation, not a color on every objective. Held, partly held, lost, too early, not tested.
- →“Not tested” is a different finding from “lost.” One is about capacity, the other about the idea.
- →Record which targets were renegotiated and why. Repeated downward revisions for one reason is next year’s most useful input.
- →Compare planned effort with real effort. Priorities that took double their allocation are why others were never tested.
- →Five days is enough when the record exists. It is three weeks of archaeology when it doesn’t.
- →Whether reasoning survives the year is decided twelve months earlier, by where catchball and check-ins were recorded.
Reflection is not a status report
Hansei is Japanese for reflection, and in lean practice it carries a sense of honest self-examination rather than review for its own sake. In hoshin kanri it is the first stage of the planning cycle: before deciding what to do next year, look hard at what this year’s plan got right and wrong.
The common substitute is a status report: a slide listing each objective with a color. That shows outcomes but not lessons. An objective can be green because the plan worked, because the market was kind, or because the target was soft. It can be red because the bet was wrong or because execution failed. The color doesn’t say which, and the difference is everything for next year’s plan.
Question 1: which bets held?
The core of the reflection is a verdict on every primary correlation on the matrix: did the priority move the KPI it was meant to move?
| Verdict | Meaning | For next year |
|---|---|---|
| Held | Priority delivered and the KPI moved as expected | Keep the logic; reuse the mechanism |
| Partly held | KPI moved, but less than claimed, or only in some places | Regrade; narrow the claim |
| Lost | Priority delivered, KPI didn’t move; mechanism didn’t hold | Retire or redesign; don’t reword and repeat |
| Too early | Lag window still open | Carry forward with the window date |
| Not tested | Priority wasn’t delivered, so the bet was never placed | Ask why before re-proposing it |
“Not tested” is worth separating from “lost”. A priority that never happened tells you about capacity and commitment, not about whether the idea works. Teams that blur the two either drop good ideas or repeat bad ones.
The diagnosis behind a “lost” verdict, whether adoption, mechanism or external forces, is covered in Outcome lag.
The grades themselves, and what a primary correlation has to state before it earns one, are in Primary, secondary, partial.
Question 2: which targets moved, and why?
List every target that was renegotiated during the year: what it started at, what it ended at, and the reason. Patterns here are revealing. If most revisions were downward for the same reason, such as an assumption about a market or a dependency on another function, next year’s targets should treat that assumption differently.
This question depends entirely on whether the reasoning was recorded at the time. If catchball happened in email and meetings, the reasons may be gone. That’s a finding in itself.
Question 3: what capacity did the plan really use?
For each priority, compare the effort planned with the effort it actually took, even roughly. Priorities that took twice the people they were allocated are the reason other priorities were never tested. Next year’s capacity check should start from these real figures, not from this year’s optimistic ones.
The most useful sentence in a reflection is usually some version of: “We committed to more than we had people for, and here is which work paid for it.”
Running it in a week
- Days 1 to 2: the strategy lead assembles the record: the matrix, monthly results by objective, renegotiated targets, and any catchball history.
- Days 3 to 4: thirty-minute conversations with each primary correlation’s two owners, priority and KPI, to agree a verdict and a one-line reason.
- Day 5: write the summary: verdicts, target revisions, capacity findings and three to five lessons for next year.
- Verdicts, not colors. Every primary correlation has one of the five verdicts.
- Both owners agreed. Each verdict was agreed by the priority owner and the KPI owner.
- Reasons recorded. Every lost or partly held verdict has a one-line reason.
- Lessons named. The summary ends with specific lessons the offsite will act on.
A worked example: a reflection summary
An illustrative reflection summary, written in the fifth day of the reflection week.
| Correlation | Verdict | Reason |
|---|---|---|
| Cost program → opex ratio | Held | Savings landed as planned |
| Pricing redesign → price realization | Held | Took two quarters longer than planned to reach most quotes |
| Service network → services share | Partly held | Worked only where call routing was changed |
| Green sourcing → scope 3 emissions | Partly held | Two of five suppliers converted; others delayed |
| Site maturity → on-time delivery | Too early | Lag window closes in Q2 next year |
| Digital portal → customer NPS | Not tested | Priority deprioritized in May for capacity |
Illustrative.
Targets revised during the year: North America services (down from 34% to 31%, contract model delayed); scope 3 (down, supplier timing).
Capacity: priorities took about 1.5 times their planned effort on average; the service operations team was overloaded from Q2.
Lessons for next year: measure priorities by adoption, not launch; plan with a 1.5× effort multiplier; resolve the service operations overload before lock; don’t re-propose the portal without first explaining why it was dropped.
What the reflection needs from the software
Read the week-one task again: assemble the matrix, monthly results by objective, renegotiated targets and catchball history. None of that is analysis. It is retrieval. Whether the reflection takes five days or three weeks is decided entirely by whether those four things sit in one system or in four.
And the decision was made a year earlier. By the time you need last year’s reasoning, you either recorded it as the year ran or you didn’t. No amount of effort in reflection week recovers a catchball negotiation that happened over email.
| What the reflection asks for | What the tool has to have kept | How to test it in a demo |
|---|---|---|
| A verdict on each correlation | Correlations as records that hold a grade and its changes over the year | Click a dot on a closed year’s matrix and show me its history |
| Why a target moved | The catchball thread that agreed the revision, with its reasoning | Show me a target that changed in Q3 and who agreed it, and why |
| Whether a priority was delivered | Check-in history against plan, not just a final status | Show me the monthly trace for a priority that ended red |
| Real effort consumed | Effort or capacity recorded against each priority | Show me planned versus actual effort for one priority |
| Last year next to this year | The X-matrix viewable across years, not overwritten each cycle | Open the 2025 matrix and the 2026 matrix side by side |
| Which hypotheses were flagged | An exception history, with dates | Show me what was flagged in June and what happened to it |
Six questions. A tool that answers two turns week one into week three.
Row 5 is the one to ask about first. A good deal of strategic planning software treats the matrix as a document for the current year and overwrites it at the next cycle. That is invisible in year one and expensive in year two, which is precisely when nobody is re-evaluating the tool. What changes when the X-matrix is live covers the difference between a drawn matrix and a matrix that keeps a record.
What the missing record costs
Put a number on the gap and it stops being an abstract point about tooling. A strategy lead spending two extra weeks reconstructing the record is the visible cost, and it is the small one. The larger cost is the lesson that never gets drawn: a priority re-proposed next year because nobody could say why it was dropped, or a mechanism retired because the lag window was never written down and the bet looked lost when it was merely early.
That is the context for comparing quotes. The published figures vary in shape as much as in size: OnStrategy lists $200 per month for unlimited users, and i-nexus publishes a free Starter tier for up to five users with $28 per user per month at the Standard tier. Per-seat pricing matters more than it first appears here, because the reflection needs the people who own the KPIs to have been in the system all year, not just the planning team.
Profit.co publishes two plans, Essentials and Enterprise, without a public seat rate; the quote depends on modules and scale. Hoshin Kanri sits in Essentials with OKRs and Goals, Balanced Scorecard and Strategy Roadmap, which matters for this stage specifically: the check-in history the reflection reads is generated by the teams running OKRs through the year, on the same data model as the matrix above them. Keep those in separate tools and question 3, real capacity consumed, has no single place to come from. The full list to total before comparing is in what drives the cost of hoshin kanri software.
The 30-day free trial is enough to load last year’s matrix and run a real reflection on it. That is the stage worth testing before buying, because it is the one you only discover is missing twelve months after the decision.
Reflection from the record in Profit.co
When the year was run in Profit.co, most of days 1 and 2 disappears. Each objective already holds its check-in history against plan, its contributing key results and its matrix correlations. The verdicts are largely written already: every correlation keeps a strength history of the Act decisions made on it during the year, and every countermeasure A3 closes each revision with a verdict on whether its check was met.
The exception queue’s history shows which hypotheses were flagged and when, catchball sessions hold the reasoning behind every target change, and the X-matrix can be viewed across years. See the hoshin kanri planning software page.
What that leaves is the part worth a strategy lead’s week: the conversations in days 3 and 4, where two owners agree what a verdict actually means. The retrieval is done. The judgment isn’t, and shouldn’t be.
Free Product Demo
Run this year’s reflection with us
A 30-minute walkthrough, tailored to your team.
Book a free demo
What happens in the call
How your teams set goals, track progress and review results.
Strategy, performance, projects and engagement on one platform, with AI.
A practical path to align teams, execute faster and drive results.
Questions this article answers
Hansei is Japanese for reflection. In hoshin kanri it is the honest review of last year’s plan that opens the planning cycle, focused on lessons rather than status.
Which bets held, which targets were renegotiated and why, and how much capacity the plan really consumed compared with what was planned.
A lost bet was delivered and didn’t move its KPI. An untested one was never delivered, so it says nothing about whether the idea works, only about capacity or commitment.
One week when the record exists: two days to assemble it, two for owner conversations, one to write the summary. Without a record, assembly alone can take three weeks.
Correlations as records with a grade history, catchball threads holding why targets changed, check-in history against plan, effort consumed per priority, and the X-matrix viewable across years.
Yes. Correlations keep a strength history, catchball sessions hold the reasoning behind target changes, and the X-matrix can be viewed across years, so days 1 and 2 are mostly already done.
