From OKR to Hoshin Kanri: what to keep, what to add

Moving from OKRs to Hoshin Kanri does not mean throwing your OKRs away. Your objectives become the annual objectives of an X-Matrix, your key results become its indicators, and the projects you already run become its initiatives. What you add is what OKRs leave open: a written direction over three to five years, an explicit link between each line, a negotiation between levels, and a fixed rhythm of reviews.

This page is for a team that already runs OKRs and feels their limits: objectives that change every quarter with no thread between them, key results nobody can tie to the strategy, reviews that depend on goodwill. It gives the mapping line by line, a migration over one quarter, and the mistakes to avoid. Two neighbouring pages: OKRs in manufacturing if you run a plant, and Balanced Scorecard, OKR or X-Matrix if a scorecard is also in use.

See also: OKR vs Hoshin Kanri: what actually differs, and how to combine them

What OKRs already give you

OKR is a goal-setting method: an objective says what is to be achieved, and key results say how you will know. Google’s guide recommends three to five objectives, about three key results for each, set from a mix of top-down and bottom-up suggestions and graded every quarter. Both of its guides ask that key results describe outcomes, not activities.

All of this carries over. A team that writes good OKRs has already done the hardest part of an X-Matrix: choosing few things, and saying how each will be measured.

What Hoshin Kanri adds

Five things that the OKR guides we read leave to the organization:

  • A direction above the year. A few long-term objectives, over three to five years, sit at the top of the matrix. The OKR guides we read describe quarterly and annual objectives, and nothing beyond.
  • Links you can audit. Each annual objective says which long-term objective it serves, each initiative which objective it carries, each indicator what it proves. An objective nobody works on, and work that serves no objective, both show on the page.
  • A named negotiation. In catchball, each level answers the level above with its own targets and its means before anyone commits. The OKR guides ask for bottom-up input; hoshin makes that exchange a step of the method.
  • Two kinds of management. Hoshin separates the few changes of the year from daily management, which keeps the existing business to standard. Many OKR sets mix the two, and routine work crowds out the change.
  • A review system. Reviews are scheduled at several levels with the matrix as their agenda, and what each review decides is written down.

The mapping, line by line

Nothing is lost in the move. Each thing you have today has a place:

  • Mission or north star → the long-term objectives, at the North of the matrix. If all you have is a mission statement, these objectives are the part to write.
  • Company objectives of the year → annual objectives, at the East.
  • Key results → indicators, at the West, each with a target and a current value. A key result that is really a task ("launch", "ship", "hire") becomes an initiative.
  • Projects, epics and bets → initiatives, at the South, each with an owner.
  • A team’s quarterly OKRs → the quarterly readings of that team’s indicators, read against the target of the year, and the milestones of its initiatives. In a cascade, a unit or a site has its own matrix that takes up lines of the one above.
  • Check-ins → a reading per period on each indicator, and a progress note on each initiative.
  • Grading → it does not carry over as it is. See the next section.

The one thing that does not map: grading

Google’s playbook separates two kinds of OKRs. Committed OKRs are expected to score 1.0, and a lower score calls for an explanation. Aspirational OKRs have an expected average of 0.7, with high variance. The "sweet spot of 60 to 70%" of the re:Work guide reads with this second kind; the pairing is ours, that guide does not make the distinction.

An annual hoshin objective is a commitment. It comes out of catchball, where the level that carries it has said what it can deliver and with what means. Read it as a committed OKR. If your culture values stretch targets, keep the stretch in the long-term objectives and negotiate the annual ones honestly: a target everyone expects to miss by a third teaches teams to discount every target.

A migration over one quarter

A schedule we suggest, not one taken from a source. It keeps your current OKR cycle running while the matrix is built:

  • Week 1: write the direction. A few long-term objectives, with the leadership team. Without them the links have nowhere to start.
  • Week 2: place this year’s OKRs. Objectives to the East, key results to the West, projects to the South. Do not rewrite them yet.
  • Week 3: draw the links and read the gaps. Objectives no initiative carries, initiatives that serve no objective, objectives with no indicator. Each gap is a decision: add, link or stop.
  • Weeks 4 to 6: one round of catchball with the teams, on the annual objectives and their targets.
  • From week 7: hold a first monthly review with the matrix as its agenda, and keep the weekly check-ins as they are.
  • End of the quarter: instead of writing new OKRs from a blank page, review the matrix. What moved, what did not, what changes next quarter.

Mistakes to avoid

The ones we see when OKRs are copied into a matrix without being sorted:

  • Copying forty OKRs. The matrix holds the changes of the year, not everything the company does.
  • Leaving tasks among the indicators. An indicator has a value that moves; a task is done or not done.
  • Skipping the North because "the mission is enough". The links then start from nothing.
  • Linking everything to everything. A link is a claim that this work delivers that objective. Draw the ones you can defend.
  • Dropping the quarterly rhythm. The habit of checking in is an asset, and the reviews build on it.

Doing it with an AI assistant

If your OKRs live in a document, an assistant connected to hoshin.app over MCP can do the placing: you give it the document, it proposes the four lists and the links, and you correct them. The build-x-matrix skill describes that conversation. We have built the matrix of a real company this way, in conversation with an assistant; it is a company of our founder. We have not yet tried it on an export from an OKR tool.

Without an account, fill in the free X-Matrix template and drop it into the checker: it returns a completeness score and the list of gaps. The choices stay yours. The assistant places and points out; it does not decide which objective to drop.

Frequently asked questions

Do we have to stop using OKRs to adopt Hoshin Kanri?

No. Your objectives become the annual objectives of the X-Matrix, your key results its indicators, and your projects its initiatives. Hoshin adds the long-term objectives above them, the links between the lines, catchball and a review system. Teams can keep their quarterly rhythm and their check-ins.

Is a key result the same as a KPI?

Close, not identical. A key result is a measurable outcome tied to one objective for a period. In an X-Matrix it becomes an indicator with a target and a reading per period. A key result worded as a task, such as "launch the new site", becomes an initiative.

What happens to the 70% rule?

It applies to aspirational OKRs, which Google’s playbook expects to average 0.7. The same playbook expects committed OKRs at 1.0. An annual hoshin objective is a commitment negotiated in catchball, so read it as a committed OKR and keep the stretch for the long-term objectives.

How long does the move take?

One quarter for a first matrix and a first review, on the schedule we suggest: one week for the long-term objectives, two to place the OKRs and read the gaps, three for a round of catchball, then the first monthly review. The first full annual cycle is where catchball takes its real place.

What happens to our OKR tool?

Keep it if the teams use it for their check-ins. hoshin.app does not connect to it. An AI assistant can read it through that tool’s own MCP server, if it has one, and carry progress into the matrix.

Sources