Skip to content

Guide with a worked tree

KPI tree: what it is, how to build one, and a manufacturing example

Short answer

A KPI tree is a diagram that splits one top outcome, such as OEE or on-time delivery, into the smaller measures that add up or multiply to it, down to the ones a single team can change this week. Read it from the top to see where the number went wrong and from the bottom to see who acts.

Last updated:

OEE77.1%

87.5% × 90.0% × 97.9%

  • Availability87.5%
    Stop time 60 min
    • Breakdowns24 min
    • Changeovers28 min
    • Waiting for material8 min
  • Performance90.0%
    Speed loss 42 min
    • Small stops18 min
    • Slow running24 min
  • Quality97.9%
    Rejects 16 parts
    • Startup rejects9 parts
    • In-run rejects7 parts
An example with sample data: one line, one 8 hour shift. Each branch multiplies to the top figure and each leaf adds to its branch.

The idea

What is a KPI tree?

A KPI tree puts one top-level measure at the root and splits it, level by level, into the measures that produce it. Each split is arithmetic: parts that add up, or rates that multiply. The tree stops at leaves that one team owns and can change this week. When the top number moves, the tree shows which branch moved it.

Most plants can tell you OEE or OTIF for last month. Fewer can say, in the same meeting, which part of it fell and who owns that part. The number sits on a board, the causes sit in three different spreadsheets, and the discussion starts from opinions.

A KPI tree puts the arithmetic on one page. Many published KPI tree examples are built on revenue, users and app metrics. The method carries over to a plant without change, but the leaves are different: minutes of changeover, rejected parts, orders short by one line.

A worked example

What is a KPI tree example for a manufacturing plant?

Take OEE for one line. It is availability times performance times quality, so those are the three branches. Each branch splits again into minutes lost or parts rejected: breakdowns, changeovers, small stops, startup rejects. The leaves are numbers an operator, a maintenance lead or a planner can act on. The sample below multiplies out exactly.

OEE77.1%

87.5% × 90.0% × 97.9%

Same figure from the other side: 740 good parts × 0.5 min ÷ 480 min planned

  • Availability87.5%

    Run time ÷ planned time = 420 ÷ 480 min

    Stop time 60 min
    • Breakdowns24 min
      Owner: Maintenance lead
    • Changeovers28 min
      Owner: Line lead
    • Waiting for material8 min
      Owner: Materials planner
  • Performance90.0%

    Ideal cycle × total count ÷ run time = 0.5 × 756 ÷ 420 min

    Speed loss 42 min
    • Small stops18 min
      Owner: Shift supervisor
    • Slow running24 min
      Owner: Process engineer
  • Quality97.9%

    Good count ÷ total count = 740 ÷ 756

    Rejects 16 parts
    • Startup rejects9 parts
      Owner: Line lead
    • In-run rejects7 parts
      Owner: Quality technician
An example with sample data: one line, one 8 hour shift. Each branch multiplies to the top figure and each leaf adds to its branch.
  1. 1

    The top follows from the branches

    Availability, performance and quality multiply to the OEE figure. Cross-check from the other side: good parts times the ideal cycle time, divided by planned time, gives the same number. If your tree does not pass both checks, a definition differs somewhere.

  2. 2

    The leaves are minutes and parts

    Of the 110 minutes of planned time lost, changeovers took the most, 28. Breakdowns and slow running took 24 each. That is where the first conversation goes. The tree does not say why changeovers are long. It says that is the branch to open an action on.

  3. 3

    Each leaf has one owner

    The maintenance lead owns breakdowns, the line lead owns changeovers, the materials planner owns waiting for material. A leaf with two owners gets neither. A leaf with no owner is the end of the tree's usefulness.

  4. 4

    The loss minutes add up too

    Stop time, speed loss and the time value of the rejects (16 parts at 0.5 minutes, so 8 minutes) add to 110 minutes. Planned time less those losses is 370 minutes of good output, which is the same 77.1% of 480.

A second example

What does an OTIF KPI tree example look like?

OTIF is the share of orders that arrive both on time and in full, so build the tree from order counts, not from percentages. Split the orders due into those delivered on time and in full and those that were late, short or both. The late and short counts point to causes a plant can fix.

OTIF, by order92.0%

Orders on time and in full ÷ orders due = 920 ÷ 1,000

92.0% + 8.0% late or short = 100.0%

  • Late or short80 orders

    30 + 40 + 10 = 80 of 1,000 orders due

    8.0% of orders
    • Late only30 orders
      Owner: Production manager, with the carrier plannerLever: Schedule attainment, pickup times
    • Short only40 orders
      Owner: Materials planner, with the packing leadLever: Finished-goods stock, pack accuracy
    • Late and short10 orders
      Owner: Plant managerLever: Fix the shortage first, then the date
An example with sample data: 1,000 orders due in a month, counted by order. The three late or short counts add to 80, and 920 plus 80 is the 1,000 orders due.
  1. 1

    Count orders, then say so

    Counted by order, one missing line fails the whole order. A 40-line order with one line short scores 0% by order and 97.5% by line. Both methods are used. Pick one unit and one tolerance for the on-time window, write them down, and use them at every level of the tree.

  2. 2

    Do not multiply the two rates

    On time was 96.0% of orders and in full was 95.0%. Multiplying gives 91.2%. The real figure is 92.0%, because 10 orders were both late and short and the product counts them as if they were independent. Some definitions multiply the rates, which is why the unit and the rule belong on the page.

  3. 3

    The leaves are different causes

    Late-only orders point at the schedule and the pickup. Short-only orders point at stock and packing. The orders that were both are the ones to look at first, because a plant that fixes the date and ships short has not fixed the order.

Top down

How do you build a KPI tree?

Start with one top outcome and write its definition. Split it with the arithmetic that really produces it, adding or multiplying, and check that the children give back the parent. Keep splitting until you reach a measure one team owns and can change. Then give every node an owner, a data source, a target and a rhythm.

  1. 1

    Pick the one outcome at the top

    Choose a measure the plant already reports and argues about: OEE, OTIF, cost per unit, first-pass yield. Write one definition for it: what is counted, the unit and the period. If two people define it differently, the tree will inherit the argument, so settle it here.

  2. 2

    Split it with the arithmetic that produces it

    Multiply for rates that chain through a process, as OEE is availability times performance times quality. Add for parts of a total, as stop time is breakdowns plus changeovers plus waiting. If you cannot write the formula, you have a guess, not a driver. Draw a dotted line and call it a hypothesis until someone tests it.

  3. 3

    Put real numbers in and check the parent

    Take last month's figures, multiply or add the children and compare with the parent. If they do not match, a definition differs between levels or a child is missing. Do this before anyone else sees the tree.

  4. 4

    Stop at a number one team can change

    Keep splitting until a measure has one owner and a lever that owner can pull this week: minutes of changeover, a count of rejects, a schedule miss. Go further and the tree becomes a list of tasks. The OEE tree here needs three levels.

  5. 5

    Give every node an owner and a rhythm

    For each node write the owner, the data source, how often it is read and the target. Branch targets should multiply or add back to the top target, or the plan contradicts itself. Put each leaf on the board where its owner meets.

  6. 6

    Decide what a red leaf does

    Write the rule next to the leaf: a red opens an action at the next meeting, and the same red twice opens an A3 or a project. Without that rule the tree is a picture of the problem and nothing more.

  7. 7

    Review it on a schedule

    Change the tree when the strategy changes, when a link you assumed turns out to be false, and at a regular checkpoint. Russell and Wille suggest a regular checkpoint, quarterly for example. Cut branches nobody has read in a quarter.

Fill it in

What goes in a KPI tree template?

A KPI tree template is a table with one row per node. Each row names the measure, its parent, the formula that joins it to its children, one owner, the data source and how often it is read, the target, and what happens when it is red. Fill the table in before you draw anything.

KPI tree template with the availability branch of the OEE tree filled in
NodeParentFormula or linkOwnerSource and rhythmTargetWhen it is red
OEENone, it is the rootAvailability × performance × qualityPlant managerStop log and counts, every shift80.0%Open the branch with the biggest loss
AvailabilityOEERun time ÷ planned time (420 ÷ 480)Production managerLine counter and stop log, every shift90.0%Open the biggest stop below
BreakdownsAvailabilityMinutes stopped, adds to stop timeMaintenance leadStop log with a reason code, every shift16 min per shiftAn action at the next tier meeting; an A3 if red twice
ChangeoversAvailabilityMinutes stopped, adds to stop timeLine leadStop log with a reason code, every shift20 min per shiftAn action at the next tier meeting; an A3 if red twice
Waiting for materialAvailabilityMinutes stopped, adds to stop timeMaterials plannerStop log with a reason code, every shift6 min per shiftAn action at the next tier meeting; an A3 if red twice
A template, with the availability branch of the sample OEE tree filled in. An example with sample data: one line, one shift.

Copy the columns into a sheet and fill one row per node, top down. The last column is the one most trees leave out. The OEE calculation guide covers the three factors, and the manufacturing KPI dashboard template is a ready sheet for the boards the leaves go on.

Four pictures

How is a KPI tree different from a scorecard, a bowling chart and a strategy map?

A KPI tree explains why one number moved. A scorecard lists the few measures a team reports against targets. A bowling chart tracks plan against actual for each measure, month by month. A strategy map shows why objectives should cause each other. Use the tree to find causes, and the others to report and to explain.

KPI tree compared with a scorecard, a bowling chart and a strategy map
AspectKPI treeScorecardBowling chartStrategy map
The questionWhy did this number move?How are we doing on the few things we report?Are we on plan for the year, month by month?Why should these objectives produce the result?
The shapeOne root, split by arithmetic down to leavesA short list of measures with targets, often groupedOne row per measure, plan and actual for each monthObjectives in layers, joined by arrows
What joins the itemsFormulas: add, multiplyA grouping such as safety, quality, delivery, cost, peopleNothing; each row stands aloneA belief about cause and effect
What changes itA definition or a driver changesA target changesEvery monthThe strategy changes
Who reads itWhoever has to find the causeAnyone who needs the statusThe monthly reviewLeaders explaining the plan
The four pictures side by side. Plants usually use all four, for different questions.
  1. 1

    Tree and bowling chart share the same measures

    The bowling chart rows are usually the top and branch measures of the trees. The chart says OEE was red in March. The tree says availability was the branch that fell. See the bowling chart entry for the monthly layout.

  2. 2

    Tree and scorecard answer different questions

    A plant scorecard shows the few measures a team reports, such as the ones on a KPI dashboard. The tree is what sits behind one of them. You draw the tree when a measure goes red and the board does not say why.

  3. 3

    Tree and strategy map differ on what a line means

    On a tree a line is a formula. On a strategy map a line is a belief that improving one objective will improve another. Mixing the two on one page is how people end up with trees that look right and add up to nothing. The plan that comes out of a map usually goes on an X-matrix, and the KPIs it names are the roots of the trees.

What goes wrong

Common mistakes when building a KPI tree

  • A tree of every metric

    Forty measures on one page is a list with lines between them. Keep the branches that matter this year. Russell and Wille suggest either focusing on the relevant branches or drawing a higher-level overview tree for the rest.

  • Children that do not add up to the parent

    If the leaves are named but the arithmetic is missing, the parent will not follow them and the tree will mislead. Until a link is tested, draw it dotted and call it a hypothesis.

  • A leaf nobody owns

    A red leaf with no name beside it waits for the next meeting, and the one after. Write one owner per leaf. If two teams share a leaf, split it until each has its own number.

  • Two definitions of one measure

    OTIF counted by order at the top and by line in the plant gives different answers for the same month. Write the unit once and use it at every level.

  • Activities instead of drivers

    Training sessions held is not a driver of OEE. Minutes of changeover is. The leaf should be the thing that moves the parent, not the effort someone made.

  • Built once and never read

    A tree that was right in January is wrong when the line changes. Put a review date on the page and delete branches nobody opens.

Sources

In LeanSuite

KPI trees, boards and goals in LeanSuite

In LeanSuite the tree, the board and the goal are built on the same KPIs.

  1. KPI trees with roll-up

    KPIs are set up as multi-level KPI trees, with roll-up from one level to the next.

  2. KPI boards

    The KPIs are shown on KPI boards, so a leaf can be put in front of the people who own it.

  3. The bowling chart

    The bowling chart shows each KPI's plan against actual, month by month, which is the monthly view of the same measures.

  4. A red KPI opens work

    A red KPI opens an action, an A3 or a project.

  5. Goals measured by a KPI

    A goal for the year is Measured by a KPI with a target and a frequency, and it shows Met or Missed for each month.

See it on your own goals

Try the interactive demo in your browser with sample data, or pilot LeanSuite on your own lines for 90 days, with the setup done by our team.

FAQ

Questions about KPI trees

  1. A KPI tree is a diagram that splits one top-level measure into the smaller measures that produce it. Each split is a formula, either parts that add up or rates that multiply, and the splitting stops at measures one team owns and can change. Read from the top it shows which branch moved the number; read from the bottom it shows who acts.

  2. OEE is the standard one. OEE is availability times performance times quality, so the root has three branches. Availability splits into the minutes lost to breakdowns, changeovers and waiting for material. Performance splits into small stops and slow running. Quality splits into startup and in-run rejects. In the sample on this page the branches are 87.5%, 90.0% and 97.9%, and they multiply to 77.1%.

  3. Many writers use the three names for the same idea: a top measure split into the measures that drive it. Some draw a line between them, usually with the KPI tree for the few measures a business is run on. The useful test is the same whichever name you use: does the parent follow from its children, and does each leaf have an owner?

  4. Pick one top outcome and define it. Split it with the arithmetic that produces it, adding parts of a total and multiplying chained rates. Check with real numbers that the children give back the parent. Stop at measures one team owns and can change this week. Give each node an owner, a source, a rhythm and a target, and write what a red leaf triggers.

  5. Yes, and it is a table rather than a drawing. One row per node, with the parent, the formula, one owner, the source and how often it is read, the target and what a red does. The template on this page has the availability branch of the OEE tree filled in. Draw the diagram after the table is complete, not before.

  6. As many as it takes to reach a measure one team owns and can change this week. The OEE tree on this page has three levels: OEE, its three factors and the losses under them. If a leaf still needs a meeting to decide who owns it, go down one more level. If a leaf is a task, you have gone one too far.

  7. The tree shows how measures relate by formula and answers why a number moved. The bowling chart shows each measure's plan and actual for every month and answers whether the plant is on plan. They use the same measures. A red month on the bowling chart sends you to the tree to find the branch that fell.

  8. A key result is usually a KPI with a finish line, so it is a node on the tree with a target and a date. The tree tells you which node to pick: the branch that fell furthest or the leaf with the most room. The guide on OKR vs KPI sets out how the two work together over a year.

  9. In LeanSuite KPIs are set up as multi-level KPI trees with roll-up, and they are shown on KPI boards. The bowling chart shows each KPI's plan against actual by month, and a red KPI opens an action, an A3 or a project. A goal for the year is measured by a KPI with a target and a frequency and shows Met or Missed for each month.

Want your own KPI tree on the board?

Bring the one measure your plant argues about most, and we will show you on a call how its tree, board and goal sit together in LeanSuite.