Book Demo →

KaizenFlow AI Certified Supervisor — Course

Module 01 · 22 min · 2 fig.

The shift, before it starts

A shift run well is decided before the line moves. This module shows you how to stand up your shift from the ranked queue so you walk the floor knowing where the loss is, what already has an owner, and what is still open.

Read the queue top-down, by dollars

Open the ranked findings queue before your pre-shift walk. KaizenFlow orders findings by verified or estimated dollar impact, not by recency or severity color. Your job is to read top-down and stop where the money stops mattering for one shift. The top three to five findings are your shift agenda; everything below is awareness, not action.

For each finding at the top, note three things: which station and reason code it ties to, whether it is a recurring loss or a one-off, and what the live tiles (OEE, downtime, scrap) say about it right now. If a finding is high-dollar but the station is already running clean this shift, it moves to watch, not work.

  • Sort by impact, read top-down, stop where the dollars stop.
  • Tag each top finding: station, reason code, recurring vs one-off.
  • Cross-check the live tile before committing it to the shift agenda.
FIG. 1.A THE QUEUE, READ TOP-DOWN KF·CS-01
01Changeover — Station 3$4,200/wkRECURRING · NO OWNER02Micro-stops — Station 1$3,100/wkCARRIED · NO CLOSURE NOTE03Scrap spike — Station 7$900/wkOWNED · FIX VERIFIED
Three findings, ordered by dollars — not recency, not severity color. The top two are unowned and recurring: that's your shift agenda. The $900 row is owned and verified, so it drops to awareness.

Separate owned from open

Every finding is in one of three states: owned (someone is on it), open (no owner yet), or carried (handed from the prior shift). Your first accountability move is to make the open set as small as possible before the huddle, not during it.

Check the handoff notes from the outgoing supervisor against the queue. A carried finding with no closure note is still open — treat it that way. Do not assume a finding moved just because a shift passed. You are standing up the shift to know exactly what is unowned the moment the crew arrives, so the huddle assigns rather than discovers.

FIG. 1.B OWNED, OPEN, CARRIED KF·CS-01
OWNERCLOSURE NOTEPRE-HUDDLEMOVEStn 3 · changeoverassign nowStn 1 · micro-stopstreat as openStn 7 · scrapwatchin placepartialmissing
The same three findings, checked against the handoff. Station 1 was 'handed over' — but with no closure note it is still open, and it goes on the assign list, not the trust list.
Key takeaway

Walk in already knowing your top three by dollars, who owns them, and what is still open — the huddle assigns work, it should not discover it.

Module quiz · question 1 of 3

You arrive 20 minutes before crew. The queue shows a $4,200/wk recurring changeover loss at Station 3 with no owner, a $900/wk scrap spike at Station 7 owned with the fix verified last shift, and a $3,100/wk micro-stop pattern at Station 1 carried from two shifts ago with no closure note.

Q1What do you do before the huddle?

Module quiz · question 2 of 3

Q2How does the KaizenFlow findings queue rank items by default?

Module quiz · question 3 of 3

The $3,100/wk Station 1 micro-stop pattern tops your queue, but the live tiles show Station 1 running clean since the shift began — three hours, no micro-stops.

Q3Where does that finding go for this shift?

Module 02 · 24 min · 2 fig.

Turning an alert into an action

An alert is not an action until it has an owner and a dollar-ranked reason. This module is the triage discipline that turns a raised finding into assigned work the crew understands and accepts.

Triage before you assign

When the AI raises a finding, resist the reflex to immediately hand it to whoever is nearest. Triage first: is this real and actionable this shift, is it the highest-value thing that person could be doing, and does the floor actually control the fix? A high-dollar finding rooted in a tooling defect is not a floor action — it is an escalation (see KF·CS-05).

KaizenFlow gives you the next-best-action recommendation alongside each finding. Treat it as a strong default, not a command. You own the call because you know who is free, who is capable, and what else is in flight.

FIG. 2.A ALERT TO ACTION KF·CS-02
Finding raisedAI + $ ESTIMATE01Real this shift?TRIAGE GATE 102Highest-value use?GATE 2 · VS ROUTINE WORK03Floor owns fix?GATE 3 · ELSE ESCALATE04Assign with why$ + REASON CODE05Log owner + targetAUDITABLE CLOSE-OUT06
Three triage gates sit between the alert and the crew — real this shift, worth the labor, floor-fixable. A finding the floor can't fix exits to escalation before it ever burns an operator's shift.

Assign with a dollar-ranked why

An assignment that lands well always carries its reason. Pair the owner with the finding's impact and the specific reason code, so the operator hears "this changeover pattern is costing about $4,200 a week and you're the one who can shorten it" rather than "go fix Station 3."

The dollar figure does two jobs: it justifies why this jumps the line over the operator's routine work, and it sets the bar for what "done" looks like. Capture the owner and a target in the finding itself so the close-out is verifiable, not a verbal promise that evaporates at shift change.

  • Owner + impact + reason code is the minimum viable assignment.
  • The dollar figure justifies the interruption and defines success.
  • Log the owner and target in the finding so close-out is auditable.
FIG. 2.B THE ASSIGNMENT THAT LANDS KF·CS-02
"GO FIX STATION 3"No buy-inno reason · no bar for done"$4,200/WK — SHORTEN IT"Buy-inowner + target logged in thefindingSame findingONE EARNS THE INTERRUPTION
Same finding, two handoffs. The dollar figure and reason code justify why this jumps the operator's routine work — and set the bar for what 'done' looks like.
Worked example — pricing the changeover loss

Station 3's changeover pattern runs about 12 minutes over standard, roughly 25 times a week. The line's loaded cost of lost run time is $14 a minute. That's where the queue's figure comes from:

Loss = overrun × frequency × $/min = 12 min × 25/wk × $14/min = $4,200/wk

That figure is the assignment's why — and its bar. If the fix holds the overrun to 4 minutes, the ledger should verify about $2,800/wk recovered (8 min × 25 × $14). Write the owner and that target into the finding: 'done' becomes a number the ledger can check, not a verbal promise that evaporates at shift change.

Key takeaway

No assignment without a dollar-ranked why — the impact figure is what earns the operator's buy-in and defines when the finding is actually closed.

Module quiz · question 1 of 3

Mid-shift, the AI raises a new finding: recurring micro-stops on Station 4 estimated at $2,600/wk, reason code 'material misfeed.' The next-best-action panel suggests your most experienced operator — who is currently running the verified $900/wk scrap fix on Station 7.

Q1What do you do?

Module quiz · question 2 of 3

Q2When you're deciding who gets a finding, what is the next-best-action recommendation?

Module quiz · question 3 of 3

You assigned the Station 3 changeover loss verbally at the line — 'you've got it, aim for a shorter setup' — and captured nothing in the finding. At shift change the operator says it went fine.

Q3What did skipping the logged owner and target cost you?

Module 03 · 22 min · 2 fig.

Coaching with the data, not at the crew

The live view is a coaching instrument or a surveillance camera depending entirely on how you use it. This module keeps it the former: coach with the data toward the loss, never at the person.

Point at the loss, not the operator

When a tile shows a problem, the coaching move is to put the data between you and the operator and look at it together. "Station 2's micro-stops climbed after the 10am changeover — walk me through what changed" invites diagnosis. "Your numbers are down" invites defense. Same tile, opposite outcome.

The live view's value is that it makes the loss visible without making it personal. Anchor every conversation to a reason code or a downtime pattern, not to an individual's output ranking. The crew will read your screen-glances; if they only ever see the board when something is wrong, it becomes a threat. Use it to recognize good runs too.

FIG. 3.A SAME TILE, OPPOSITE OUTCOME KF·CS-03
"YOUR NUMBERS ARE DOWN"Defenseaimed at the person"WALK ME THROUGH 10AM"Diagnosisaimed at the loss · same dataSame tileTWO CONVERSATIONS
Put the data between you and the operator and the conversation starts at the loss. Point the same tile at the person and it starts at defense.

Coach the pattern, in private, with a next step

One bad data point is noise; a pattern is a coaching opportunity. Wait for the trend before you act, then have the conversation off the floor when it concerns one person. Bring the specific finding, ask what they're seeing, and leave with one concrete next step the operator agreed to — not a reprimand.

Never broadcast individual performance comparisons in the huddle or on shared displays. Aggregate and process losses are shared; named underperformance is private. Crossing that line is the fastest way to make the whole platform read as policing, and once the floor games or hides reason codes, your data — and every finding built on it — degrades.

  • Wait for a pattern; one data point is noise.
  • Individual coaching happens off the floor, one next step at a time.
  • Shared screens show process losses, never personal leaderboards.
FIG. 3.B NOISE, THEN A PATTERN KF·CS-03
NORMAL RANGEPATTERN, NOT NOISEONE-OFF · NOISESHIFTMICRO-STOPS
One bad point stays inside the story of a normal shift. Three shifts of climb after material changes is a pattern — that's when the off-the-floor conversation happens.
Key takeaway

Coach with the data toward the loss, never at the person — the moment the board feels like surveillance, the crew games the reason codes and your findings rot.

Module quiz · question 1 of 3

The live view shows one operator's station with rising micro-stops over the last three shifts, clustered right after material changes. The rest of the line is steady.

Q1How do you address it?

Module quiz · question 2 of 3

Q2What belongs on the shared huddle display, and what stays private?

Module quiz · question 3 of 3

Station 2's tile shows one rough hour of micro-stops this morning. The operator has run clean for months, and the rest of the shift has been normal.

Q3What's the right coaching move?

Module 04 · 20 min · 2 fig.

The daily huddle, on one board

A good huddle is ten minutes, one board, and zero surprises. This module gives you the structure to run a stand-up driven by ranked losses and yesterday's closed loops — not a status meeting that sprawls.

One board, three moves

Run the entire huddle off the ranked-loss board. No slides, no separate spreadsheet, no roundtable of "how's everyone doing." The discipline of one board is what keeps you at ten minutes and keeps the crew anchored to dollars.

Three moves, in order: close yesterday (which loops did we close, what did they recover), name today's top three open losses by dollar impact, and confirm an owner and a next step for each. That is the whole meeting. If something needs a deeper conversation, name it, assign it, and take it offline — the huddle's job is to align, not to solve.

  • Closed loops first — it shows the board works and earns the crew's attention.
  • Today's top three open losses, by dollars, with owners named live.
  • Anything that needs solving gets taken offline, not worked in the huddle.
FIG. 4.A THREE MOVES, EVERY DAY KF·CS-04
Close yesterdayWHAT IT RECOVEREDTop three lossesBY DOLLAR IMPACTOwner + next stepNAMED LIVE10 MIN · ONEBOARD
The whole meeting: close yesterday, name today's top three by dollars, leave with owners. Anything that needs solving gets named, assigned, and taken offline.

Yesterday's closed loops earn today's credibility

Always open with what got closed and what it recovered, pulled straight from the verified-savings ledger. This is not a victory lap — it is the proof that the board reflects reality, which is what makes the crew trust the next set of assignments.

When a loop closed but the savings aren't verified yet, say so plainly: "fix is in, ledger confirms next cycle." Honesty about pending verification protects your credibility more than claiming wins early. A crew that has seen you over-claim once will discount every number on the board afterward.

FIG. 4.B THE CREDIBILITY LOOP KF·CS-04
Loop closesFIX IN PLACE01Ledger verifiesOR SAY 'PENDING'02Huddle opens withitPROOF, NOT VICTORY LAP03Crew trusts theboardASSIGNMENTS LAND04CREDIBILITY COMPOUNDS
Opening with verified closed loops is not a victory lap — it's the proof the board reflects reality, and it's what makes the crew accept the next set of assignments.
Key takeaway

Ten minutes, one board, three moves — open with verified closed loops, name today's top three by dollars, leave with owners; solve nothing in the room.

Module quiz · question 1 of 3

Crew is at the board. Yesterday you closed a changeover loop (recovery verified in the ledger) and a scrap loop (fix in, savings pending next cycle). Today's queue tops out with two recurring losses, both unowned.

Q1How do you run the ten minutes?

Module quiz · question 2 of 3

Q2A finding raised in the huddle clearly needs deeper root-cause work. What happens to it?

Module quiz · question 3 of 3

Yesterday's scrap fix went in and the line looks better, but the verified-savings ledger hasn't confirmed the recovery yet. You're about to open the huddle and want the win on the board.

Q3How do you present it?

Module 05 · 18 min · 2 fig.

Escalating what the floor can’t fix

Some losses the floor simply cannot fix — they recur because the root cause lives in tooling, design, or process owned elsewhere. This module is knowing when to stop re-assigning and start escalating, with the evidence already attached.

Recognize a floor-unfixable loss

The signal to escalate is recurrence despite competent floor action. If a finding has been assigned, worked, and closed by capable operators and still returns shift after shift, the root cause is not on your floor. Re-assigning it a fourth time is not leadership — it is burning crew time on a problem they don't control.

Watch for the tells: the same reason code on the same station across multiple shifts and owners, anomaly alerts firing on a predictable cadence, or a failure-forecast flag that keeps coming true. When the pattern outlives the people you've thrown at it, route it up.

FIG. 5.A THE PATTERN OUTLIVES THE PEOPLE KF·CS-05
ASSIGNEDWORKED +CLOSEDCAME BACKShift 1 · Op AShift 4 · Op BShift 7 · Op CShift 9 · your moveescalate——happenedpartialnot yet
Three capable owners, three competent closes, same reason code back within days. When recurrence survives good floor work, the root cause is not on your floor — the fourth move is up, not sideways.

Escalate with the evidence attached

A clean escalation is one engineering or CI can act on without re-investigating. KaizenFlow lets you route a finding up with its full chain intact: the recurring loss history, the reason codes, the dollar impact from the ledger, the anomaly or forecast data, and what the floor already tried. Send that, not a verbal "Station 3 keeps breaking."

The dollar impact is what gets your escalation prioritized against everything else in engineering's queue — the same logic you use on the floor applies upstream. Attach the verified or estimated recurring cost and the count of how many shifts and owners have touched it. Evidence plus dollars moves a finding; a complaint does not.

  • Route the finding itself, with its history and reason codes intact.
  • Lead with recurring dollar impact — it sets priority in engineering's queue.
  • Include what the floor already tried so no one re-investigates from zero.
FIG. 5.B THE ESCALATION PACKAGE KF·CS-05
Finding, routed upLoss history3 shifts3 ownersReason codesbearing wearDollar impact$3,800/wkrecurringForecast datafired true ×3Floor alreadytriedfixes, outcomes
Everything engineering needs to act without re-investigating, routed up with the finding itself. The dollar line is what sets its priority in their queue.
Key takeaway

When a competently-worked loss keeps returning, the root cause isn't on your floor — escalate the finding with its history, reason codes, and dollar impact attached, not a verbal complaint.

Module quiz · question 1 of 3

A Station 5 bearing-wear downtime loss, $3,800/wk, has been assigned and closed by three capable operators over two weeks and keeps returning on a predictable cadence. The failure-forecast flag has fired correctly each time. A fourth operator is free.

Q1What do you do?

Module quiz · question 2 of 3

Q2What makes an escalation something engineering can act on without re-investigating?

Module quiz · question 3 of 3

Your escalation lands in an engineering queue that already holds a dozen requests from other lines and shifts.

Q3What gets yours prioritized?

Course complete.

You’ve worked all 5 modules. Sit the free practice exam to see if you’re ready for the real assessment — same format, instant grading, keyed back to the modules.

Take the practice exam → Get certified