Performance board

A performance board tracks the vital few key performance indicators (KPIs) that matter to a function, in the shape continuous-improvement teams call a bowler chart: one block of three rows per measure, read left to right across the periods of a year. The Target row is what was promised, the Actual row is what happened, and the Gap row is the arithmetic between them. In front of them sits the jump-off point (JOP): the last real measurement before the plan started, so the size of the improvement being asked for is on the page rather than in somebody's memory. It sits on the Actual row, because that is what it is — an actual, not a target anybody set and not a gap. At the end sits a year-to-date (YTD) column rolling up the periods that have closed. Colour lands on the Actual row and nowhere else, because a chart where everything is coloured is a chart where nothing stands out. A period is green if the actual meets or beats target and red if it misses; a period that has closed with no number is blue, and a period that has not happened yet is left white. Those last two are the distinction most boards lose: a blue square says nobody looked, which is a different problem from a miss and gets fixed differently. The Gap row stays uncoloured and literal — a minus sign means below target and nothing more, since on a lower-is-better measure below target is the good news, and the Actual row's colour has already said so. Every red gets a response, not just a color: containment (a quick action that protects the customer or the next process now) and a short note on the board saying what's being done and who owns it. Log it in the abnormality log too (a cause seen first-hand at the gemba — the place the work happens — never speculation) and open a countermeasure with a real owner and a target date. The board also flags a KPI that has missed two or more periods running. Treat that as a prompt to check whether containment is holding, not as a test of whether the miss is a real change or ordinary ups and downs: two points in a row can't tell you that. To judge whether performance is genuinely shifting, plot the KPI on a run chart (its values in time order around a centre line) and look for a sustained pattern over enough points, such as a long run on one side of the median. Sources: David Mann, Creating a Lean Culture (daily management and visual boards); Perla, Provost and Murray, "The run chart: a simple analytical tool for learning from variation in healthcare processes", British Medical Journal (BMJ) Quality & Safety (2011).

KPI matrix

Describe the function and what matters to it — a first-pass set of KPIs gets drafted, each with a target for every period and a jump-off point (JOP) recording the last real measurement before the plan began. Enter actuals yourself as each period actually closes; a real number is never invented, and the Gap row is computed rather than typed so the board can never disagree with itself.

Coverage

Computed for you from your KPIs, once actuals are entered — nothing to enter here.

Abnormality & root cause log

Log one whenever a period comes in red — a cause seen first-hand where the work happens and a 5-why root cause, not a guess.

Countermeasures

Every red gets at least a containment action (fast, protects delivery now) with an owner and a note on what's being done; add a permanent countermeasure (fixes the root cause) when the cause is understood.

Narrative

Drafted from the board actually built.

The example below is a fabrication cell's quality board: a full twelve-month bowler read in April, so eight months are still to come and sit white. First-pass yield has missed two months running, and the first containment closed on time but didn't hold, so a permanent countermeasure is open. Scrap rate had a single red month that got the same treatment — logged, root-caused and countermeasured — before it came back under plan. On-time delivery shows the fourth colour: February is blue because the month closed and the number was never posted, so its year-to-date (YTD) figure averages three months, not four. With only four monthly points, no trend is called; every measure stays on a run chart.

• Letting a red pass with no response because it "looks like a one-off".

  Every red gets containment and a note on what's being done — that's how the board protects the customer while you learn more. Whether a miss is a one-off can't be judged from a single point or two in a row; it needs a run chart with enough history.

• Treating two reds in a row as proof of a real change, or one red followed by a green as proof it was just noise.

  A couple of points can't separate ordinary ups and downs from a genuine shift. Judge trends on a run chart, and don't credit a countermeasure for a recovery until the run chart shows it holding.

• Logging an abnormality's cause from a guess instead of direct gemba observation.

  A root cause written from speculation is usually wrong, and a countermeasure built on the wrong root cause doesn't actually fix anything — go and look before writing the cause down.

• Tracking so many KPIs that none of them get real attention at the review cadence.

  The vital few discipline applies here the same as everywhere else in this catalogue — a shorter list that genuinely gets discussed beats a long one that gets skimmed.