This page is for an agent or production manager who already has several shop-floor metrics available, line efficiency, work-in-progress backlog at a bottleneck stage, and defect rate among them, but has never tested which one actually gives enough warning to act on before a shipment date is at risk. At work, a person watching every metric with equal attention reacts to whichever one moves first, not necessarily the one that actually predicts the outcome that matters. This lesson mainly teaches the decision flow: it shows how to test a candidate metric against an order's own real outcome, rather than assuming any metric that looks concerning is the one worth acting on.
Meridian Sourcing manages Order B, 3,000 pairs of jeans with two washes, on behalf of a European buyer, on Delta Garments' own hybrid line that also carries a forecast add-on quantity. Order B (jeans)'s own 22-week timeline includes a 3-week washing stage that has caused delays before. Karim tracks line efficiency, washing work-in-progress backlog, and defect rate on this order, but has never checked which of the three actually gave the earliest reliable warning the last time a shipment ran late.
You need: no previous lesson from this track, though lessons T26-C03-L01 and L02 supply useful background on dashboards and line efficiency.
By the end of this lesson you can:
Not covered here: building a dashboard around one real question, covered in lesson T26-C03-L01, and calculating a line's own efficiency, covered in lesson T26-C03-L02.
| Term | Meaning |
|---|---|
| Leading indicator | A metric that changes early enough, before an outcome occurs, to give time to act on it. |
| Lagging indicator | A metric that only confirms an outcome after it has already happened, too late to prevent it. |
| Bottleneck stage | The single stage in a production sequence whose own capacity limits how fast the whole order can move, regardless of how fast other stages run. |
| Work-in-progress backlog | The quantity of an order sitting unfinished at a specific stage, waiting to be processed. |
| Warning window | The amount of time between a metric first signalling a problem and the deadline the problem would actually affect. |
| Predictive test | Checking a candidate metric against an order's own past actual outcome, to see whether the metric would have given useful warning. |
| False alarm | A metric signalling a problem that, on later evidence, did not actually affect the order's own final outcome. |
The general rule is to test any candidate metric against an order's own real past outcome before trusting it: a metric only counts as predictive if it moved early enough, ahead of the deadline it is meant to protect, to give time to act, not merely early enough to look concerning in hindsight. A metric that only confirms a delay after it has already happened is a lagging indicator, not a warning.
| What can differ | The shapes you will meet | Why it matters |
|---|---|---|
| Export order with a single fixed bottleneck stage | Order B (jeans)'s own washing stage is a known bottleneck; a backlog building there gives earlier warning than a line-wide efficiency average that blends the bottleneck with faster upstream stages. | A bottleneck stage's own backlog reacts to the exact constraint that actually controls the shipment date; an average can stay healthy-looking while the bottleneck itself falls behind. |
| Hybrid order mixing a confirmed quantity and a forecast add-on | A line running Order B (jeans)'s own confirmed 3,000 pieces alongside a forecast add-on needs its bottleneck backlog tracked by quantity actually committed to a ship date, not blended with forecast stock that has no fixed deadline. | Blending committed and forecast quantities into one backlog figure can mask a genuine risk to the committed pieces behind healthy-looking forecast stock. |
| Domestic reorder with no fixed shipment date | Cedar & Co.'s reorder has no port cut-off, so a metric predicting a stock-out risk, not a shipment-date risk, is the one worth testing for that order. | The outcome a metric needs to predict changes with the order; a metric proven predictive for one outcome is not automatically predictive for a different one. |
| Order managed through an agent reporting to a buyer | Meridian Sourcing needs a metric that gives it enough warning window to inform the European buyer before the buyer's own contingency options close, not just enough warning for Delta Garments' own internal planning. | The warning window a metric needs to provide depends on who has to act on it and how much lead time that person's own next step requires. |
Delta Garments' records show that on the previous order run through this line, the shipment left 4 days late. Meridian Sourcing asks Karim to check, in hindsight, which of the three metrics he tracks, line efficiency, washing work-in-progress backlog, and defect rate, would have given useful warning before that delay became unavoidable.
Washing work-in-progress backlog gave a 9-day warning window, more than four times line efficiency's 2-day window and the only one of the three that gave Meridian Sourcing enough lead time to inform the buyer with real options still open. Line efficiency moved too late to be useful for this specific outcome, and defect rate gave no signal at all, despite all three metrics being tracked on every shift.
A common mistake is Meridian Sourcing, or any agent or manager, watching line efficiency, backlog, and defect rate with equal attention, reacting to whichever one moves first on a given day, without ever having tested which one actually predicted a real past delay early enough to matter. What this costs is real: on Order B (jeans)'s own previous late shipment, defect rate stayed calm the entire time and line efficiency only moved 2 days out, so anyone relying on either of those two metrics alone would have had no useful warning at all, while the one metric that actually mattered, washing work-in-progress backlog, was tracked on the same dashboard the whole time without being trusted as the leading indicator it actually was.
Not every number a factory tracks on its shop floor is equally good at warning someone before a shipment runs late, and the only way to know which one actually is good at it is to check it against a real delay that already happened. A metric that only moves a day or two before the deadline gives almost no time to react, while a metric tied to the exact bottleneck stage that controls the timeline can give far more warning, if anyone bothers to check which one that is instead of assuming every tracked number matters equally. Testing a candidate metric against a genuine past outcome, rather than trusting it because it is on the dashboard, is what turns a list of numbers into an actual early-warning system.
With the predictive test's own results in hand, showing washing work-in-progress backlog as the only metric that gave useful warning last time, Karim still wants to keep sending Meridian Sourcing all three metrics in every weekly update, unranked, arguing that more information is always better for the agent to have.
Should Karim keep reporting all three metrics with equal weight, given what the predictive test just showed? Decide before opening the answer.
No, not with equal weight. The predictive test already showed that washing work-in-progress backlog gave a 9-day warning window on Order B (jeans)'s own last delay, while line efficiency gave only 2 days and defect rate gave none at all.
First, sending all three metrics unranked risks Meridian Sourcing reacting to whichever one moves first, which the test showed is very likely to be line efficiency or defect rate, the two metrics that actually gave the least useful warning last time.
Second, more information is not automatically better if it is not clearly ranked by how predictive it actually is; an agent under time pressure needs to know which single figure to watch most closely, not three figures presented as equally important.
Karim should keep tracking and sharing all three metrics, since defect rate and line efficiency still carry other value, but should clearly flag washing work-in-progress backlog as the leading indicator Meridian Sourcing should watch most closely for this specific outcome, based on what the predictive test actually showed.
Relying on line efficiency alone would have given Meridian Sourcing only 2 days to inform the European buyer before Order B (jeans)'s own shipment left late, against 9 days if the agent had been watching washing work-in-progress backlog instead, a difference of 7 additional days in which contingency options, an airfreight upgrade, a partial shipment, or a revised delivery date, were still genuinely open. The final lesson, T26-C03-L04, brings this decision together with the dashboard and efficiency work from lessons T26-C03-L01 and L02 into one complete worksheet.
Open the course workbook's "Predictive metric test" sheet. Enter each candidate metric, the actual past outcome it is being tested against, and how many days before that outcome each metric first signalled a problem; the sheet ranks the candidates by their own warning window with a live formula.
[Download the Dashboards and shop-floor analytics worksheet](../../templates/T26-C03-dashboards-and-shop-floor-analytics-worksheet.xlsx)
Download the Dashboards and shop-floor analytics worksheet (Excel)
1. On Order B (jeans)'s own last late shipment, line efficiency gave a 2-day warning, washing work-in-progress backlog gave a 9-day warning, and defect rate gave none. Which metric should be treated as the leading indicator for this outcome?
2. A factory tracks five shop-floor metrics on every order but has never tested any of them against a real past late shipment. What is the most likely problem with this approach?
| EN | AR | FR | TR | Meaning |
|---|---|---|---|---|
| Leading indicator | مؤشر مبكر | Indicateur avancé | Öncü gösterge | A metric that changes early enough, before an outcome occurs, to give time to act on it. |
| Lagging indicator | مؤشر متأخر | Indicateur retardé | Gecikmeli gösterge | A metric that only confirms an outcome after it has already happened, too late to prevent it. |
| Bottleneck stage | مرحلة الاختناق | Étape goulot d'étranglement | Darboğaz aşaması | The single stage in a production sequence whose own capacity limits how fast the whole order can move, regardless of how fast other stages run. |
| Work-in-progress backlog | تراكم العمل الجاري | Encours en attente | Yarı mamul birikimi | The quantity of an order sitting unfinished at a specific stage, waiting to be processed. |
| Warning window | نافذة التحذير | Fenêtre d'alerte | Uyarı penceresi | The amount of time between a metric first signalling a problem and the deadline the problem would actually affect. |
| Predictive test | الاختبار التنبؤي | Test prédictif | Öngörü testi | Checking a candidate metric against an order's own past actual outcome, to see whether the metric would have given useful warning. |
| False alarm | إنذار كاذب | Fausse alerte | Yanlış alarm | A metric signalling a problem that, on later evidence, did not actually affect the order's own final outcome. |
You now know how to choose the metric that actually predicts a late shipment. The final lesson brings this course's three lessons together into one working worksheet. Back to the course map