The Insight Forge

Wayfinder Supply Co. · analytical field case

The Case of the Shattered Lanterns

Starlight Lanterns are coming back damaged. Start with the return spike, follow the shipments through warehouses and packaging, then compare carriers. Carry useful findings between the chart rooms to build an explanation you can check.

Synthetic case file.Synthetic Wayfinder Supply Co. shipments; no customer or carrier records are real. Outcomes are complete thirty days after shipment, so the rates shown here are not right-censored.

Current questionWhen did the problem appear?
Full dataset in view9,600 / 9,600 rows
Thirty-day return rate by shipment date
May 15–29 · completed outcomes · baseline 3.7%
static
The author has marked 3 features on this chart: a threshold line labeled "Nonincident baseline 3.7%", a highlighted band labeled "Incident window (proposed)", and a threshold line labeled "Insert B rollout". A line chart of rate by time, split by series. rate ranges from 0.02 (1.7T) to 0.1 (1.7T), with a mean of 0.05 across 90 points. This is an alerting chart; watch for points that break from the rest.
  • The author has marked 3 features on this chart: a threshold line labeled "Nonincident baseline 3.7%", a highlighted band labeled "Incident window (proposed)", and a threshold line labeled "Insert B rollout". A line chart of rate by time, split by series. rate ranges from 0.02 (1.7T) to 0.1 (1.7T), with a mean of 0.05 across 90 points. This is an alerting chart; watch for points that break from the rest.
    • X axis: time, 1.7T to 1.8T.
    • Value axis: rate, 0.02 to 0.1.
    • Series Seven-day return rate: The author has marked 3 features on this chart: a threshold line labeled "Nonincident baseline 3.7%", a highlighted band labeled "Incident window (proposed)", and a threshold line labeled "Insert B rollout". rate ranges from 0.02 (1.7T) to 0.1 (1.7T), with a mean of 0.05 across 90 points.
    • Annotations: 3 marked features.
Skip to data table
The incident rises to 9.4% against a 3.7% nonincident baseline.

Incident band is proposed; collect and test it before accepting a claim.

Current semantic artifact · anomalyReturn-rate spike

May 15–29: 152 returns among 1,617 completed shipments, 9.4% versus a 3.7% baseline.

n = 1,617
Rule-based interpreterRule-based grounding

This is an alerting chart; watch for points that break from the rest.

Reader groundingThe author has marked 3 features on this chart: a threshold line labeled "Nonincident baseline 3.7%", a highlighted band labeled "Incident window (proposed)", and a threshold line labeled "Insert B rollout". A line chart of rate by time, split by series. rate ranges from 0.02 (1.7T) to 0.1 (1.7T), with a mean of 0.05 across 90 points.

First ingredient
Select an item in the satchel
Second ingredient
Select an item in the satchel
Recipe resultNo compatible recipe yet
Compatibility preview

Choose two satchel items. Ingredients are referenced, never consumed.

Anomaly + DenominatorAppraise the Signal
Anomaly + SegmentCut a Cohort
Filter + PathFrame a Hypothesis
Hypothesis + CounterevidenceTest the Attribution
Insight + Saved ViewBind Knowledge to a View
Final lesson
Each room gives you another view of the same case. Collected findings keep their dates, comparison group and shipment counts as you carry them forward.
How the forge is built (optional)

The rooms are ordinary Semiotic charts with scoped evidence attached. The satchel and forge are plain React. The list below is for people wiring the same pattern—skip it if you came for the story.

  • Chart clicks and hovers flow through onObservation.
  • Each room can describe itself and build a screen-reader path (describeChart, buildReaderGrounding).
  • The Appraisal Desk runs validation, diagnosis, and accessibility checks.
  • Saved views round-trip through config helpers so an artifact keeps its chart definition.