Algorithm Engineering / Interactive Demo

Supplier Matching & Ranking Engine

Hard requirements decide eligibility. Weighted trade-offs decide the order — and the page shows how much each weight actually moves it.

ReactTypeScriptExplainable ScoringSynthetic Data
18 / 24Eligible after hard requirements
25 → 28.3%Cost: weight vs observed influence
34 testsAlgorithm and failure states

24 synthetic supplier profiles · browser-local · nothing sent to a server

A visible weight can still mislead.

A weight of 25 does not mean 25% of the outcome — influence also depends on how much the data actually varies. An opaque score hides that, and hides infeasible choices along with it.

Separate eligibility from preference, then expose how the ranking is formed.

01ValidateSchema and ranges
02FilterHard requirements
03CalibrateP10/P90 anchors
04ScoreBounded utilities
05RankStable tie-breaking
06ExplainContributions and gaps

Excluded suppliers keep every failed requirement, so screening and ranking stay separately auditable.

Show the intended priority and the observed influence together.

Auto calibrationP10 / P90

Good and poor anchors adapt to the scale of each valid dataset.

Boundary buffer25 / 100

A supplier at the initial lead-time or quality threshold does not receive near-zero utility.

Observed influenceWeight × spread

A descriptive diagnostic shows the current ranking leverage—not a causal effect.

Scoring calibrationBalanced
CriterionWeightRangeInfluence
Cost2528.3%
Lead time2016.1%
Quality2015.9%
Reliability2022.7%
Risk1517.0%
18Eligible6Excluded88.6Best score
01Arca Components88.6
02Meridian Manufacturing78.5
03Lumen Precision75.6

Change hard requirements without moving the scoring scale. Compare presets, edit anchors, read every exclusion reason, export the ranking.

Synthetic profiles. No contract terms, volume tiers, or negotiation. Observed influence is a descriptive diagnostic, not a causal effect.