20,000 practical research leads: method

How 20,000 leads were made, and what they are worth. The leads are composed, not written one by one: 40 practical domains × 25 transport or degradation mechanisms × 20 measurement traps = 20,000. Each is a real claim, because every primitive is a real thing and the sentence names a specific inferential failure. None has been tested. The score ranks practicality and testability, not truth.

1. The composition

A lead is domain × mechanism × trap, phrased as "in D, M is inferred from a bulk proxy, but T is unmeasured; test by X". This construction is deliberate: the failures this programme has documented are almost always of that form — a quantity inferred from a proxy while a specific confounder goes unmeasured.

2. Domains (40)

Domain	Leads
LEDs	500
additive manufacturing	500
adhesives	500
bearings	500
brewing	500
coatings	500
dairy processing	500
desalination	500
electric motors	500
filtration	500
gas sensors	500
humidity sensors	500
lenses	500
lubricants	500
machining	500
packaging	500
paints	500
paper	500
pressure sensors	500
pumps and valves	500
seals and gaskets	500
textiles	500
thermal insulation	500
thermoelectrics	500
water treatment	500
welding	500
batteries	500
building materials	500
concrete	500
displays	500
food preservation	500
fuel cells	500
heat exchangers	500
optical coatings	500
photovoltaics	500
refrigeration	500
steel structures	500
sterilisation	500
supercapacitors	500
drug manufacturing	500

3. Traps (20)

Measurement trap	Leads
a derived variable	1000
a unit convention	1000
baseline subtraction	1000
humidity cross-sensitivity	1000
model freedom	1000
reference drift	1000
sampling bias	1000
saturation	1000
truncation	1000
a non-representative sample	1000
a threshold artefact	1000
ageing during measurement	1000
averaging over heterogeneity	1000
coincident mechanisms	1000
detector dead time	1000
instrument resolution	1000
temperature mismatch	1000
an operator-dependent protocol	1000
an unmeasured active fraction	1000
calibration transfer	1000

4. Scoring

Practicality = 6 − domain cost tier (1 = common industrial, 3 = regulated or clinical). Testability = 6 − test cost (1 = reanalysis or a table, 3 = multi-site or assay). Total 2–10. The ranking favours leads that are cheap to test in a valuable domain.

5. What this is not

Not 20,000 discoveries. It is 20,000 questions, of which perhaps a few hundred are worth asking.
Not independent. The same mechanism appears in 40 domains; a result in one transfers, which is a feature for search and a hazard for counting.
Not tested. See page 12 for the handful that were.

6. Files

Catalogue pages 01–10 hold 2,000 leads each, ordered by score. Page 11 holds the top 50 with testing detail and practical use. Page 12 records the tests actually run. The full set is also Incoprea_2K_all_leads.csv.

Incoprea Research · 20,000-lead project · 3 November 2026. Systematically composed leads, scored for practicality and testability. No experiment performed.