Methodology3 min read

NULL Should Never Mean Favorable: How Default-Optimistic Screening Corrupts Results

A data-stitching audit revealed that missing phonon data defaulted to 'stable,' missing patent counts defaulted to 'novel,' and missing PFAS flags defaulted to 'clean' across our screening pipeline. Every screen that treats NULL as favorable silently promotes unvalidated candidates. Here's what we found and how we fixed it.

LG
Lattice Graph Research

NULL Should Never Mean Favorable

The Problem

When you screen millions of materials for commercial viability, missing data is the norm, not the exception. The question is: what does your pipeline do when a field is NULL?

A comprehensive data-stitching audit of our warehouse and screening pipeline revealed a systemic pattern: every screen defaulted NULL to the favorable verdict. Missing data didn't mean "unknown" — it meant "good."

The Specific Defects

ScreenNULL InterpretationWhat It Should Mean
Phonon stabilityMissing phonon data = stable (weight 1.5, can reach S-tier)No evidence of stability — flag as "no-data"
Thermal conductivitymeasured_kappa IS NULL passes the κ gateConductivity unknown — cannot pass
PFAS screeningpfas_clean is None → no penaltyPFAS status unknown — flag for review
Patent whitespacepatent_count == 0 (from unreliable name-keyed KG lookup) → full whitespace score 1.0Zero patents found by a broken query — not the same as "no patents exist"
Supply chain scoringScorer exception → neutral 0.5 supply scoreScoring failed — report as error, not neutral
Novelty scoringFormula missing from coverage → "novel"Not in our data — that's a data gap, not novelty

How It Propagated

The NULL-favorable pattern compounds through the pipeline:

  1. A compound with no phonon data is scored as "stable" (weight 1.5)
  2. With no patent hits (because the patent matcher was broken), it scores as "novel"
  3. With no PFAS flag, it's "clean"
  4. The compound climbs to the top of the rankings — entirely on the strength of missing data

This is how candidates get promoted to Tier 1 based on the absence of negative evidence rather than the presence of positive evidence. The pipeline rewards ignorance.

The Additional Data Quality Issues

The audit uncovered compounding defects in the data layer itself:

  • ~50% of formula_reduced values are not formulas — 1.17M numeric IDs, 498k comp_<hash> tokens, 445k El_El stubs, and filesystem path fragments. No consumer filters on parse_ok.
  • The patent-evidence chain is defective at every link. dim_patent.title is actually MAX(claim_text). The patent-composition linker regexes ANY capitalized token pair without an element-symbol whitelist — linking patent 8248032 ("notebook computer charger") to NaS3Ti.
  • Cross-source identity resolution is inconsistent. The same physical material gets different composition_id values across sources, while distinct polymorphs sharing a formula get merged.

The Fix: Three-Way Verdicts

Every screen should emit one of three verdicts:

  • PASS — positive evidence meets the threshold
  • FAIL — evidence contradicts the criterion
  • NO_DATA — insufficient evidence to judge

NO_DATA is not PASS. NO_DATA is not even neutral. NO_DATA means the compound cannot be ranked on this axis, and any ranking that treats it as favorable is lying about its confidence.

We implemented this by adding an evidence_level column to every screening verdict and refusing to promote candidates with more than two NO_DATA verdicts across critical axes.

Apply this methodology to your materials

Lattice Graph automates computational screening so you can focus on what matters — the lab.

Explore the Platform