English · Русский · Українська

In 90 seconds

Before We Build currently has no compatibility percentage, rank, formula, or recommended weights. Before measurement could be responsible, the project would need to define the outcome, validate each construct and instrument, test pair mechanisms, quantify uncertainty, replicate results, and establish safety limits.

What “measure compatibility” would mean

Compatibility is always compatibility for an outcome, domain, population, and time horizon. “Can this pair coordinate a three-month project?” and “Will this marriage remain safe and faithful?” are different questions and cannot share an undefined universal score.

Shared example

For relocation, a study could separately examine decision clarity, factual accuracy, burden balance, freely given consent, execution, and later regret. Combining these into one number would hide value choices about which outcome matters and whether a safety failure can be traded against convenience.

Prerequisites

Required evidence includes a clear construct, reliable indicators, valid instruments, appropriate samples, outcome follow-up, rival-model comparison, calibration on data separate from development, uncertainty estimates, subgroup analysis, and external replication.

What may be investigated later

Researchers may compare qualitative profiles, threshold or gate models, additive models, and interaction models. These are study designs, not current BWB algorithms. Any chosen model must follow the meaning of the outcome and evidence rather than theoretical elegance.

Useful alternatives now

Use an evidence table, unresolved-question list, level-specific resources and frictions, safety screening by qualified processes, and a decision log. These preserve important differences without pretending that uncertainty is a precise quantity.

What must not be inferred

Individual typology test totals are not pair compatibility scores. A high average cannot cancel coercion, violence, exploitation, or a non-negotiable conflict. More decimal places do not make an unvalidated judgement scientific.

Questions before any model

  • What outcome is being predicted and who chose it?
  • Which errors are most harmful?
  • Is the measure reliable and comparable across groups?
  • How is uncertainty communicated?
  • Which conditions are gates rather than trade-offs?

For a researcher

The technical design space and rejection criteria are documented in Compatibility Measurement Methods. The staged evidence programme is in Validation Program.

Next reading

Use Evidence Workflow and Relocation Walkthrough for the current non-scoring workflow.