A conjoint rescue is often misunderstood.

It is not a re-analysis, a modeling tweak, or a set of post-hoc adjustments designed to “make the numbers work.” Those interpretations persist because they are common in blogs and software documentation, not because they describe what rescue actually is.

At its core, a conjoint rescue is an independent determination of whether a completed study is capable of supporting the decision it was commissioned to inform.

What a Conjoint Rescue Is Not

A rescue is not:

  • Respondent cleaning
  • Utility smoothing or re-estimation
  • Constraint application to force plausibility
  • Re-running the same analysis with different settings
  • Validation theater after conclusions are already decided

These activities assume the original study was fundamentally decision-capable. In many rescues, that assumption is precisely what must be examined.

What a Conjoint Rescue Actually Involves

A rescue focuses on judgment, not mechanics.

It involves:

  • Reconstructing the original decision the study was meant to support
  • Auditing what the experimental design could and could not identify
  • Stress-testing the evidence under decision-relevant scenarios
  • Determining whether gaps are analytic, structural, or unrecoverable
  • Making explicit which conclusions are defensible — and which are not

The outcome of a rescue is not always a corrected result. In some cases, it is a determination that the study cannot safely support the decision at all.

Why Rescue Exists — and Why It Is Rare

Rescue is analytic governance applied late, when the decision clock is already running.

In most situations, better options exist earlier:

  • conducting a post-mortem after the decision to prevent recurrence, or
  • restarting with analytic governance in place from the outset.

Rescue only makes sense when:

  • fieldwork is complete,
  • timelines cannot move, and
  • the cost of being wrong exceeds the cost of intervention.

That combination is uncommon — which is why true rescue work is rare.

When a study fails silently, rescue is often the only way to determine whether decisions should proceed, be revised, or be stopped altogether.