RESEARCH DESIGN INTELLIGENCE

Find out if your study can work before you run it.

ForskAI tests whether your planned design can recover the signal you care about, before recruitment, fieldwork, or data collection begins.

The Recovery TestA planned study is stress-tested: a latent signal is hidden in measurement noise across a sample of size 32, and a recovered estimate emerges as the design’s recoverability is evaluated, resolving to a pass, risk, or fail result.latent signalrecovered estimatemeasurement noise · planned N = 32RECOVERABILITY0%FAIL

THE PROBLEM

The failure usually appears after the money is gone.

A study can be well-intentioned, well-funded, and still unable to recover the thing it was built to measure. The model may not identify. The scale may saturate. The planned N may be too small. The analysis sheet may quietly alter the data. ForskAI moves those failures to the design stage, when they are still cheap to fix.

The question
Can this design recover the signal?

ForskAI starts before data collection: with assumptions, instruments, planned analyses, and failure modes still visible.

THE APPROACH

Specify the target. Stress-test the design. Fix what cannot recover.

ForskAI turns a research plan into a recoverability test. It checks whether the planned measurement and analysis can recover the intended signal, then shows what has to change before the study begins.

  1. STEP 01

    Specify

    Define the construct, measurement plan, sample, assumptions, and intended analysis.

    The question is translated into testable design conditions.

  2. STEP 02

    Stress-test

    Simulate whether the planned design can recover the signal under realistic noise, missingness, and ceiling effects.

    Failure is useful here because it becomes a design rule.

  3. STEP 03

    Decide

    Return a PASS / RISK / FAIL result with the assumptions, thresholds, and recommended fixes attached.

    The result is design evidence, not a validity claim.

DESIGN PILOT

The expensive failure is not a negative result.

It is discovering too late that the design could never recover the signal.

Design Pilot

PER STUDY DESIGN

€5k–15k

  • Design specification
  • Recoverability stress test
  • PASS / RISK / FAIL result
  • Assumptions and thresholds
  • Recommended fixes
  • Written design-risk report
Run a design pilot

One planned study, stress-tested before you spend it — a clear pass, risk, or fail result, the assumptions behind it, and the report to back it.

Final price depends on model complexity, measurement plan, number of scenarios, and whether evidence support is included.

OPTIONAL ADD-ON · EVIDENCE PILOT

An optional check on the assumptions behind the design.

A recovery test is only as good as the assumptions you feed it — the effect you expect, the variance you assume, the base rates you take for granted. Evidence Pilot checks what the current literature actually supports for those assumptions, so a design starts from defensible inputs instead of hopeful ones.

The ConvergenceSeven study estimates with wide confidence intervals narrow and pool into a single summary diamond at an effect of 0.18, 95% confidence interval 0.12 to 0.24.

It is an optional add-on, not a literature-review tool. It grounds the inputs to a design test and never decides for you.

Secondary by intent: ForskAI leads with research design. Evidence support is there to strengthen the assumptions behind a Design Pilot, not to compete as a standalone review.

PRICING

From one study to an institution.

  • Design Pilot

    €5k–15k

    One study

  • Lab / group retainer

    €20k–40k / year

    Repeated designs

  • Institution

    €60k–150k / year

    Shared method support

  • Enterprise

    Custom · €150k+

    Pharma, CRO, private deployment

Optional evidence support add-on: typically €2k–8k, priced after scoping.

Institution and enterprise enquiries: [email protected]

THE BOUNDARY

What ForskAI is not.

ForskAI does not validate your study, guarantee results, give clinical advice, or turn a weak construct into a strong one. It tests whether the design you are planning has a reasonable path to recovering what you say you want to measure.

  • A PASS is design evidence, not validity.
  • A FAIL is not a failed project. It is an early warning.
  • Roadmap capabilities must be marked as roadmap.
  • Human statistical and domain review remain part of the process.

Bring us one planned study.

We will show where it can recover, where it is at risk, and what has to change.

Run a design pilot