Design it right.Make it ready.Prove its integrity.
Study Manager turns a scientific question into an execution-ready study, guides and monitors its conduct, and keeps the complete path to defensible evidence in one record — so the study runs once, and every reviewer gets an answer.
Prior NIH SBIR Phase II awardee · Mapped to the NIH rigor criteria and the EQIPD Quality System — see the mapping
- A design sized for the comparisons you will actually report
- A recorded, stratified allocation you can reproduce from its seed
- Bench sheets and worklists generated from the approved design
- Problems surfaced while the study can still respond
- An analysis of what was collected, under the designed comparison
- A submission-grade SEND package and ARRIVE-oriented documentation from the record
- The underpowered cohort
- The repeat study, and the months it costs
- “How did you choose n?” — with no answer
- Four littermates in one arm nobody noticed
- The statistician who arrives after the data
- ARRIVE responses reconstructed at publication
The science should not disappear between design and execution.
When design, execution, data and documentation come apart, rigor becomes difficult to demonstrate and problems are discovered too late. Study Manager keeps the reasoning, the plan, the observations, the changes and the evidence connected to the study they belong to.
- Rationale in the investigator’s notes
- Power calculation in a spreadsheet
- Protocol in a document nobody executes from
- Collection forms built apart from the approved design
- Instrument data in separate file formats
- Deviations and changes in email
- Analysis outside the study record
- ARRIVE responses reconstructed after completion
Plan → Design → Operationalize → Conduct → Assess → Document.
The design intelligence is the starting point. The connected scientific record is the platform. Integrity is not a stage — it is what every stage has to preserve, and the documentation is the proof that it did.
- 01Available — shown as steps
Plan
What question must the study answer?
Objective, hypothesis, biological target, primary outcome and the intended decision are defined — and whether the study is exploratory or confirmatory.
In the product today: The Design Assistant asks what the study must establish, reads the objective, resolves a design family and offers alternatives.
- 02Verified
Design
What study can produce the required evidence?
Factors, groups, comparisons, timing, allocation, effect assumptions, sample size and analysis intent are resolved — with every assumption and its source recorded.
In the product today: Conditions, species, outcomes, factors and the Analysis LaunchPad: sizing, power, simulation and scenarios, all design-time.
- 03Available — shown as steps
Operationalize
Is the design ready to be executed rigorously?
The design becomes an executable plan: timeline, protocol-defined events, collection setup, QC rules, capture routes and a printable site protocol.
The integrity gate — nothing proceeds past what is undecided.
In the product today: Timeline & sampling plan, the measurement protocol builder, collection setup from the approved design, the handoff sheet.
- 04Verified
Conduct
Is the approved plan being followed, and are problems visible in time?
Enrollment with stratified allocation, worklists by event, bench and record sheets, barcode identity, file and analyser imports through quarantine, a QC queue, amendments with reasons, and a monitoring sweep.
In the product today: A named module with an honest state line — not set up yet, ready to collect, collection in progress, needs review.
- 05Available — shown as steps
Assess
Did the study produce the evidence it was designed to produce?
Planned, executed and observed evidence compared: did the variance and recruitment assumptions hold; what the designed comparison shows on the collected data; what the next study would need.
The integrity check — did the intent survive execution?
In the product today: The post-collection sizing readout and the engine-computed analysis run, inside the Conduct module today.
- 06Available — shown as steps
Document
Can the study be explained, reviewed, reproduced and delivered?
A SENDIG 3.1.1 submission package validated against the FDA rule set, ARRIVE-oriented readiness and export, CSV exports, and the amendment and approval history — from the record, not reconstructed.
The integrity record — evidence that it was upheld.
In the product today: The Reports page with its readiness cards and the SEND builder; being rewritten around the record.
Stage status follows the capability ledger’s lifecycle table (2026-09-18): “Verified” means a user can open the stage by name; “shown as steps” means the work exists across steps without a stage screen. The stage ribbon in the product is being built.
Know what evidence the study needs before you run it.
Study Manager does not simply calculate sample size. It asks whether the proposed design can produce the evidence the scientific question requires — and it holds the number until the study has said enough. Real engine outputs, inputs stated.
Adjusting for a baseline covariate you already measure.
Two arms, continuous outcome, ρ = 0.70. Three arms: 53 → 28. Verified 2026-09-15.
Sizing for the comparisons the paper will report — each dose vs vehicle under an exact Dunnett adjustment — instead of “do the arms differ at all.”
Four arms, d = 0.8, α 0.05, power 0.80. Verified 2026-09-11.
Weighting a shared control √k : 1, as the literature recommends — and the engine sizes what it recommends.
41 fewer at eight arms. Refused where treatment arms differ among themselves. Verified 2026-09-15.
From a study worth running to a study ready to run.
The approved design becomes the operational framework — timeline, protocol-defined events, capture routes, QC rules, responsibilities and the readiness evidence — before a cohort is committed. The integrity gate.
Shown as a composite assembled from existing surfaces (design readiness, collection setup, handoff sheet). A single readiness review screen is in development.
- Scientific readiness
- Experimental-design readiness
- Timeline and procedure readiness
- Subject, group and treatment readiness
- Measurement and data-collection readiness
- Randomisation and blinding readiness
- Roles and responsibilities
- ARRIVE reporting readiness
- SEND terminology readiness
Design the study once. The collection environment follows.
Data collection that understands the study it is collecting: worklists and bench sheets generated from the approved plan, identity by barcode, files and analyser feeds through quarantine, and problems surfaced while the study can still respond.
Collect
Worklists by event in cage or rack order; a bench sheet across animals or a record sheet per animal; categorical values from the design’s SEND-coded option lists; out-of-range values flagged and confirmed with a reason; corrections as amendments with a controlled reason code.
Scan
Ear tags, cage cards and tube labels through a USB wedge or the device camera (Code 128, Code 39, DataMatrix, QR, EAN-13); a scan station whose tray becomes a bench sheet, a custody event or an enrollment; a phone paired by QR as a desk’s scanner.
Import and connect
CSV and XLSX results and ASTM/HL7 analyser messages into a quarantine, resolved by tube then animal, units converted exactly (UCUM) or held with both units named, then accepted through the same rules as the bench.
Early warning
A monitoring sweep names what needs attention, in study context, with who hears what decided in setup.
- A planned event is at risk or overdue
- An external result is overdue
- A value is outside its expected range and awaits review
- A missed critical event has raised a deviation
- An imported record needs resolution
- A record is incomplete for its event
Notification delivery (email/SMS) is not yet built — the sweep records what it found and the screens show it. Phone-camera scanning requires HTTPS. Verified by execution: 81 tests in the integrated module, 2026-09-19.
Did the study produce the evidence it was designed to produce?
After collection, the platform shows whether the design’s variance and recruitment assumptions held, analyses what was collected under the designed comparison with every number from the engine, and sizes the next study on what was actually seen. The integrity check.
Planned
- Target enrollment and allocation
- Scheduled events and measurements
- Biological target and minimum meaningful difference
- Analysis intention and comparisons
Executed
- Actual enrollment and allocation, by seed
- Completed, missed or changed events
- Documented deviations and amendments
- Audit history and provenance
Observed
- Observed difference, pooled SD, recruited n
- The designed comparison on the collected data
- Assumptions held or missed
- What the next study would need
Illustrative composite of the three-column view; the Observed column is being built. Today the readout and the analysis run live in the Conduct module’s Review workspace. No “observed power” is reported, by decision (Hoenig & Heisey, 2001).
Rigor should not be reconstructed at publication.
The integrity record: documentation generated from the study as it was designed, conducted and amended — not written afterwards.
SEND submission package
SENDIG 3.1.1: SAS Transport datasets for 19 domains, Define-XML 2.1, Dataset-JSON 1.1, an nSDRG, a transformation report tracing every value, an eCTD m4 layout. Terminology from SEND CT, never guessed. Validated against the FDA SEND rule set (Pinnacle 21 Community: 0 rejects, 0 errors).
ARRIVE-oriented readiness and export
Readiness checks for randomisation, blinding and sample size, and design and analysis summaries from the record. Supports ARRIVE reporting; never “ARRIVE compliant.”
The amendment and approval history
Who changed what, when, and why — with a controlled reason — from enrollment through review.
Exports with a declared contract
Enrollment, data plan, collection status and analysis rows as CSV, blinding decided server-side.
SEND caveats travel with the claim: validated with Pinnacle 21 Community; no PC/PP, SUPP--, RELREC or pooled findings; feature-gated for the founding cohort. We say “validated against the FDA rule set,” never “FDA-accepted.”
Thirteen real-world designs, nine areas — and what the design step changes for each.
Seeded from published studies and configured for analysis planning. Each card names the one decision the platform makes differently for that kind of study.
- OncologyPDX tumour growth inhibitionrepeated-measures factorial · tumour volume
Baseline tumour volume as a covariate — the requirement nearly halves (37 → 20 per group at ρ = 0.7).
- OncologyAnti-PD-1 + myeloid agent combinationrepeated-measures factorial · tumour volume
The interaction is sized, not assumed. Baseline-burden stratum and implant cohort declared up front and balanced at enrollment.
- OncologyDrug × radiation xenograftfactorial · tumour volume
Per-term power: main effects and the drug × radiation interaction each get a number.
- Toxicology · IND-enablingLiver safety margin, multi-dose (rat)one-way, multi-dose · ALT
Each dose vs vehicle under an exact Dunnett adjustment; the shared control is weighted. The 28-day tox study is also the SEND demonstration.
- Metabolic · hepatologyNASH fibrosis drug × dietfactorial · fibrosis area
Is response modified by diet? The interaction is the estimand, and it is sized as one.
- Metabolic · hepatologyDiet-induced NASH model selection (rat)one-way, four diets · NAS score
A multi-arm ordinal outcome: the engine asks which contrast the study will report and sizes that.
- NeuroscienceAlzheimer's model, longitudinal cognition + biomarkerrepeated measures · memory trajectory
Simulation-based power for the trajectory, with a reproducibility assessment.
- NeuroscienceEpilepsy model severity assessmentone-way · Irwin screen score
Model characterisation with an ordinal primary outcome, declared exploratory and recorded as such.
- ImmunologyIBD anti-inflammatory (colitis)one-way pilot · histology score
A pilot declared as a pilot. Its data seeds the confirmatory study's expected values and ρ.
- Infectious diseaseInfluenza antiviral challengerepeated measures · viral RNA (log10)
Trajectory sized by simulation; sex-specific arms; litter and arrival weight balanced at enrollment.
- Gene therapy · rare diseaseVector dose × genotype severityrepeated-measures factorial · grip strength
Dose-response trend across dose rank; genotype as a declared factor.
- Behavioural · neuroimmuneMaternal asthma → offspring behaviourfactorial · social interaction
A mechanism study: the 2×2 interaction is the claim, and it is sized as the claim.
- Aging · geroscienceDiet restriction, lifespan body weight (DO mice)repeated measures · body weight
A longitudinal trajectory in genetically diverse animals; variability sources declared.
- Your study
Two-group and multi-group, baseline-adjusted, factorial, cross-over, dose-response, binary and ordinal outcomes, longitudinal, non-inferiority. If the shape is here, the design step is ready.
Engine capabilities, for completeness →
AI that can’t invent a fact.
Study Manager’s assistants draft, rewrite and explain — from a closed list of designs the engine can actually run, using only what you wrote. They never supply a dose, a comparator, an effect size or a number. The numbers come from the engine, and the engine tells you when it won’t.
Planned: an Expert Guidance rail drawn from the rigor literature and your study’s own state, and an optional expert statistician review at the readiness gate.
- Objective composer
Makes your stated objective complete and well-formed, using only what you already wrote, and says which of six slots it touched.
- Design drafter
Turns the objective into a design family the engine can execute — chosen from the list, never free-form.
- Timeline draft
A first draft of the study timeline from the design, for you to edit. It has never produced a timeline the app couldn't render.
- Stat Assistant
The narrative beside the number: what the requirement means, which assumption is doing the work, what would move it.
Of 1,173 animal-research publications from five leading institutions, 1.4% reported how they chose their sample size.
The three practices every rigor framework asks for, and the share of publications that reported them. Denominators differ because a practice is only counted where it would have applied.
Macleod et al., PLOS Biology 2015 (1,173 in-vivo publications from five leading UK institutions). All figures verified against source before publishing.
NIH and EQIPD ask for the same things. The readiness review checks them.
NIH’s rigor criteria and the EQIPD Quality System converge on one list for a study that makes a claim: a pre-specified hypothesis and analysis, a sample size justified before starting, randomisation and blinding, and outcomes traceable to data. Study Manager records each one — and says which it doesn’t yet.
See the requirement-by-requirement mapping →Rigorous experimental design — randomisation, blinding — is a scored review criterion. In 2026 NIH asked the field for tools that help researchers implement and document rigor practices.
Simplified Peer Review Framework; Highlighted Topic 66, April 2026.Eighteen core requirements from 29 founding institutions in eight countries, free to use. For a knowledge claim: hypothesis, analysis and sample size defined and documented before starting.
Bespalov et al., eLife 2021. Study Manager is not EQIPD-certified; the mapping is ours.- Investigators
Design a study that can answer the question and preserve the justification behind it — and answer Reviewer 2 from the record.
For PIs → - CROs
Give sponsors a traceable design, execution plan, study record and evidence package.
For CROs → - Programs & core facilities
Standardise readiness across labs and see which studies are ready, active, at risk or complete.
For programs → - Funders & quality communities
The requirement-by-requirement mapping to NIH rigor criteria and the EQIPD Quality System.
Rigor frameworks →
Help shape the next generation of rigorous preclinical study management.
The statistical engine has a decade behind it, built under NIH SBIR Phase I and Phase II awards. The workspace around it is being rebuilt now, with the people who will use it. Founding partners get early access, direct founder support, documented influence on what gets built, and a free year. We publish what’s verified and what isn’t.