Everything between your question and the paper.

You do — all three. You are the data controller; we act as processor under a written agreement. The database structure, the instruments, the analysis scripts and any custom software we write for you are yours, delivered as source with documentation.

We keep no rights over your data and no claim on your results. Where we have built a platform component that is genuinely ours, such as STATLAB's engine or COHORT's delivery machinery, you get a licence to it for the life of the study, and an exit path if you ever want to leave. We agree that path in writing before you start.

Yes, and it usually strengthens the application. We provide a work package description, a costed budget, CVs, and the methods and data management text for the technical parts. Reviewers respond well to a named partner with a track record where the alternative is “a student will build a database”.

Danish and EU schemes we have costed work packages for include Innovation Fund Denmark, Novo Nordisk Foundation, Independent Research Fund Denmark, Horizon Europe and EU4Health. If the grant is not awarded, you owe us nothing for the application work.

REDCap is always hosted by your institution. It is licensed to non-profits, so we cannot host it and do not try to. We build inside the instance your university or hospital already runs. See REDCap for exactly where that boundary sits.

Our own platforms, and anything we build for you, are hosted wherever the study requires: your own datacentre, one of ours, a partner datacentre in Denmark or Germany, a partner datacentre in the United States where that is appropriate, or public cloud. We decide this with you for each study on policy grounds, never by product tier or by default.

Where data cannot leave the environment it lives in, we work inside that environment under your project authorisation. We have collaborated with Statistics Denmark, the CPR register and Statens Serum Institut, and one of our services runs inside the Capital Region of Denmark's own datacentre. For consortia where no site may share raw data at all, we run federated analysis instead: the models go to each site and the records stay there.

Yes. You publish the analysis, and the assistant is not part of it. STATLAB produces standard R or Python using standard packages. The output is the same code a statistician would write, arrived at faster, and you can read every line of it.

We give you a methods paragraph describing exactly what was run, a machine-readable record of package versions and random seeds, and a container that reproduces the numbers. Where a journal asks about AI tool use, that record is what you disclose. We do not let a number into a manuscript unless you can personally trace it to code you have seen.

It is the normal starting point, and the only one we work from. REDCap is excellent at what it does, and all of our REDCap work happens inside an instance the institution already holds the licence for: designing projects properly, fixing ones that grew sideways, and adding the pieces REDCap deliberately leaves to others.

Where REDCap genuinely runs out — a hundred thousand invitations a wave, adaptive item banks, a participant app, live device data — we bolt our own platform alongside it through the API rather than replacing it. Your coordinators keep the interface they know. See REDCap and the survey platform.

A single-arm cohort database with clean instruments: two weeks. A multi-site randomised trial with e-consent, blinding and monitoring views: four to six weeks. A questionnaire platform for a national wave, including translations and a reminder ladder: four to eight weeks. A custom participant app in the stores: eight to twelve.

Analyses are faster: a well-specified survival or mixed-model analysis on a locked dataset takes days. What moves those dates is almost always approvals and the state of the existing data. We will tell you which of the two you are actually waiting for.

Builds are quoted as fixed fee against a written scope, so a grant can carry them as a line item. Retained support is a monthly hour pool. Hosting of our own platforms is per study or per institution, priced on participants and storage rather than on how successful your recruitment turns out to be. Your REDCap costs you nothing extra, because it is already yours.

We will also say when you do not need us: some of what people ask for is forty minutes of REDCap configuration that a research assistant can do with a written instruction. We would rather send the instruction and keep the relationship.

That is a large share of what we do. The usual pattern: recruitment is halfway, the data is in three formats, the person who built it has left, and something needs to be defensible by a deadline.

We start with a two-week assessment of what exists, what is recoverable, what has to be re-collected, and what the honest options are. You get that as a written report with a recommendation, and it is useful to you whether or not you continue with us. Occasionally the recommendation is to change the analysis rather than the data, and we say so.