The goal is that you need us less.
Most of what research groups pay consultants for is knowledge, and knowledge can stay in the group after we leave. We teach the things we are otherwise hired to do: database design, applied statistics, reproducible analysis, governance, qualitative method. On your own data, with your own protocols as the exercises.
A course that ends with everybody nodding has failed. It should end with somebody having built the thing badly, been told why, and built it again.
Teaching what we sell looks commercially odd for about a year. Then the groups we taught come back with harder, better-specified problems and with somebody in the room who can tell us when we are wrong. We value that work most.
It also fixes the thing that actually limits research groups: a database built by whoever was free, an analysis nobody can re-run, and a governance answer that lives in one person's head until they graduate.
So the courses are small and hands-on, and the exercises use your own protocols and data, because a tidy textbook example never resembles a real study.
Six people, six different courses.
A PhD student, a study coordinator and a professor need almost nothing in common from a course on the same subject. Pick who is asking. The recommendation changes, and so do the length and the price.
What we can actually teach.
Every one of these is something we are otherwise paid to do, which is the only qualification for teaching it we consider sufficient. Each is built from your material before it is taught, so the exercises use your own study.
Design it once, properly
How to turn a protocol into a data dictionary: field types, coded values, validation, longitudinal event grids, and the four decisions that decide whether cleaning takes a week or a quarter.
- Hands-on in your own REDCap instance
- Participants leave with a built, reviewed project
- Half a day to two days depending on the group
The methods your field actually uses
Survival analysis, mixed models for repeated measures, regression with the functional form checked, and missing data handled as a mechanism rather than as a default.
- In R, on your own dataset or a realistic synthetic one
- Assumption checks as a habit
- A course series over a term, or an intensive week
So the numbers survive the reviewer
Version control that a non-programmer will genuinely use, project structure, dependency pinning, containers, and the discipline of a frozen analysis dataset.
- Git for people who do not want to learn Git
- From script-on-a-laptop to a re-runnable supplement
- Usually the highest-value two days a group can spend
Not the compliance lecture
What the rules actually require of a study, written for people who have to apply them: lawful basis, pseudonymisation, transfers, DPIAs, and what to do when a participant withdraws.
- Worked through with your own submissions
- Includes the questions to put to your own DPO
- Half a day, and it prevents months
Codebooks and coder agreement
Thematic analysis and the framework method taught as practice: writing code definitions that two people can apply the same way, and what to do when they cannot.
- Taught in CORPUS, on your own transcripts
- Double-coding exercise with real disagreement
- Reporting to COREQ without inventing the answers
For groups who will do several
Protocol, search strategy, screening discipline, extraction and risk of bias, with the search taught by an information specialist.
- Build a real search strategy for your real question
- Screening calibration on fifty records
- Two days, or a supervised review alongside your first one
Where it helps, and where it lies
An honest working course: what current models are good at in a research workflow, what they are dangerous at, how to disclose their use, and how to check their output.
- Hands-on with the tasks in your own week
- Prompting, verification, and knowing when to stop
- The disclosure statement your journal will ask for
The checklist as a design tool
CONSORT, STROBE, PRISMA, TRIPOD+AI and COREQ applied at design time, the way they were intended, so the week before submission is not spent ticking boxes.
- Methods and results sections, drafted in the session
- Responding to reviewers without conceding the argument
- Runs well as a series of short sessions
Five shapes. Pick the one your calendar allows.
| Format | Length | Best for | What it does not achieve |
|---|---|---|---|
| Workshop | Half a day to two days | One subject, one group, hands on their own material. The most common request and usually the right one. | Lasting fluency in a language or a method. That takes practice afterwards. |
| Course series | 6–10 sessions over a term | Statistics or reproducible analysis, where spacing matters and people need to try things between sessions. | Nothing much — it is the format that works best, and the hardest to schedule. |
| Supervised project | Alongside a real study | Learning by doing the actual work with somebody watching. A first systematic review, a first trial database. | Speed. It is deliberately slower than us doing it. |
| Embedded period | 1–6 months, part time | A department that wants capability rather than a deliverable — see the engagement models. | Being cheap. It is the most expensive option and often the best value. |
| Conference or seminar | 45–90 minutes | Raising the floor across a large audience, or making a case to a leadership group. | Any actual skill transfer. Useful, but do not mistake it for training. |
People learn this by breaking it.
The single biggest difference between a course that changes practice and one that does not is whether participants built something and had it critiqued. So roughly two thirds of every session is hands on, and the material is their own.
So preparation is more important than delivery. Before a course we read your protocol, look at your existing database or analysis, and build the exercises from what is actually there — including, with your permission, the mistakes.
- Small groups — four to twenty. Above that it is a seminar, and we will call it one
- Your data — or a synthetic dataset built to resemble it, where governance requires
- Two thirds hands on — the rest is framing and critique
- Danish or English, and the materials in both where a group is mixed
- Materials you keep — slides, code, exercises and solutions, reusable internally
- A named teacher who does this work for a living
- Follow-up — one session a few weeks later, when the real questions have surfaced
- Honest scoping — if a course will not fix your problem, we say what would
What groups ask before booking.
Can this count as formal PhD course credit?
That is your graduate school's decision, not ours, and it varies by institution. What we can do is provide what they need to make it: learning objectives, hours, assessment, and attendance records in the format they use.
Several of our courses have been run as part of doctoral programmes with ECTS attached. Ask your graduate school first — it is much easier to design a course to their requirements than to seek approval afterwards.
Can you teach on our data, given the governance?
Usually yes, and where we cannot, a good synthetic substitute is nearly as effective. Teaching on real data is better because the awkwardness is real — the miscoded field, the impossible date, the two rows for one participant — and that awkwardness is the actual lesson.
Where the material cannot be shown to an external teacher, we generate a synthetic dataset that preserves the structure and the problems without any real person in it. See secure data handling.
What does it cost?
Priced per course rather than per participant, so a department filling the room is not penalised for it. A half-day workshop is a small fixed fee; a term-long series is quoted against the preparation it needs, which is where most of the work sits.
Where we are already engaged on a project, teaching the relevant part to your team is often included, because a group who can maintain what we built is cheaper for us to support.
Are the courses tied to your platforms?
No, and we would rather they were not. Statistics is taught in R. Reproducible analysis is taught with Git and containers. Databases are taught in the REDCap your institution already runs.
Two courses do use our own tools, because they are the subject: qualitative coding in CORPUS and AI-assisted analysis in STATLAB. In both cases the underlying method is taught so that it transfers to other software, and we say plainly which parts are tool-specific.
Can one of you join our group for a while instead?
Yes — a secondment for one to six months, part time, sitting inside the group as staff rather than delivering a project from outside. It is the format that transfers the most capability and the one departments come back for.
It works best with a stated goal beyond “help us”: a named system to build with your people, or a named capability the group should own by the end. Without that it drifts into being cheap labour, which serves nobody.
Will you tell us we do not need a course?
Regularly. A surprising number of training requests are really a request for a decision, a written standard, or one person's afternoon. Teaching twelve people to design databases will not help if what the department actually needs is one agreed template.
So the first conversation is diagnostic, and it is free. Sometimes the recommendation is a one-page standard and a review, and that is what we will propose even though it is the smaller invoice.
Everything here is taught.
Tell us who is in the room.
How many people, what they already know, and the thing that went wrong recently which prompted the thought. From that we propose a format, a length and a price — or tell you that a written standard would serve you better than a course.