Doctoral research
Joining is not staying
What makes an organisation contribute data to shared infrastructure, and are the terms that recruit it the terms that keep it?
Applied doctoral research at DCU Business School, funded by the DCU Connect Scholarship and carried out alongside practice.
The problem
Most digital platforms fail on adoption rather than on engineering.
The system works. The data model is sound. The dashboards load. And then the people it was built for carry on emailing spreadsheets to each other, or the partner organisations that signed up to contribute data quietly stop contributing, and the business case evaporates eighteen months after the money was spent.
The uncomfortable part is that the reasons were usually knowable in advance. They were simply never measured. Adoption gets handled with workshops, readiness assessments and communication plans, all of which describe attitudes. None of them produces a number you could have put in the investment case.
Why platforms fail on adoption develops this argument at length.
What the research asks
Shared data infrastructure only works if organisations contribute to it. A sector platform, a federated analytics arrangement, a data space, a regulator's reporting hub: each depends on participants handing over something valuable, under terms they did not write.
This research measures what those organisations will actually accept.
Specifically, it asks which governance terms move the decision: ownership, permitted use, aggregation thresholds, control rights, exit provisions. And it separates two decisions that are routinely treated as one - the decision to join, and the decision to keep contributing. There is good reason to think the terms that recruit an organisation are not the terms that retain it, and that the gap between them explains a great deal of quiet platform failure.
Why discrete choice experiments
The method is borrowed rather than invented, and that is the point.
Discrete choice experiments are the standard instrument in health and transport economics, where governments use them to set policy, price public services and value things that have never been sold. They work by presenting people with realistic packages that differ on the attributes that matter, observing what they choose, and estimating from those choices how much each attribute is worth.
They answer a question a survey cannot. Asked directly, every respondent says every factor is important. Forced to choose between options that are each better on something and worse on something else, they reveal what they will actually trade away, and at what rate.
The method is almost entirely absent from information systems and technology work. Applying it to organisational adoption decisions, rather than to individual consumers, is the methodological contribution of the study.
Estimation is in R, using the Apollo choice modelling package: attribute elicitation through structured interviews, an efficient experimental design, then mixed logit and latent class models to capture the fact that organisations do not all want the same things.
Six Years on One Question sets out the question in full, including what would make me wrong.
Applied research, by design
This work is funded by the DCU Connect Scholarship, part of Dublin City University's strategic effort to connect university research directly with industry so that it produces business and societal outcomes alongside academic ones.
That funding model shapes the research rather than merely paying for it. The work runs alongside professional practice instead of replacing it, the question was drawn from a real platform problem rather than from a literature gap, and the output is designed to be usable by an organisation making a decision, not only publishable by a journal that reviews it.
The practical consequence: every phase produces something a business could act on before the thesis is finished.
Where this applies
The method generalises well beyond the thesis case. Five areas where it answers a question that currently gets answered by opinion.
Adoption risk, before the investment is approved
Will the intended users adopt this, and what would have to change for them to do so? Predicted uptake under alternative designs, by segment, before the build.
For transformation directors and investment committees deciding whether a platform business case survives contact with its users.
The terms of data sharing and collaboration
Which governance terms get organisations to contribute, and which keep them contributing. A terms sheet ranked by its effect on participation rather than drafted by analogy.
For consortium leads, sector bodies, platform operators and regulators.
Pricing, packaging and access
What customers will give up, and at what price, for something that does not yet exist. Willingness-to-pay by segment, and the attribute actually holding people on the lower tier.
For commercial leadership and product teams who have run out of road on A/B tests.
AI adoption in regulated settings
Will the people who have to work alongside an AI system accept it, under which guardrails, and where must a human stay in the loop for it to be used at all? A human-in-the-loop specification derived from what users will accept rather than from what the risk function prefers.
Governance that gets used
Metric definitions, lineage and access control designed around what the people who must follow them will actually tolerate. The difference between a policy that is agreed and a policy that is obeyed.
These are the application areas I am building into a professional services capability, in partnership with firms who have the client relationships and the delivery capacity to carry it.
Partner with the research
Innovation Partnership
Enterprise Ireland Innovation Partnership funding lets a company collaborate with an Irish research institution on a defined problem, with the state carrying a substantial share of the cost.
Most such partnerships begin by finding a researcher and shaping a question. This one starts further along: the question is defined, the method is built and tested, and the researcher is registered and funded.
This suits an organisation that:
- operates or is building shared data infrastructure that depends on other organisations contributing
- is deciding the terms of participation and would rather measure them than negotiate them
- has a platform with disappointing uptake and no diagnosis
- needs to price or package access to data and has no empirical basis for it
What it would involve: a scoped problem, a defined deliverable, co-funding through Enterprise Ireland, and findings your organisation can act on while the academic work proceeds in parallel.
Start with a conversation - we can establish in twenty minutes whether the fit is real.
Where the work is now
Registered October 2026 at DCU Business School, part-time across six years.
- Phase 1 · 2026 to 2027
Structured interviews with the organisations whose decisions the study models, to establish the attributes that actually drive participation. Nothing in the choice experiment is designed until this is finished; the interviews determine the attributes, not the other way round.
- Phase 2 · 2027 to 2028
The discrete choice experiment itself, fielded across the population of potential contributors, pre-registered before data collection.
- Phase 3 · 2028 onwards
Estimation, latent class analysis to identify segments that respond differently, and the terms sheet that follows.
A published route choice model has already been replicated end to end in R, so the estimation approach is tested ahead of any primary data.
Findings will be published as they clear review. There are none yet, and I will not pretend otherwise.
Supervision and affiliation
Doctoral researcher and DCU Connect Scholar, DCU Business School, Dublin City University.
Supervised by Prof. Tim Hubbard and Prof. Pierangelo Rosati.
Contact
If you are designing a platform, a set of sharing terms, or a price whose success depends on other people choosing to participate, the useful conversation happens before the build rather than after it.
The form reaches me directly. Chronology, roles and education are in the CV.