I have started funded doctoral research into why organisations refuse to share data. Here is the question, why it has not been answered, and what would make me wrong
Last month I set out five ways platforms fail on adoption rather than on engineering, and ended by admitting something unsatisfying. When you ask five people why adoption will or will not happen, you get five confident and conflicting answers, and nothing that would tell you which of them is right.
I have decided to spend six years finding out.
As of this month I am a doctoral researcher at DCU Business School, supervised by Professor Tim Hubbard and Professor Pierangelo Rosati, and funded by the DCU Connect Scholarship.
The argument that never resolves
An organisation is invited to contribute data to a shared platform. The room divides.
Someone says the real obstacle is administrative burden, because nobody has the staff time. Someone says it is control, because once the data leaves you cannot govern what happens to it. Someone says it depends entirely on who owns the thing. Someone says people will do it if the price is right, and someone else says price is irrelevant because it is a matter of principle.
All five positions are plausible. All five are held sincerely. And the decision gets made by whoever is most senior or most persistent, because there is no evidence that would settle it.
Why it has not been answered
There is a large research literature on why organisations adopt or resist new systems. It is genuinely good, and it establishes which factors matter: switching costs, habit, perceived value, the weight of the incumbent system.
But it is built almost entirely on surveys. Respondents rate statements on scales, and the analysis establishes that a factor correlates with intention. That tells you a factor matters. It does not tell you what it is worth relative to another factor, which is the only form in which the answer is usable by someone designing a platform.
Meanwhile there is a method that does exactly that. Discrete choice experiments present people with realistic packages that differ across several features and ask which they would take. From those choices you recover how much weight each feature carries, and what people will trade for what. Health economics uses it to design services. Transport uses it to price roads. Environmental economics uses it to value things that have no market at all.
Information systems research barely uses it. That gap is the whole of my project.
The specific question
What determines whether organisations contribute data to shared infrastructure, and do the conditions under which they would join differ from the conditions under which they would continue?
The second half matters more than it looks. Almost everyone treats participation as a single decision. There is good reason from behavioural research to think that agreeing to join something and agreeing to stay in it are different decisions, evaluated against different reference points. If that is right, everything learned at launch about what recruited people tells you very little about what will keep them, and platforms designed on recruitment evidence end up governed on the wrong assumptions.
Why this is applied research, and what that changes
The Connect Scholarship is part of Dublin City University's strategic effort to tie university research directly to industry, so that it produces outcomes a business can use alongside the ones a journal will accept.
That is not a detail about funding. It changes what the research is.
It means the work runs alongside professional practice rather than replacing it, which is why I am still building platforms while studying why people decline to use them. It means the question came from a real problem rather than from a gap in a literature review. And it means each phase has to produce something usable before the thesis is finished, because a finding that only arrives in 2031 is no help to someone deciding something in 2027.
It also means the research needs organisations in it, which I will come back to.
Why charities, and why Ireland
Not because this is a charity problem. Because Ireland happens to be an unusually good laboratory for a general one.
Ireland built this kind of infrastructure and lost it. An independent nonprofit maintained the only live database of Irish nonprofits from 2015 to 2022, assembled from the published returns of more than 20,000 organisations. It was wound up when its lead government funder withdrew and no replacement coalition formed.
The European Union then created a legal form for exactly this kind of body. Across the entire Union, fewer than forty have registered. None in Ireland, despite a designated regulator being in place since 2024. The machinery exists and almost nobody uses it.
And there is a natural comparison available. A particular reporting standard is mandatory for charities in the United Kingdom and voluntary in Ireland. Irish organisations therefore reveal, in their filed accounts, whether they choose to disclose beyond what is required of them. That gives me something rare in this kind of work: a way to check what people say against what they actually did.
The same structure appears wherever an intermediary asks a constituency to contribute data it will then use. Health research networks. Education data partnerships. Municipal data sharing. Industry benchmarking. Charities are where the question is legible, not where it is confined.
The part I have to declare
I am professionally involved in a proposed shared platform for the Irish charity sector whose funding model assumes charities will accept revenue being generated from aggregated data they contribute.
That is precisely the assumption this study puts at risk. It would be straightforward to design a study that confirms what I already believe, and it would be worthless.
So the design is built against me. The attributes tested in the survey come from interviews with organisations, not from the initiative's assumptions. A sample of interview transcripts will be independently coded by someone other than me. And I have committed in the proposal, before collecting anything, to publishing findings that contradict the premise.
The proposal itself went through seven rounds of review, four of which removed something I was attached to. It is roughly half the scope it started at.
What would make me wrong
Three results would undermine the thing I set out to show, and all three are genuinely possible.
Commercial reuse of contributed data might weigh almost nothing next to administrative burden. My central concern would then be a preoccupation of platform designers rather than of the organisations they design for.
Organisations might decline to contribute on any terms at all. One national membership body has already told me it would be unusual for their members to share what they consider their data with anyone beyond their funders and their regulator. If that is the settled position of a sector, no governance arrangement fixes it, and that is a finding worth having.
And joining and staying might be governed by identical terms. That would mean a widely accepted behavioural finding does not extend to organisational decisions, which is also worth knowing.
Each of those outcomes is publishable. The study is designed so that none of them is a failure.
What happens next
The method is tested even though none of my own data exists yet. I have replicated a published route choice model end to end in R, using the Apollo package, so the estimation approach is known to work before any responses arrive. That is a small thing and it matters, because a study that discovers its method does not run, after fieldwork, has wasted a year.
The first phase is interviews, and it will not begin until ethics approval is in place next year. Before then there is a great deal of reading and a research design to finish specifying.
On organisations in the research. Because this is funded as applied work, there is a route for an organisation to be more than a respondent. Enterprise Ireland's Innovation Partnership scheme lets a company collaborate with a university on a defined problem with the state carrying much of the cost, and in this case the question is already defined and the method already built. If you operate or are building shared infrastructure that depends on other organisations contributing to it, that conversation is open.
I will keep writing here, mostly about method rather than findings, because findings are three years away and anyone claiming otherwise is not doing research.
The next piece is about what data lineage actually costs and what it buys. It turns out to be the same question wearing different clothes: the price is paid by the people who submit the data, and the benefit accrues to the people who read the figures.
I am a product and platform consultant, a doctoral researcher and a Connect Scholar at DCU Business School. Client, entity and vendor identities are omitted throughout.