Many scholarly infrastructure business models fit institutions poorly. They tie cost to transactions, outputs, or rising use in ways that make adoption harder to sustain over time. That is a problem for libraries and research institutions, which are trying to normalize and support good practice, not disincentivize against it by tying payments to countable units. DataCite’s recent move away from per-DOI pricing offers a useful example of a different approach. It shows what it can look like when an infrastructure organization designs its funding model in a way that better matches institutional values and institutional realities.
The issue is not whether institutions understand that infrastructure costs money. Of course they do. In some cases, they will even choose to build and maintain systems themselves, not because doing so is necessarily cheaper or more efficient, but because predictable responsibility can feel easier to manage than open-ended financial exposure. The real question is whether infrastructure can be supported in ways that are stable and aligned with how institutions actually work, while providing sustainable funding to those providing the infrastructure. Transactional infrastructure models are not only competing against one another. They are also competing against institutional instincts toward local control, internal ownership, and operational predictability.
Too often, the answer is no. Many infrastructure models still assume that use should be counted and billed one unit at a time. That logic may work in settings where value is primarily individual and immediate. However, scholarly infrastructure is usually not one of those settings. Repositories, PID systems, metadata services, curation platforms, and related infrastructure create value that institutions sustain collectively so those capabilities remain broadly and equitably available over time. They help preserve the scholarly record, improve practice, support compliance, and make core capacities available to the communities institutions serve.
This is why collective models matter. A transactional model ties cost to individual acts of use. A collective model ties support to sustaining shared capacity. That distinction is not just financial. It shapes whether institutions can confidently encourage adoption, whether the entire community is rewarded for broad uptake or punished by it, and whether infrastructure actually behaves like and is treated like a shared investment rather than just another vendor.
The issue is not just paying, but paying unpredictably
Institutions around the world understand that scholarly infrastructure costs money. The harder question is whether those costs can be structured in ways institutions can support responsibly over time. By ‘transactional model’, I mean a model in which institutional cost rises with discrete acts of use, outputs, or participation, rather than through a stable collective commitment to sustain the infrastructure as a whole.
For most institutions, the central problem is not cost itself but unpredictability. Libraries and research organizations can plan for recurring commitments when those commitments are understandable, stable, and tied to a known scope of support. What becomes difficult is a model in which ordinary adoption, routine use, or successful growth creates open-ended financial exposure. At that point, infrastructure stops feeling like an investment and starts feeling like a moving target.
That changes the internal conversation institutions have about infrastructure. Instead of asking how to support participation, improve practice, or build durable capacity, institutions begin asking what happens to the budget if those efforts succeed. Once that happens, the pricing model is no longer neutral. It begins shaping behavior. This is one reason transactional models can create friction in institutional settings. Research institutions operate through annual budgets, recurring commitments, and long-term operational responsibilities. They can usually absorb predictable commitments much more easily than uncertain growth in costs tied to usage. When ordinary success becomes financially difficult to forecast, it becomes harder to normalize infrastructure, harder to advocate for adoption locally, and harder to sustain support over time. In some cases, institutions may also become more cautious about dependence itself, particularly when local ownership appears easier to predict than external variable costs.
Transactional pricing creates perverse incentives
The problem with transactional models is not only that it can be harder to budget for. It is that it can create incentives that run directly against the purpose of the infrastructure itself. When cost rises with each deposit, identifier, publication, or workflow, institutions can find themselves paying more precisely when the infrastructure is being used in the ways they have been encouraging all along.
That is the core perverse incentive. Libraries and research institutions are often trying to move communities toward better practice: more data sharing, stronger metadata, broader use of PIDs, better preservation, and more consistent compliance with policy and funder expectations. They are often the very outcomes that infrastructure providers, libraries, and research offices all say they want. But if each incremental success also increases cost, then business models start to punish adoption rather than support it.
The tension may not always be stated explicitly, but it shows up in institutional decision-making all the time. A library that wants to normalize good practice across the research lifecycle has to consider whether each additional use carries a budget consequence. Once that happens, pricing models are no longer neutral. It is shaping behavior. And it is doing so in a way that makes institutions more cautious about advocating for the very practices they are supposed to enable.
This is especially problematic in infrastructure, where the goal is not to reduce demand but to build uptake and shared norms. If adoption increases cost, then the pricing models are working against the purpose of the infrastructure. That is not a minor design flaw. It is a sign that the models are poorly matched to the role the infrastructure is meant to play.
Infrastructure does not produce one-to-one value
A transactional model assumes a fairly direct relationship between user, benefit, and payment. That logic may work in settings where the value is primarily individual and immediate but scholarly infrastructure is usually not one of those settings. The immediate user may be a researcher depositing a dataset, registering an identifier, or using a metadata or curation service, but the value created does not stop with that act or with that person.
The benefits of scholarly infrastructure are often distributed across time. They include stewardship of the scholarly record, support for better practice across the institution, the existence of shared capacity that can be relied on when needed, and value that extends well beyond the immediate user to librarians, curators, educators, compliance staff, future researchers, and the entire research ecosystem. In other words, the person using the service is not always the only one who benefits, and often not the only one the institution is trying to support.
Pricing models do not just recover cost. They also express an assumption about where value resides. When infrastructure is priced transaction by transaction, it privileges the most visible unit of activity and can understate the wider institutional and communal benefit being sustained. It treats infrastructure as if institutions were merely reimbursing individual use rather than sustaining institutional research capacity. This is one reason the fit can feel so poor from an institutional perspective. Institutions are often not paying simply to cover a series of isolated acts. They are paying to sustain continuity, readiness, and the conditions for good practice over time. That is a different kind of value proposition. And it is one that is often misdescribed when the funding model is built around one-to-one transactions.
Collective models fit how institutions actually work
Collective models often fit institutional operating realities better than highly transactional ones. Research institutions do not support repositories, PID systems, metadata services, and preservation infrastructure simply as narrow transactional services. They support them as part of the broader research environment: shared capacities that enable stewardship, compliance, discoverability, and continuity over time.
That distinction matters operationally as much as philosophically. Institutions generally manage infrastructure more effectively when support can be framed as a stable commitment with a clear communal rationale, rather than as a variable bill tied to successful engagement. A collective model fits that reality because institutions are usually trying to sustain shared research capacity, not support one researcher transaction at a time.
This does not eliminate the need for accountability or careful governance. It does mean, however, that the financial model begins to match the way institutions actually justify, budget for, and sustain infrastructure over time.
Predictability is a feature on both sides
Predictability benefits infrastructure organizations as much as institutions. A collective funding model does not eliminate risk, but it changes how risk is managed. Instead of pushing financial variability back onto institutions one transaction at a time, infrastructure providers must think more deliberately about scope, staffing, reserves, governance, and sustainable growth. That can be a strength rather than a weakness. A predictable model encourages a clearer relationship between mission, service levels, and community support. It pushes organizations to be more transparent about what baseline funding supports, where additional investment is needed, and how expansion should occur responsibly over time.
Transparency is essential. Communities should be able to understand how fees are determined, what the baseline covers, what reserves exist, and how future changes will be evaluated. A sustainable collective model should also plan for continuity, including governance transitions and responsible wind-down where necessary. In that sense, predictability is not simply a concession to institutional budgeting concerns. It is part of a healthier governance model for infrastructure itself. For institutions, predictability can make support easier to justify and sustain. For providers, it creates a more intentional and legible operational model grounded in long-term stewardship rather than continual transactional expansion.
Collective does not mean crude or unfair
A collective model does not require pretending that all institutions are the same, and it does not require a single identical fee for everyone. The more important distinction is whether the model is structured around sustaining shared capacity. Infrastructure is itself inherently collective: its value emerges not only from individual acts of use, but from the existence of shared systems that institutions and communities can rely on over time. A model can still be collective while using formulas, tiers, or contribution levels that reflect institutional size, capacity, or role. Not every differentiated pricing model is transactional in the problematic sense; a model can still be tiered or formula-based and remain collective if it is designed to sustain infrastructure. This does not have to remain abstract. DataCite recently announced a new membership model that moves away from per-DOI pricing and toward a collective approach, with fees adjusted using country-level economic indicators. It offers a useful example of how an infrastructure organization can differentiate contributions without reverting to transactional logic.
One of the easiest objections to collective funding is the claim that it is somehow simplistic or unfair. In reality, the fairness of a model can recognize institutional differences without reducing infrastructure support to a running tally of routine use. In practice, some of the strongest collective arrangements are precisely the ones that are transparent about how costs are distributed and why. Formula-based approaches can make it possible for larger institutions to contribute more, for smaller organizations to participate meaningfully, and for the whole community to see how growth, accountability, and shared support fit together. That is a very different logic from a model where each additional act of participation simply adds to the bill. So the choice is not between a crude flat fee and finely tuned fairness. The better contrast is between models that fund infrastructure predictably and models that meter normal use as though the infrastructure were simply a series of individual transactions.
Infrastructure has to be more than a new vendor
There are already plenty of vendors in the scholarly infrastructure space. The point of infrastructure is not simply to add more providers to that landscape. It is to create a different kind of relationship, one grounded in stewardship, accountability, transparency, and a clearer connection between what is being funded and the broader benefit it supports. When that distinction fades, something important is lost.
This is why the funding model matters beyond pricing mechanics. If a membership organization or mission-driven infrastructure relies too heavily on transactional logic, it can start to look and feel like an ordinary vendor relationship. The institution is no longer supporting shared capacity with a communal rationale. It is managing exposure, counting usage, and evaluating the arrangement as one more procurement decision, often against the perceived predictability of local control. At that point, the infrastructure may still be valuable, but its business model is no longer doing much to distinguish it from the market logic it was supposed to counterbalance.
That does not mean infrastructures should ignore sustainability, discipline, or cost recovery. It means that their distinctiveness should show up in the model itself. If a community is being asked to support infrastructure because it is mission-driven or built for long-term collective benefit, then the way it is funded should reflect those same qualities. Otherwise, the rhetoric starts to drift away from the reality of the relationship.
The issue is not simply whether institutions can afford one more service. It is whether the scholarly community is building genuine infrastructure or just reproducing the vendor landscape in a slightly different form. If infrastructures adopt vendor logic, it risks becoming just another vendor in a space that already has too many. We do not need more infrastructure providers acting like vendors. We need infrastructure that actually operates like a investment.
The OA conversation offers a useful parallel
Research infrastructure and open access publishing are not the same thing, but there is still a useful parallel. Institutions have already spent years grappling with business models that tie costs to discrete units of activity. APC-based publishing models are one familiar example. The issue is not simply that institutions pay for publishing. It is that costs can become difficult to predict and difficult to sustain as adoption grows.
That tension is one reason institutions have often been drawn toward models that create more stable and legible forms of support, whether through negotiated agreements, collective funding arrangements, or Diamond OA approaches that move away from per-article transactions. In some cases, these efforts have also produced collective funding experiments, including models such as the Open Library of Humanities and SciPost, that attempt to sustain publishing infrastructure through broader community support rather than article-level billing. These approaches are not interchangeable, but they reflect a familiar institutional preference: supporting systems through models that are easier to normalize and sustain over time.
The same tension appears in scholarly infrastructure. Institutions are not only evaluating whether a service is useful. They are evaluating whether the funding model fits the operational realities of long-term institutional support. In that sense, the OA conversation is useful not because it offers a perfect analogy, but because it demonstrates a broader and recurring challenge across scholarly communication: how to sustain these systems without making adoption financially destabilizing.
The pitfalls of collective models
Collective models are not perfect, and the argument for them is weaker if that is ignored. As communities grow, governance can become more complicated. Not every participant wants deep engagement, shared decision-making, or a strong sense of communal identity. Some institutions may still evaluate a collective arrangement through a narrow transactional lens, asking whether local usage "justifies" the cost even when the benefit is broader and less easily measured. Those are real challenges.
There are also legitimate questions about scope and edge cases. Some services or institutions may place unusually large demands on an infrastructure. Some kinds of support may be bespoke enough that they do not fit neatly within a shared baseline model. A collective approach does not eliminate those questions. It simply means that they should be handled as contextual decision points within a framework, not used to define the logic of the whole system.
It is also true that collective models require explanation. Institutions need to understand what they are supporting, how the model works, and why the benefit is greater than a count of immediate uses. Providers need to be transparent about what the baseline covers, how contributions are determined, and where additional needs may arise. Without that clarity, a collective model can start to feel vague or unconvincing, especially in institutional settings where every recurring cost is under scrutiny.
But none of those tradeoffs change the larger point. The existence of complexity does not mean transactional pricing is better. It means collective models have to be designed carefully, governed well, and explained in a way that makes the communal value visible.
The real choice
When it comes to infrastructure, the real choice is not between paying for it and getting it for free. Institutions have never sustained research support that way. The choice is between models that sustain shared institutional capacity and models that tie cost to participation. That distinction is important because it shapes not only what institutions pay, but how they understand the relationship, how they justify it internally, and how confidently they can encourage broader adoption.
If institutions are going to sustain repositories, PID systems, metadata services, curation platforms, and other scholarly infrastructure as part of their remit, then the business models behind those services need to reflect the kind of value they create. In most cases, that value is not only singular and immediate. It is institutional, communal, and long-term. It shows up in stewardship of the scholarly record, in support for better practice across the institution and in benefits that extend well beyond the immediate user.
That is why this conversation is ultimately about more than pricing. It is about what kind of infrastructure ecosystem the scholarly community wants to build. For institutions, that means moving away from models that punish ordinary success and toward models that support capacity building, predictable commitment, and long-term stewardship. DataCite’s recent move away from per-DOI pricing offers one important example of what that transition can look like in practice. We need more infrastructures to also operate like investments in the future.
Acknowledgements: Thank you to Kathleen Gregory, Dorothea Strecker, Maria Gould, and my colleagues at California Digital Library (Günter Waibel, Catherine Mitchell, Miranda Bennett, Kurt Ewoldsen) for their help with this article.
Copyright © 2026 John Chodacki. Distributed under the terms of the Creative Commons Attribution 4.0 License.
