Membership & Pricing

By your institution's own annual research output: the number of works attributed to your ROR identifier in the metric year of the edition in force. Not full-time-equivalent staff, not Carnegie class, not self-reported revenue, and nothing you have to disclose to us. Nobody has ever been quoted anything, so what follows is how a quote would be built, not how one is built. The exact query, the snapshot date and the resulting figure would be printed beside the price, so that no band ever had to be accepted on trust; and where the figure was wrong — a merged institution, a missing affiliation string, a hospital counted twice — you would say so and we would re-run it in front of you.

That annual figure is not published on this site yet. The organisations directory carries a column headed “Articles”, and it is the obvious thing to read against the table — but it is the lifetime cumulative count OpenAlex keeps. Measured on one named institution: Ege University's row reads 50,811, which read against the annual thresholds is the top band, I6. Its actual output in 2025 was 4,110 works, which is band I4. It is a Turkish institution, so the only figures it could ever be quoted are the lira ones: following the old instruction would have put it at 280,000 TRY where the annual figure puts it at 190,000 TRY — 90,000 TRY too much, and $6,500 on a dollar list no Turkish buyer would be invoiced from. Across the directory the same misreading puts 2,366 of the 9,483 education institutions with 1,000 or more cumulative works into the top band, where their annual output puts 122; in Türkiye it puts 104 of 214 universities in the top two bands, where the annual figure puts three.

Why we are not simply rescaling the published column. The ratio between the lifetime column and the annual figure is not a constant: across those 9,483 institutions its quartiles are 7.0, 10.4 and 16.3, and the top twentieth is above 32. It tracks how old an institution is and how far back its record is indexed, not how much it publishes now. Dividing by any single factor would move the error around rather than remove it.

What has to be built is one script. OpenAlex publishes counts_by_year on every institution record. We have run that harvest once, off this site, for all 126,410 organisations, in order to cut the thresholds against the real distribution. What is not done is loading it into a side table and printing each institution's annual output on its own page. That sits on the membership work list, and no institutional quote goes out before it is done.

A price list keyed to a variable nobody has counted is a guess with a table around it. So on 4 September 2026 we read OpenAlex's counts_by_year for every one of the 126,410 organisations in the directory and counted where they land on their 2025 output. Universities is all 24,418 education institutions; buyer pool is the 9,483 of those carrying 1,000 or more works in their lifetime — the ones with a library, a serials budget and a reason to read this page.

What the count changed. Two boundaries, both at the top: I5 now starts at 6,000 works a year instead of 8,000, and I6 at 12,000 instead of 20,000. The two most expensive rungs had held 251 and 35 universities on the whole planet. 260 organisations move up one band as a result, 229 of them universities; nothing below 6,000 works a year changes, and nothing moves down. Nobody has ever been quoted a price, so no invoice changes — but the old thresholds were published, so the correction is printed rather than made quietly.

What it did not change. 73% of universities land in band I1 and would pay the $750 floor: most organisations holding a ROR identifier publish a few dozen works a year and will never buy this. Among the 9,483 that plausibly might, the ladder separates them: 31% at the floor, 40% at $1,500, 19% at $3,000 and 10% above. Türkiye's 214 universities sit 43 / 82 / 77 / 9 / 3 / 0 across the six bands. Raw upstream counts occasionally break — one facility record reports 18,940,310 works for 2025 alone — and that is what the correction route is for, and why the query and the snapshot date would be printed beside the price.

Call S the sum of what the member institutions would each pay on their own band. The consortium pays the smaller of S and the larger of 60% of S and $3,000. The floor exists because a consortium contract, a shared contact and one reconciliation run would cost us about the same whether there were three members or thirty. The outer bound is the promise that matters: the floor can never carry the price above S, so joining a consortium is never more expensive than not joining one. Until 3 September 2026 the row read “× 0,6, floor 3,000” with no upper bound, and that version would have charged two band-I1 institutions 3,000 for a joint licence they could have bought separately for 1,500 — printed here because a schedule that has been wrong should say where.

Three stretches: at or above S = $5,000 the full 40% applies; between $3,000 and $5,000 the floor takes part of the discount, which is what a floor is for; below $3,000 the members pay exactly what they would have paid one by one — no saving, no penalty, and still one negotiation. Where 60% of a sum lands on a half dollar it is rounded down, like every discount on this page. In Türkiye each member is its own contracting authority and would be invoiced member by member, which keeps every member under the direct-procurement limit and is not a split purchase, because there is no single purchase being split.

The Organisation band and the Team tier both list publishers among their buyers; Premium is sold to one person, and a journal's editor is one person. Against that, the promise to journals below is that nothing about your own titles is sold to you. Those sentences survive together only if the line between them is written down: you would buy the same reader-side service everyone else buys, and the journals you publish or edit yourself would be cut out of every part of it. It is one rule, and it governs every paid tier on this page — Premium, Team, the institutional bands and Organisation alike.

• ✓ May buy. Reconciliation, standing and dated reports, cross-edition views, a change feed and the support obligation — computed over a list you send us, with your own titles struck out. And the dated bulk editions, which are exempt on a stated principle rather than by oversight: an edition is the whole catalogue, the same file for every licensee, so it carries your titles for the same reason it carries everyone else’s and singles nobody out. Both of the fee policy’s tests hold on it — accepting a title moves no fee, and delisting one removes it from every copy alike — and the exemption is exactly as wide as that reason. Filter an edition down to a named list and it stops being the undifferentiated whole: it is then a deliverable, and the cut-out applies to it like the rest.
• ✕ May never buy. Anything whatever about your own titles: a report, a feed, an alert, a reconciliation row, an early sight of an edition, or a private answer on whether one of them is listed, scored, flagged, under observation or removed. Three places would carry all of that for nothing. One of the three exists. Free, permanent, and readable without an account: the public profile, carrying the catalogue record and the citation metrics on every journal, and the subject quartiles and the Crossref deposit scorecard on the journals they have been computed for — 23,426 of the 62,643 journals hold a quartile at all, and 6,743 have no Crossref record to score. That unevenness sits in the data rather than behind a gate: a deliverable would be built from the same store, so on your own titles it could say no more than the profile already says, and where the profile is thin it would be thin in the same places. The criteria report and the panel’s evaluation surface are meant to be free and permanent too, and today are neither issued nor built. So the cut-out removes work an editor would never have needed to buy, once the free half of that sentence is finished — which it is not yet.
• ✕ May never be priced by. Titles held, titles listed, acceptances or seals. The Organisation fee is flat, and the only thing that can override it upward is the buyer's own research output — a quantity we neither control nor influence.

Which titles are your own, and which are not. A journal is yours if you are, or act for, the party that would hold its account in the publisher panel: its publisher, its owner, or the editorial office that decides what it prints. It is not yours because you have published in it, refereed for it, or sit on its board from outside. A researcher's list is made of the journals they read and publish in — that is what the tier is for — and a rule that caught the reader would be worse than the gap it closed.

How it would be applied, and the limit that leaves. We will not ask you to declare an employer, and we will keep no register mapping a buyer to a publisher: the second undertaking under what the licence does not buy is that the evaluation side will never be told who holds a licence, and a register of that kind is precisely the thing that undertaking is about. So the rule would be applied where the deliverable is made, by the person making it, out of two things and no third: the titles you name as your own when you send the list, and what the catalogue already records about who publishes them. A title that came out would be named in the deliverable and marked struck out rather than quietly absent; if one were missed, we would remove it, reissue the report and refund the period. Now the limit, because a rule that hides one is worse than no rule. A buyer who names nothing, and whose connection to a title the catalogue does not record, is not caught. There is no check at the door and we are not proposing to build one: the only two checks that would work are the declared employer and the buyer-to-publisher register we have just refused, and refusing them costs exactly this. What is left is a term of use with an auditable object rather than a lock — the guarantee is that a report either contains a title or shows it by name as removed. Nothing here is gated, so there is nothing a gate could protect. What is at stake is not what a publisher might learn — all of it is public — but whether we would ever take money in exchange for our attention to a journal's own record.

What the rule does not reach, named rather than left to its author. Both tests below are run against the buyer's own titles, so neither of them sees the purchase an editor is most likely to actually make: Premium on a list of competitors' journals. That sits outside the rule, deliberately. A list of journals you do not control is the ordinary reader-side list this tier exists for — a researcher's list is made of journals they do not own either, and a rule that caught the one would catch the other. The two tests still pass on it: the fee is flat, so nothing about a competitor's acceptance moves it, and delisting a competitor would neither raise the invoice nor lower it. What we will not pretend is that such a motive would be invisible to us; the list would be in front of the person producing the report. The line we would hold is the deliverable's contents, not the buyer's reason for wanting it — and every number in it is one a free reader can already read for nothing.

The rule is the two tests the fee policy sets for itself, applied to every tier. Is the revenue tied to acceptance? No — accepting a thousand of a publisher's journals tomorrow leaves its invoice on the same figure. Does delisting cost us money? No — removing one of its journals removes nothing from what it bought, because its journals were never in it. Until 5 September 2026 this rule named only the Organisation band and the Team tier, which left the cheapest tier on the page as the one a journal's editor could have used to buy a standing report on their own journal for $59 — printed here because a rule with a hole in it should say where the hole was. None of this has been tested by an actual sale: nothing is on sale, no deliverable has been produced, and no title has yet been struck out of anything.

The $750 floor is set where a contract, an invoice and a year of support would stop costing more than the fee, and it is roughly a third of the median article processing charge in our catalogue, $2,697. The steps are ×2, ×2, ×1.83, ×1.55, ×1.41 — deliberately decelerating, because serving a large university would cost us almost exactly what serving a small one would. The $12,000 ceiling sits inside the $10,000–30,000 the one structural comparable, Overton, charges per institution on the same criterion — OpenAlex annual output — and near its floor, which is the right neighbourhood for a smaller corpus that has, today, evaluated no journals at all. We would rather be visibly cheap than defend a number we invented.

Nobody else in this market publishes an institutional price; librarians reconstruct what their peers pay through freedom-of-information requests. Crossref and DOAJ publish complete banded schedules in a plain table, and so does this page. The full argument, with every source, is in the pricing document behind this page, and every figure in it is checkable.

The lira list is a separate list, not a conversion, because Turkish law forbids pricing a service contract between residents in a foreign currency or indexing it to one. The intention is to fix it once a year at the edition freeze — published in July, in force from the following 1 January — so that a university can budget it in the previous year's allocation round. No lira list has been published or been in force yet: the figures in the table above are the first ones cut, and they are a proposal like every other number on this page — so the yearly rhythm just described is a plan, not yet a practice.

Its ceiling is set by procurement rather than by competition: every band, VAT included, stays under the 340,391 TRY direct-procurement limit for 2026, so no Turkish public buyer would need a tender. The upper-middle-income discount is already inside the lira figures and is not applied twice. The Organisation band is 280,000 TRY as well, because its most likely Turkish buyer is a public body under the same limit.

Nobody in Türkiye can pay us until the operating entity is registered and an e-Arşiv/e-Fatura route exists beside it.

• ✕ No number, ranking, profile, export or query route that a free reader cannot get. There is no gated content to gate.
• ✕ No priority in any evaluation queue, and no expedited or out-of-window evaluation, at any price, for anyone.
• ✕ No consideration of any kind for the institution's own journals or its university press. Buying a licence has no effect on how anything you publish is evaluated, and never will.
• ✕ No seat on the advisory board, no influence over the criteria, and no advance sight of an edition.
• ✕ No faster support than anyone else. What the licence would add is a remedy if we miss the target, not a shorter queue.

An IP range would never change what a page shows. It would change who we bill and who we thank. Everywhere else in this industry an IP range decides what a visitor may see; here that is impossible, because the metrics are CC BY 4.0 and data.html publishes the read-only key in full. Recognition would switch Premium-level service on for everyone on the campus network, give the institution a monthly usage count, let the page say “recognised via your institution”, and tell us who to thank in public. None of it is built: no range is matched today, and nothing here reads an IP address to decide anything.

• ✓ IP ranges with CIDR and IPv6, because tenders require them
• ✓ Institutional e-mail domain recognition — the door that works at home and on a phone, shipped at the same time, because IP matching fails silently
• ✓ An “are we recognised?” self-test page, on the development plan
• ✕ Federated single sign-on only when a real consortium asks for it by name

Matching ranges means processing IP addresses, so the privacy notice gains a section before any of this goes live: counts per range per month, no per-person profile, a stated retention period.

We are not COUNTER-audited, and we do not claim to be. COUNTER Release 5.1 compliance requires an independent audit we have not had. We will provide a plain monthly usage report and call it that. If five institutional customers ask for real COUNTER reporting by name, it goes on the roadmap.

We do not sell an availability or throughput guarantee yet. Our own load test saturated the shared database instance this site runs on at roughly 75 requests per second on author search, and one route already times out at the 8-second ceiling — measured and published. It becomes available when the metric workload moves to its own server.

We have no customers, no logo wall and no uptime record. There is no “most popular” badge on this page because nothing here is popular yet; every badge on a paid column says “proposed price”, which is the truth, and the one on the free column says what the free column is. What we can offer instead is the published formula behind every number, the errata feed, the open licence, the disclosed processor region and this price list. The repository deposit of each frozen edition under its own DOI is planned and not done, and no institutional licence has been published or signed either. Weigh us on the open licence, which is real and already granted; not on the deposit, and not on an agreement that does not exist.