Procurement is the part of governance that happens before the vote

A DAO that needs an audit, a market maker, a treasury manager or an RPC endpoint has to buy it. The vote that approves the spend is the visible half; the invisible half is how the counterparty got in front of the DAO in the first place. Someone drafted a scope, someone decided who was allowed to bid, someone set the deadline, and someone scored the answers. That is procurement, and in most DAOs it is either undocumented or done by whoever happened to know a vendor.

The formal instrument for doing it in the open is the request for proposals: a published scope, a stated set of things a proposal must address, a submission window, and an evaluation the DAO commits to in advance. The service-provider page covers the shapes of the resulting relationship — the open programme, the exclusive mandate, the operating company. This page is about the call itself: what a DAO writes down, who answers, and what the exercise actually produces.

The honest finding across the record is that an open call is worth running even when it fails on its own terms. Lido's drew no public bids and the deal it was meant to source happened anyway. What the call bought was a written rubric, a public price, and an argument the community could have before the money moved rather than after.

The open call: Lido, 2022

Lido ran the cleanest small example. In February 2022 a contributor posted a temperature check asking whether the DAO should hire a market maker at all, noting that LDO was “virtually unlisted” on centralized exchanges while UNI, COMP, AAVE and SUSHI were everywhere. The sentiment check offered four options and closed with 57% for opening a call for market makers to put forward terms. Only then did the RFP go up.

The RFP itself is five headings long, and the headings are the interesting part because they are the DAO stating what it does not know: proposed deal terms (“possibly: treasury loan, embedded call options, length”), assurances about the firm's background, case studies of previous DAO engagements, whether the firm can assist with listings, and the success KPIs. Publishing that list did two things at once. It told bidders the DAO already understood the loan-plus-embedded-call structure and would ask about it, and it gave delegates a checklist to hold any eventual proposal against.

Then nothing happened. The thread ran to five posts, all from community members — one asking how it would affect the LDO price, one asking that other Lido assets be included in scope. No market maker posted terms in public. The instruction the original poster gave a fortnight earlier is worth quoting for what it assumed: receiving proposals is “non binding and as such would only bring free optionality to the DAO.” The optionality was free; it was also, in the event, empty.

The unsolicited bid: GSR and Olympus

Lido's own sentiment-check thread pointed at the precedent it was hoping to reproduce: market makers who “work via this process” and disclose specific deal terms in the open, with OlympusDAO cited as the example. That thread is instructive on its own. GSR posted a full market-making proposal for OHM directly to the Olympus forum: a loan of 25,000 OHM for liquidity provision, repayable either in kind at 0% interest or by buying out 12,500 OHM at $15,625,000 after six months and 12,500 OHM at $18,750,000 after twelve.

The forum then did the thing a private negotiation cannot. A reply decomposed the term sheet into what it actually was — a six-month call and a twelve-month call on OHM, granted with no premium paid — and asked GSR to revise the deal with real collateral or a premium. Nobody had to run a scoring rubric for that to happen. The proposal was legible because it was public, and being legible is what got it priced.

This is the counterweight to the tidy procurement story. An unsolicited bid posted where the whole DAO can read it produces most of the transparency benefit of a formal RFP, and produces it without the DAO having to run a process at all. What it does not produce is comparability: one bid scrutinised is not the same as four bids ranked.

The open tender: Arbitrum's procurement committee

Arbitrum DAO is the only DAO to date that has built a standing procurement function rather than running calls ad hoc. The Arbitrum DAO Procurement Committee (ADPC) was established by an on-chain vote and has since issued its own RFPs under a common framework, describing the first of them as the industry's first strategic-sourcing procurement framework.

The design borrows from public-sector purchasing and is worth naming precisely, because it is not an RFP in the one-off sense. The ADPC runs a panel: providers apply once to be whitelisted as Sellers against a head agreement with pre-agreed terms, conditions and pricing, and thereafter Buyers — project teams receiving subsidies, in the security case — order off the panel under a work-order template instead of negotiating from scratch. The Security Services Panel (June 2024) made that framework mandatory for anyone drawing on the security subsidy fund; the RPC Service Providers Panel (January 2025) ran the same machinery on a non-mandatory basis, with the framework noting that this could change if a subsidy programme were attached.

The DAO also uses the plain form. Its treasury management RFP (January 2025) put 25M ARB in scope with a specific brief — 10M ARB deployed in ARB-only on-chain strategies, 15M converted to stablecoins over three months with applicants required to state how they would minimise slippage and market impact — and took submissions on the public forum, where firms posted their strategies and got told off in-thread for exceeding the requested format. That is a different transparency trade from the panel: every bid is readable by every competitor.

Tender probity, written down

The most reusable artefact the ADPC produced is not a panel — it is the explanation it gave when a delegate asked why 3 of 12 applicants were cut. The committee declined to give per-applicant reasons, then set out the rules it had held itself to, and those rules are the transferable part:

• No target number. The committee had no pre-decided panel size and waited to see the applications as a whole before deciding — the cut came from the scoring producing “2 distinct groupings”, with a group of 9 stronger in aggregate, not from a quota.
• The deadline is the deadline. The overall date was extended once, for EthCC, and applied to everyone; no individual applicant was allowed to amend a submission afterwards, on the stated grounds that a deadline nobody enforces is not a deadline.
• Questions answered to the whole field. Applicant questions were answered to all applicants, anonymised, so asking a question carried no competitive cost — and the number or type of questions asked was explicitly excluded from scoring.
• Criteria published before submissions. The evaluation criteria were in the RFP, and the committee ran the comparison against them rather than against its own impressions.

The committee's own name for this is tender probity: avoiding favouritism and giving every applicant a level playing field. It is the same set of commitments a public buyer makes, and it exists for the same reason. A DAO that awards work without them has a procurement process that cannot be distinguished, from the outside, from a DAO that awards work to friends. Where that distinction collapses entirely, the failure belongs to the capture literature rather than the procurement one.

A panel is a barrier as well as a marketplace

Five months after the security panel closed, Oak Security posted in the same thread: clients wanted to build on Arbitrum and use the subsidy, but wanted to work with Oak, which was not whitelisted. Were there plans to add providers, or to run another RFP? The answer was that the subsidy fund was still in its pilot phase, that the committee would take the learnings, and that a process for adding providers was possible if the community wanted the fund expanded — conditional on the pilot succeeding and on the DAO's continued appetite.

Nothing in that exchange is bad faith. It is the structural cost of the panel model stated plainly. A whitelist attached to a subsidy is not just a convenience for buyers; it is a determination of who can be paid, and every provider outside it now has to win a governance argument rather than a commercial one. The one-off RFP has the opposite property: it is worse at repeat purchasing and better at staying open, because the next purchase reopens the field by default.

The same asymmetry runs through the renewal problem: scrutiny is loudest at the first award and quietest at every award after it. Procurement design is largely a choice about where to put the friction — at entry, where it filters, or at renewal, where it corrects.

What an open call actually surfaces

Set the three records side by side and the pattern is not “RFPs work.” It is that a call surfaces different things depending on how thick the supply side is.

• A thin market answers privately or not at all. Lido's call drew zero public bids from firms that were, demonstrably, willing to do the deal. Market makers will state terms in a forum when they are pitching an incumbent relationship and will not when they are one of several being compared.
• A thick market answers in public and stays comparable. Twelve security firms answered Arbitrum's tender; ENS DAO drew 26 applications requesting $12.2M against $3.25M available in its service-provider programme. Both fields were deep enough that a rubric had something to discriminate between.
• The rubric outlives the call. Lido's five headings and Arbitrum's evaluation criteria are reusable whether or not the specific call produced bids — and they are the part a DAO can be held to next time. What the resulting score decides once the bids are in is a separate question, and usually less than it looks: scoring and review gates.
• A refusal is information. An empty thread tells the DAO the terms it advertised were not attractive at that price, which is worth learning before a vote rather than after one.

None of this is free. Running a scored tender is real governance operating cost — a committee, a legal framework, an evaluation period measured in months — and the ADPC's frameworks required a paid committee and outside legal drafting to exist at all. The cheapest honest position is Olympus's: let bids arrive in public and let the forum price them. The most expensive is a standing panel. Most DAOs should be somewhere in between, and should at minimum write the rubric down.

How Caper approaches this

A caper has no procurement layer, because it has no standing organisation to procure for. There is no committee, no head agreement, no whitelist — a caper's treasury moves when a proposal passes and settles, and that is the whole apparatus. What that removes is the renewal problem: there is no incumbent contract quietly rolling over, because there is no contract.

Two design choices are relevant to the part of procurement DAOs find hardest, which is what a member can do when they lose the argument about a spend. First, proposing and voting cost fixed fees rather than being free, so contesting an award has a price and so does spamming one — and those fees are platform-wide rather than per-caper, held on the shared logic component and changeable only by the protocol admin badge, not by any individual caper's own governance.

Second, a member who has lost the argument can leave with a share rather than only sell. A caper's exit pays out a share of the treasury computed as (t · v) / (V · T) — the member's governance tokens times their votes cast, over total votes times circulating supply — which is the same formula that weighs their vote. Exiting requires surrendering both the governance tokens and the soulbound vote tokens. The practical consequence for the procurement question is narrow but real: a member who thinks a spend is capture does not have to win a renewal vote to stop funding it, and the share they leave with is priced off how much they participated, not how large their bag is.

References

• Lido Research — Market makers and CEX Listings (2 Feb 2022), the four-option sentiment check that authorised the call.
• Lido Research — Request for Proposals: Market Making (23 Feb 2022), the five-heading rubric; five posts, no vendor proposals.
• Olympus DAO Forum — Market Making Proposal from GSR, the unsolicited bid and the community's option-pricing objection.
• Arbitrum Forum — RFP: The ADPC Security Services Panel and Procurement Framework (19 Jun 2024), including the whitelisting update (9 of 12) and the tender-probity explanation.
• Arbitrum Forum — RFP: RPC Service Providers Panel and Procurement Framework (31 Jan 2025).
• Arbitrum Forum — [RFP Process] Request for Proposals: Treasury Management Services (22 Jan 2025), the 25M ARB scope.
• Arbitrum Forum — Request for Proposals: Treasury Management Services, submissions thread (6 Feb 2025).