The OTA prototype-to-production transition

Where a prototype becomes production

Every prototype other transaction ends one of two ways. The government buys the capability, or the work stops and the file closes. The provision that decides which one happens is 10 U.S.C. § 4022(f), and by the time anyone reads it the outcome is usually already set — by the acceptance criteria in the agreement, by whether a program office has production money in a budget year that has not arrived yet, and by two sentences most firms never negotiate.

The scale of the gap is public. GAO reported that DoD obligations through other transactions grew from $1.8 billion in fiscal year 2016 to over $18 billion in fiscal year 2024. Of that fiscal 2024 total, over $16 billion was prototype work and about $2 billion was production. GAO's finding in the same report is the more useful one: DoD does not know the extent to which prototype other transactions directly resulted in production awards, and GAO recommended the department build a systematic way to track it.

Enormous prototype spending, a thin production tail, and no measurement of the connection between them.

What section 4022(f) says now

The operative language is short. Under § 4022(f)(2), a follow-on production contract or transaction may be awarded to the participants in the transaction without the use of competitive procedures, notwithstanding the competition requirements of chapter 221 of title 10, and — in the statute's own words — "even if explicit notification was not listed within the request for proposal for the transaction," if two conditions are met: competitive procedures were used to select the parties for participation, and the participants successfully completed the prototype project.

That "even if" clause is newer than most commentary on this authority. Congress added it after the leading protest decision on the subject, which turned on exactly that point.

The FY2025 NDAA, Public Law 118-159, moved the section again. Section 816 lowered the approval level for large prototype transactions: a prototype other transaction expected to cost more than $100 million but not more than $500 million now rests on a written determination by the head of the contracting activity — or by the agency director for DARPA, the Defense Innovation Unit, and the Missile Defense Agency — rather than the senior procurement executive, who now handles the tier above $500 million with 30 days' notice to the congressional defense committees. Section 817 rewrote the follow-on provision itself.

Three additions matter to anyone working under a consortium. Paragraph (f)(1) now states that a transaction includes all individual prototype subprojects awarded under it to a consortium of United States industry and academic institutions. Paragraph (f)(3) allows a follow-on award when the department determines an individual prototype or subproject was successfully completed. Paragraph (f)(4) states that such an award is not contingent on successful completion of all activities within that consortium. Before those paragraphs, a firm whose own subproject succeeded could be held hostage by unrelated work under the same base agreement.

Paragraph (f)(5) is the sleeper. It says the follow-on may be awarded using the other transaction authority in subsection (a), under chapter 137 of title 10 — ordinary procurement contracts — or under procedures the Secretary of Defense establishes by regulation. The government picks the instrument. The firm does not.

The two conditions that decide, and the one that fails

Competitive procedures for the original selection is the condition firms rarely lose. A consortium request or a commercial solutions opening with a real field of offerors checks that box permanently.

Successful completion is the condition that fails, and the case law shows how. In Oracle America, Inc., B-416061 (May 31, 2018), GAO reviewed the Army's award of a follow-on production other transaction to REAN Cloud with a not-to-exceed value of $950 million. GAO sustained the protest on two independent grounds. First, the prototype agreement contained no provision for a follow-on production award, and GAO rejected the argument that loose solicitation language about possible follow-on production was enough. Second — and this is the ground that still bites — the prototype project had not been successfully completed. The Army had modified the prototype agreement to add enclave migration work, and that work was unfinished when the production award issued. Completing "parts" of a project was not completing the project.

Congress has since overtaken the first ground. Nothing has overtaken the second. A firm that treats "successfully completed" as a formality is relying on a government determination made against criteria it never wrote.

Measurable acceptance criteria are the whole defense here. A detection rate at a stated false-alarm rate. A latency budget under a stated load. A completed integration against a named system, verified by a named test. Language about performing to the government's satisfaction transfers the entire question to whoever occupies the chair at the end of the period of performance.

The mirror-image risk is scope drift. The Army's agreement grew mid-performance and the completion standard grew with it. Every modification that adds work moves the finish line, so treat each one as a moment to restate in writing what completion means for the modified agreement.

Where the regulation and the statute diverge

DFARS 206.001-70 supplies the competition exception on the FAR side. Current through the May 7, 2026 change, it excepts follow-on production contracts for products developed under § 4022 prototype authority when the contracting officer receives sufficient documentation from the agreements officer showing that the other transaction solicitation and agreement included provisions for a follow-on production contract, and that the applicable thresholds and the requirements at § 4022(f)(2)(A) and (B) were met.

Read that against the statute, which now permits the award even where explicit notification was absent from the request for proposals. The two are not aligned.

The practical consequence runs one direction only. Never rely on the statutory relaxation to excuse a missing clause. A contracting officer applying the regulation as written will look for the provision, will not find it, and the file will stall while lawyers argue about which text governs. Get the provision into the agreement; it costs nothing at negotiation.

What the public numbers show

Four measured figures frame the transition better than any narrative. GAO's September 3, 2025 report supplies three: the fiscal 2024 split between prototype and production dollars, and the finding that of 18 selected weapon systems using prototype other transactions, ten planned to switch to standard contracts for production. GAO's February 27, 2025 review of the Defense Innovation Unit supplies the fourth — DIU made 450 prototype awards from fiscal years 2016 through 2023 and reported that 51 percent of completed prototypes transitioned to production.

A completed prototype that reaches production is roughly a coin flip even at the office with the strongest published record, and the production award is more often a standard contract than another other transaction. Anyone planning a business around automatic conversion is planning around the wrong number.

The appropriation is the real gate

Statutory eligibility is necessary and never sufficient. The money has to exist, in the right form, in the right year.

The purpose statute, 31 U.S.C. § 1301(a), says appropriations shall be applied only to the objects for which the appropriations were made. Prototype work is normally research, development, test and evaluation money. Buying production quantities normally takes procurement money. These are different appropriations with different rules, and no amount of enthusiasm from a program office converts one into the other.

Timing compounds it. The Congressional Research Service describes the Planning, Programming, Budgeting, and Execution process as calendar-driven and typically beginning more than two years before the expected year of budget execution. A sponsor who decides mid-prototype that production should follow is deciding about money two budget cycles out, and the prototype ends long before that money exists. That interval is where capabilities die, and it is not a contracting problem at all.

Production money that exists for this problem

One program is built specifically for the seam. APFIT — Accelerate the Procurement and Fielding of Innovative Technologies — started in fiscal year 2022 and provides procurement funding for innovative projects that have finished development and are ready to transition into operational use. Awards run between $10 million and $50 million per project. The program office reports over $1.4 billion awarded to more than 75 companies through fiscal year 2025.

The eligibility rules are specific. The prime vendor and majority awardee must be U.S. based and must be a small business, a nontraditional defense contractor, or an entity whose aggregate defense obligations over the preceding year fall under a stated ceiling. Funding is strictly for procurement and cannot be used for anything requiring further research and development after award, so all such work should be finished by the end of the selection process. Early-stage projects are excluded.

The mechanic that surprises firms most: proposals are submitted by a government representative or sponsor to the relevant organization lead, not by the company. The program does not award contracts either — the acquisition is negotiated separately between the company and the responsible contracting activity. The real work is finding a sponsor who will carry the submission, and that work starts during the prototype.

Service-level transition funds sit alongside it. The Air Force runs strategic and tactical funding increases that extend Phase II work with matched outside money — different rules, same purpose.

The follow-on is often a FAR contract, not an OT

Section 4022(f)(5) lets the government award the follow-on under other transaction authority, under ordinary procurement authority, or under regulation. GAO's finding that ten of eighteen reviewed weapon systems planned standard contracts for production tells you which way the department leans on larger efforts.

This is the part of the transition that changes a firm's operations most, and almost nobody plans for it. A prototype other transaction can run on ordinary commercial books, with payment against demonstrated milestones. A FAR production contract brings back the machinery the other transaction set aside: FAR Part 31 cost principles where the contract type calls for them, Cost Accounting Standards coverage once the thresholds are met, certified cost or pricing data where applicable, DCAA involvement, and standard DFARS data rights clauses attaching by operation of the contract.

An accounting system adequate for accumulating costs by contract line is not something a firm assembles in the six weeks between a completion determination and a production award. Build toward it during the prototype, when there is time and no deadline attached.

Data rights across the seam

Intellectual property is negotiated fresh in an other transaction because the FAR clauses that would supply defaults are absent. What firms miss is that the negotiation happens twice: once at the prototype, and again at whatever instrument carries production. The second one inherits whatever the first established, in front of a contracting officer used to standard clauses.

Attach the background IP schedule to both instruments. Anything unlisted at the prototype is harder to claim at production, when the government has been using it for a year.

License per deliverable, not per project. Government purpose rights in an integration layer with restricted rights in a core model is a normal, negotiable structure. A project-wide grant made at prototype speed governs the production award too.

Name the open-source and third-party components and their licenses. A copyleft dependency found inside a delivered container after a production award is a worse week than the same discovery during a prototype.

Treat data as its own asset class. Training data, fine-tuned weights, evaluation sets, and prompt libraries carry independent value, and a clause drafted around "technical data" often fails to say which of them it reaches.

Where prior SBIR work sits underneath the capability, the position is stronger and travels differently. Under the SBIR and STTR policy directive, a funding agreement that derives from, extends, or completes work performed under a prior SBIR or STTR award carries Phase III status and SBIR data rights by nature, whoever funds it. Separately, 15 U.S.C. § 638(r)(4) directs agencies to issue Phase III awards, including sole-source awards, to the firms that developed the technology, and FAR 6.302-5(b)(4) supplies the competition authority. There is no ceiling, and the awarding agency need not be the one that funded the earlier phases. For a firm with SBIR history, Phase III and § 4022(f) are two doors into the same room.

Commercial solutions openings after the FY2026 NDAA

The FY2026 National Defense Authorization Act, Public Law 119-60, signed December 18, 2025, reworked the commercial solutions opening authority at 10 U.S.C. § 3458 in section 1823. The word "innovative" came out, so a commercial solutions opening now reaches commercial products, commercial services, and nondevelopmental items — a narrow innovation instrument turned into a general commercial acquisition tool. The review requirement loosened: a solicitation may be paired with a peer review, a technical review, or an operational review, whichever fits. And the section now carries explicit follow-on production authority, including sole-source awards, out of a competitively conducted opening.

That last change is a second path to production, parallel to § 4022(f) and no longer dependent on the prototype framing. Firms selling something already commercial, and awkward fits for "prototype" language, have a cleaner story to tell.

The demand signal points the same way. In a March 6, 2025 memorandum, the Secretary of Defense directed components to adopt the software acquisition pathway as the preferred pathway for all software development components of business and weapon system programs, and to use commercial solutions openings and other transactions as the default solicitation and award approaches under it.

Protest exposure at the seam

The forum picture changed materially in 2025, and it changed at the transition point specifically. GAO's position remains narrow: because an other transaction is not a procurement contract, GAO generally declines to review the award, while agreeing to consider whether an agency improperly used a non-procurement instrument to procure goods or services — the door the 2018 Oracle decision walked through to reach the follow-on production award.

At the Court of Federal Claims, Raytheon Co. v. United States, No. 24-1824C (Fed. Cl. Feb. 24, 2025), had Judge Bonilla describe the court as the de facto forum for bid protests involving other transaction agreements, distinguishing an earlier decision on the ground that the products at issue were government-specific rather than broadly commercial. Then in Telesto Group, LLC v. United States, decided June 2, 2025, Judge Hertling drew a phased map: jurisdiction at the outset of an other transaction and again once the prototyping phase concludes and the agency decides to proceed with a proposed procurement, but not over discretionary agency conduct during performance of the prototyping phase itself.

Read together, the two decisions say something counterintuitive. The prototype performance period is the least reviewable stretch of the arrangement; the transition decision at the end is the most reviewable. A firm that loses a down-select mid-prototype has thin recourse. The terms that protect a firm during performance therefore have to be in the agreement, because the courthouse is closed during that stretch.

The instrument choice, side by side

The sequence, and where it stalls

Step five is where efforts die, and it is the only step a firm cannot execute itself. Everything before it is preparation for a conversation somebody else must have inside a budget process.

Terms to fix before the prototype starts

• A follow-on production provision, stated plainly, even though the statute no longer strictly requires the notification
• Completion criteria expressed as measurable, testable results with named thresholds and named test conditions
• A background IP schedule listing every pre-existing model, library, dataset, and component
• Per-deliverable license grants rather than one project-wide grant
• Named third-party and open-source components with their licenses
• Explicit treatment of training data, weights, evaluation sets, and prompt libraries as separate assets
• The sponsor's answer, in writing if possible, on which appropriation and which fiscal year buys production

None are exotic asks. All are ordinary at negotiation and nearly impossible to obtain afterward.

Common objections, answered plainly

Bottom line

The transition is not a contracting event. Three things have to line up at once: a documented determination that the prototype was successfully completed against criteria written down in advance, an appropriation of the right kind in the right year, and a sponsor willing to spend political capital on the instrument. The statute has gotten steadily friendlier — the notification precondition relaxed, consortium subprojects carved out, approval levels pushed downward, a parallel path opened through commercial solutions openings. None of that supplies the money or the sponsor.

What a firm controls is the first item and its own readiness for the third. Write the completion criteria. Keep the follow-on provision in the agreement. Build the accounting capability during the prototype. Ask which appropriation buys production while there is still time for the answer to change something. The rest is a program office deciding whether it wants the capability badly enough to fight a budget cycle for it, and no instrument creates that.

Frequently asked questions