The model that carried the last program was built by a subcontractor. Two years later the recompete is on the street, past performance is scored, and the capture team has to decide what to claim. Whose experience is it? The answer is not one answer. It depends on what the solicitation asks, what the prime managed, what the subcontractor did, and above all on decisions that should have been made before the work started and usually were not.
This is written for the capture director or proposal manager who has to write the volume, and for the program manager who will be asked to sign a reference. The mechanics are learnable and mostly unglamorous. Getting them right is worth several evaluation points on a scored factor, and getting them wrong produces either a claim an evaluator discounts or a good reference that nobody can obtain because the person who could sign it has moved.
Three different things that all get called past performance
Most of the confusion in this area comes from treating three separate records as one.
The government's evaluation of the prime contract. Recorded in the contractor performance assessment reporting system, written by government personnel about the contract, and attached to the prime's record. It rates quality, schedule, cost control, management and small business participation. It is about the contract holder. A subcontractor does not receive a rating of its own out of that system.
What a prime may claim in a proposal. An offeror describes contracts it holds. It may also, where the solicitation permits, present the experience of subcontractors and other team members, with the evaluator deciding how much weight that carries given the role the team member will play on the new effort.
What a subcontractor may claim in its own proposals. This comes from references: a letter or a completed past performance questionnaire from a named person at the prime or, sometimes, at the government customer, describing what the subcontractor did, over what period, at what value, and how it performed.
These three are produced by different mechanisms and on different schedules. A prime that has all three lined up on a program has an asset it can use for years. A prime that has only the first has a record about itself and no way to prove which parts of the technical work its team can repeat.
What makes a subcontractor past performance citation carry weight with an evaluator
Editorial weighting, illustrative rather than measured. The last row is deliberately low: a number without a scope tells an evaluator nothing about capability.
What an evaluator is actually doing
Past performance evaluation is a prediction, not a reward. The evaluator is asking whether this offeror, with this team, is likely to perform this requirement well. Everything about how a citation is written should serve that question.
Three tests run in the evaluator's head, usually in this order.
Is it relevant? Relevance is judged on similarity of scope, magnitude and complexity. A subcontract with a modest dollar value on a large, complex program can be highly relevant if the technical scope matches, and a large contract with different work can be barely relevant at all. This is why the citation should describe the technical environment and the scale of the system, not only the subcontract value.
Is the experience coming with the team? An evaluator who sees a strong technical citation attached to a team member with a small role on the new effort will discount it, and should. The citation is only predictive to the extent that the same organization, ideally with some of the same people, holds the same kind of scope now. Match the citation to the workshare, and say plainly in the volume what portion of the new effort that team member performs.
Is it confirmable? Evaluators verify. They read the government's own record on the contract, they send questionnaires, and sometimes they call. A citation that describes work in language nobody at the customer would recognize is worse than no citation. Write it so a government technical lead reading it would nod.
How to write the reference so both parties benefit
The reference document is where most of the value is created or lost. A weak reference is not one that says something bad; it is one that says nothing specific. The pattern that works has five parts.
The setting, stated in technical terms. The program's purpose, its scale, the number of sites or users if that is describable, the technologies in the environment, and the security destination. This is what carries relevance. "Supported a data modernization effort" carries none.
The scope the subcontractor owned, drawn as a boundary. What the team was responsible for, expressed as deliverables and interfaces, and what it was not. Ambiguity here is what lets a competitor argue in a debrief that the citation overstates the role.
The outcome with a measurement in it. A baseline and a result, or a schedule commitment and what happened against it. Any number that can be stated without disclosing anything sensitive is worth more than a paragraph of praise. If a number cannot be released, state the change in factual terms: a nightly load that previously failed weekly now runs without operator intervention; a review cycle that took months completed in one pass.
How the team performed as a party to manage. Responsiveness, schedule discipline, how problems were surfaced, whether the team worked cleanly inside the program's reporting and configuration practice. Evaluators care about this because it predicts the cost of managing the team.
The signer's basis of knowledge. A named person, their role on the program, and the period over which they observed the work. A reference from someone who supervised the work directly is worth several from people who heard it went well.
The prime benefits from writing it this way for a plain reason. When the prime later claims that team member's experience in its own volume, the evaluator will read the same specifics from the other side. Two documents describing the same work in the same concrete terms is a stronger file than a general assertion and a general endorsement.
Set it up before the work starts
Everything above is easy at the beginning of a program and hard at the end. The people who could sign move to other programs. The metrics that would prove the improvement were never baselined. The scope in the subcontract says "technical support services," which is unusable.
Six things belong in the subcontract or the teaming agreement.
- A scope statement written as deliverables and interfaces. If the subcontract scope reads as a labor category and a level of effort, it will produce a citation that reads the same way. Write the scope the way you would want the reference to read.
- A reference commitment with a name and a clock. The role that will sign, a commitment to complete a standard past performance questionnaire within a stated number of business days of a request, and an obligation that survives the end of the period of performance. Attach a duration, because references become hard to get after a reorganization.
- Permission to describe the work. An express statement that the subcontractor may describe the scope of its own work in proposals and qualifications, subject to nondisclosure terms and customer sensitivity. Without it, a nondisclosure clause read strictly can bar the very citation both parties want.
- A baseline measurement obligation. Record the state of the metric the workstream will be judged on before the team starts, and report it in the monthly. This costs one hour at kickoff and is impossible to reconstruct later.
- A reciprocal commitment. The subcontractor should be obliged to supply the prime with performance narrative and data on request as well, for use in the prime's own proposals and in its performance evaluation responses.
- A point of contact who is not the one at risk of leaving. Name a role, not only a person, and keep a copy of the reference material in the program file so the prime can produce it even after the individual is gone.
| Record | Who it is about | Where it comes from | How to secure it |
|---|---|---|---|
| Government performance evaluation | The prime contract holder | Government assessing officials on the contract | Perform, respond to the draft evaluation carefully, keep the narrative factual |
| Prime's claim of team experience | The team as proposed | The prime's own volume, where the solicitation permits it | Match every citation to the team member's workshare on the new effort |
| Subcontractor reference letter | The subcontractor's scope | A named individual at the prime with direct knowledge | Commit the signer role and a response clock in the subcontract |
| Past performance questionnaire | Whoever the questionnaire names | Completed by a reference on the evaluator's form | Keep the reference's contact current and warn them before a submission |
| Measured technical outcome | The work itself | The program's own monthly reporting | Baseline before the team starts; record the delta each month |
| Customer recognition | Individuals or the team | The government program office | Ask at the time it happens and file it; it cannot be recreated |
Common ways the citation goes wrong
The scope is described in contract language rather than technical language. A citation that says the team provided analytic support services under a task order tells an evaluator nothing that predicts performance. What predicts performance is the system, the data volumes, the interfaces, the deployment target, and what changed as a result.
The citation and the workshare do not match. A team member cited for building an analytics platform, then proposed for five percent of a new effort with no analytics scope, weakens the volume rather than strengthening it, because it invites the evaluator to conclude the team is assembled for its resumes.
The reference cannot be reached. The named individual has changed employers, the questionnaire goes to a shared mailbox, and no response arrives before the due date. Naming a role as well as a person, and keeping the material in the program file, prevents this.
The outcome is stated in adjectives. Successful, effective and high quality are the words that appear in every citation an evaluator reads that day. A recorded baseline and a recorded result are what separate one from the rest.
Nondisclosure terms are read to bar the description. Both parties intended the citation to happen and neither wrote the permission down, so counsel takes the conservative reading and the work becomes uncitable. One sentence in the subcontract avoids this entirely.
Where subcontractor citations lose their value, ranked by how often it happens
Editorial weighting, illustrative rather than measured. The last row is deliberately low: most lost citations describe work that went fine.
The technical specifics that make an AI citation credible
Citations for model and data work fail differently from citations for ordinary software. The reason is that an evaluator with any technical depth has seen claims that did not survive contact with production, so the credible citation is the one that shows the writer knows where such systems break.
Include the things a knowledgeable reader looks for.
How the evaluation was constructed. Say that a held-out set was reserved and that the partition rule prevented records sharing a subject, a site or a time window from appearing on both sides. This is the single most common source of results that look excellent and then collapse, and naming it signals competence better than any adjective.
What was stored for every decision. The input record, the model version, the feature values, and the threshold in force. Systems whose outputs a federal user acts on need this, and describing it tells an evaluator the team has built something a government user could actually rely on.
What monitoring shipped with the system. Input distribution and outcome rate monitoring, with thresholds and a named owner. It is the difference between a delivered capability and a delivered demonstration.
How the system was authorized. The environment it runs in, the control evidence produced, and the review it passed. Working to the NIST SP 800-53 control set and the FedRAMP baseline appropriate to the destination is worth stating plainly, because the ability to get a system through review is often the scarce skill.
What the receiving organization got at the end. Infrastructure as code, a build pipeline, a runbook, and a handover in which their team deployed while the builders watched. An evaluator reading that knows the capability survived the contract.
How we work with primes on this
Precision Federal builds AI systems, data platforms, cloud infrastructure and full-stack web and mobile applications, and delivers them into production inside federal agencies. We are a small business and we work as a specialist subcontractor and teaming partner to large primes. We treat the citation record as part of the deliverable, and we set it up at kickoff rather than at the end.
In the first weeks that means three concrete things alongside the engineering. We agree the scope statement in deliverable and interface terms, so it can be described later without argument. We baseline the metrics the workstream will be judged on before we change anything, and we report them in the monthly in a form the prime can quote. And we agree in writing who signs a reference, within how many days, and what each party may say about the work.
The prime keeps the customer relationship and the contract, and everything we build is delivered under the subcontract terms, with our pre-existing tooling named as background and licensed so nothing in the delivered system is blocked for a future maintainer. We supply the prime with performance narrative and technical detail for its own volumes and evaluation responses on request, without being chased for it.
Pricing takes one of two shapes. A bounded increment prices as a firm fixed-price milestone against written acceptance criteria, which puts schedule and technical risk on us. A continuing workstream prices as a committed team at a stated allocation, with named engineers and a substitution path in the subcontract.
The first step is one email with a one-page brief: the program or the pursuit, the workstream, the environment named by product, what data exists and who grants access, the security destination, and the date that matters. We return a scoped, priced statement of work.
Bottom line
Nobody wins a scored past performance factor with a subcontract value and a satisfactory adjective. The evaluator is predicting whether this team performs this requirement, so the citation has to show matched scope, matched technical environment, a measured outcome, and the same organization holding the same role on the new effort. All of that is produced by decisions made at kickoff: write the subcontract scope in deliverables, baseline the metric before the work starts, commit the reference signer and the response clock in writing, and grant each party permission to describe its own work. Primes that do this on every subcontract accumulate a file they can bid from. Primes that do not spend the week before a proposal is due looking for someone who still remembers the program.
Frequently asked questions
Generally yes where the solicitation permits presenting the experience of team members, and the weight depends on the role that team member will hold on the new effort. An evaluator discounts a strong technical citation attached to a partner with a minor role, because past performance is a prediction about how this team will perform this work. Match each citation to the proposed workshare and state plainly in the volume what portion of the effort that team member performs.
Not from the government's performance assessment system, which records evaluations about the prime contract and attaches them to the contract holder. A subcontractor's usable record comes from references: a letter or a completed past performance questionnaire from a named person with direct knowledge, describing scope, period, value, technical environment and how the work was performed. That is why the reference commitment belongs in the subcontract at the start rather than in an email at the end.
Similarity of scope, magnitude and complexity to the new requirement. A modest subcontract on a large program can be highly relevant when the technical environment matches, and a large contract for different work can be barely relevant. So the citation should describe the system, its scale, the technologies involved and the security destination, not only the dollar value. Write it so a government technical lead who knew the program would recognize the description.
Four things. A scope statement written as deliverables and interfaces rather than labor categories. A named signing role and a commitment to complete a questionnaire within a stated number of business days, surviving the end of performance. Express permission for each party to describe the scope of its own work in proposals, subject to nondisclosure and customer sensitivity. And an obligation to baseline the workstream's metric before the team starts, so an outcome can be stated in numbers later.
By showing the writer knows where such systems fail. Name how the evaluation was partitioned so records sharing a subject, site or time window did not appear on both sides. Say what was stored for every decision: input record, model version, feature values, threshold in force. Say what monitoring shipped with the system and who owns it. Say how the system was authorized and what the receiving team got at handover. Specifics of that kind read as capability; adjectives do not.