Almost everything written about mentor-protégé programs is written for the protégé. The small firm learns how to find a mentor, what to ask for, and how to write the agreement. The mentor's side is treated as charity with a compliance benefit attached. That framing costs large primes real money, because a mentor-protégé relationship, structured deliberately, is three things a capture organization otherwise buys separately: an entity that can compete for set-aside work the prime cannot touch alone, credit against subcontracting commitments that the prime is already obligated to meet, and a specialist engineering team the mentor gets to shape over years rather than qualify in the two weeks before a proposal is due.
This is written for the person inside a large prime who owns that decision: a small business liaison officer, a strategic partnerships lead, a capture director who keeps losing small-business set-asides to teams they could have been on. The question is not whether to run a mentor-protégé program. Most large primes already do. The question is whether the program produces bidding capability or a spreadsheet.
Three things the mentor is buying
Separate them, because they have different economics and a given relationship rarely delivers all three well.
Access to a market the prime is barred from. A large business cannot bid a small-business set-aside. It can, through an approved SBA mentor-protégé joint venture, be part of an entity that does. Set-aside work is a large share of federal services and technology spending, and a prime that has written it off is writing off a portion of its addressable market. The joint venture is the legal mechanism; the mentor-protégé approval is what makes the joint venture eligible despite the mentor's size.
Credit that the prime owes anyway. Above the threshold, a large business receiving a contract with subcontracting possibilities has to submit a small business subcontracting plan, and then perform against it. Every prime knows the difference between a plan that is met and a plan that is met with substance. Development assistance provided under a mentor-protégé agreement, and subcontract dollars to a protégé doing real technical work, count in a way that a reseller pass-through does not.
A specialist bench the mentor helped build. This is the one that gets undersold. On AI, data engineering, cloud migration, and modern application work, primes routinely need a team that has actually put a model or a platform into production inside an agency, with the authorization paperwork done. Finding that team during a proposal is expensive and uncertain. Having a protégé whose engineers have worked inside the mentor's programs for two years, on the mentor's tooling, under the mentor's process, is a different asset entirely.
What a mentor gets back, weighted by how often it is the deciding reason
Editorial weighting, illustrative rather than measured. The last row is low on purpose: ownership complicates affiliation analysis and is rarely the reason a program starts.
Two programs, and they are not interchangeable
Primes conflate them constantly, and the conflation produces agreements that cannot do what the capture team assumed. The SBA All Small Mentor-Protégé Program and the Department of Defense Mentor-Protégé Program are separate instruments with separate purposes.
The SBA program is the one that makes joint ventures work. An SBA-approved mentor-protégé relationship allows a joint venture between the mentor and the protégé to compete for contracts set aside for the protégé's small business category, without the mentor's size disqualifying the venture through affiliation. That exception to affiliation is the point of the program. If a capture team wants to bid a set-aside with a large partner's past performance and resources behind it, this is the path.
The Department of Defense program is a development instrument, not a bidding instrument. It is oriented toward the mentor providing developmental assistance to the protégé, and it carries mechanisms through which a mentor may be reimbursed for, or credited with, the cost of that assistance. It does not, by itself, create joint venture eligibility for set-aside competition. A prime that signs a DoD agreement expecting to bid small-business set-asides has bought the wrong instrument.
Nothing prevents a mentor from having both, aimed at different outcomes. What is fatal is assuming one does the other's job. Before any agreement is drafted, write one sentence naming the pursuits the relationship is supposed to enable, and confirm the instrument reaches them.
| Dimension | SBA mentor-protégé | DoD mentor-protégé | Ordinary subcontract |
|---|---|---|---|
| What it primarily enables | A joint venture that can pursue set-aside work | Funded or credited development assistance to the protégé | Workshare on a contract the mentor already holds |
| Approval needed | SBA approval of the mentor-protégé agreement | Agency review and acceptance under the program | None beyond the prime's own subcontracts process |
| Set-aside eligibility | Yes, through the approved joint venture | No, not on its own | No |
| Where the mentor's benefit shows up | New pipeline the mentor could not otherwise bid | Reimbursement or credit for assistance provided | Delivery capacity and subcontracting performance |
| Typical commitment | Multi-year, with reporting on assistance delivered | Multi-year, with a documented development plan | Contract-length, renewed per task order |
| Common mistake | Signing it with a protégé who cannot carry workshare | Expecting it to open set-aside bidding | Calling it a partnership when it is a purchase order |
One more thing to check before drafting anything: what the protégé is eligible to be a protégé for, and whether the mentor already has agreements at the program's limit. Programs cap how many protégés a mentor may carry and how many mentors a protégé may have, and those limits move. Ask SBA or the agency office rather than relying on a memory of the rule from three years ago.
What a mentor actually commits
The written commitment is developmental assistance, and every prime that has run a program knows the failure mode: a signed agreement, a quarterly call, a slide with green boxes, and nothing that changed either firm. The assistance that matters is the assistance a capture team would want to have already happened when a proposal lands.
Access to real work, early. The single most valuable thing a mentor gives is a scoped piece of a live program, at a size the protégé can absorb, on a schedule that is real. Two engineers on a genuine task for six months teaches more, and proves more, than any number of training sessions. It also produces the thing the mentor needs most from the relationship: performance evidence written by the mentor's own program managers.
Process, tooling, and the mentor's way of working. Configuration management, the mentor's software factory and pipelines, how a change request moves, how a CDRL is produced and reviewed, how earned value reporting works if the program uses it. A protégé that already writes deliverables in the mentor's format is a protégé who costs the program office nothing to onboard.
Security and compliance uplift. Help getting the protégé's own environment to the control set the mentor's programs require, and help understanding what flows down and why. This is the assistance that most often turns a promising technical firm into one a subcontracts manager can approve quickly.
Business systems maturity. Accounting systems that can carry a cost-reimbursement subcontract, purchasing and property practices, indirect rate structure, and the internal controls a government audit expects. Technical firms are frequently strong at engineering and immature here, and this is the cheapest gap for a mentor to close.
Pipeline visibility. Telling the protégé what the mentor is chasing eighteen months out so the protégé can build the specific capability the pursuit needs. Withholding it and then complaining the protégé lacks the skill produces nothing.
How to pick a protégé whose engineering the mentor would put in a bid
Most protégé selection runs on socioeconomic category and personality. Both matter and neither predicts whether the firm can carry scored workshare. The selection question is narrower than it looks: would the capture team, with no program obligation at all, put this firm's technical approach in front of an evaluator and sign under it?
Ask for a system, not a capability statement. A capability statement is a marketing document; a system is a fact. Ask what they have deployed into a federal environment, who the government sponsor was, what the system does now, and what broke on the way there. Engineers who have shipped into an agency will answer with specifics about identity integration, data access approvals, authorization boundaries and the week the deployment slipped. Firms that have not will answer with methodology.
Look at the engineering surface directly. Ask to see how they structure a repository, how tests run, how a deployment happens from a clean checkout, how they handle secrets, and what their infrastructure looks like as code. Fifteen minutes of that tells a chief engineer more than a past performance volume.
Ask the data-rights question early. On defense work especially, what the government gets in the delivered software and data, and what stays with the developer, is decided by markings and by what was developed at private expense. A protégé that has thought about this, can explain its own position, and can produce an assertions table, is a protégé whose work will not blow up a delivery months later. A protégé that has never considered it will hand the mentor a problem.
Ask how they handle controlled unclassified information today. Not what they plan. What is in place: where data sits, who has access, how it is monitored, what happens on an incident, and whether the environment matches what flows down on the mentor's programs. Ask for the current assessment and the plan of action behind it. This is a checkable answer.
Confirm the boring things. Active registration, current representations and certifications, no active exclusions, insurance at the limits the mentor's subcontracts group requires, and financial statements that show the firm can carry payroll across a payment cycle on a cost-type subcontract. Most failed protégé relationships fail here, not on engineering.
Test appetite for the joint venture. A joint venture is more work than a subcontract and requires the protégé to take on managing responsibility and real accountability. Some firms want the credential and not the obligation. Ask directly whether they intend to be the managing venturer on a pursuit, and what their capacity looks like if the venture wins.
Protégé selection criteria, weighted by how well each predicts scored workshare
Editorial weighting, illustrative rather than measured. The last row is low because it is the criterion most programs actually select on.
The engineering questions that separate a real technical protégé from a résumé
This is where a mentor's chief engineer earns the program its return, and it is worth being specific about what to probe. Take an AI or data pursuit, since that is where most primes currently feel the gap.
How does the model get its data, and what happens when the source changes? A model in a demo reads a static extract. A model in production reads from a system that changes schema without telling anyone. Ask what their pipeline does when a column disappears: does it fail loudly, fail silently, or quietly start producing wrong answers. The right answer involves schema contracts, validation at ingest, and an alert that goes to a person. A firm that has not been burned by this will not have an answer.
What does the system do when the model is wrong? Every deployed model is wrong sometimes. The question is whether the surrounding software treats the output as a suggestion with a confidence and a human path, or as a fact. Ask what the interface shows the user, what gets logged, how a bad output is reported, and how that report becomes training data. This is the difference between a system an agency will actually run and a pilot that quietly gets switched off.
How is a model change released? Ask whether models are versioned as artifacts, whether the training data and code that produced a given model can be recovered, whether there is an evaluation set that gates promotion, and whether the government can roll back. Model governance frameworks in the federal space keep converging on the same requirement: know what version produced which decision, and be able to reproduce it.
Where does the evaluation number come from? If a firm quotes accuracy, ask for the dataset, the split, the date, and what a baseline scored on the same set. Numbers without those four things are decoration. An honest engineering firm will also tell you where the model performs badly and on which subpopulation.
What is the authorization path? Ask which boundary the system will live inside, whether the components are already authorized there, what the control inheritance looks like, and who writes the body of evidence. A firm that answers this fluently has done it; a firm that treats accreditation as somebody else's paperwork will cost the program six months.
How do they hand it over? Source, build pipeline, infrastructure as code, environment configuration, runbooks, and a rehearsal where the receiving team deploys while the original engineers watch. A protégé who does this by habit makes the mentor look good at the end of a program, which is when it matters most.
Structuring the relationship so it produces bids
The agreement is a compliance artifact. The operating rhythm is what determines whether anything comes of it. Programs that work share a small number of habits.
- Name the pursuits. The agreement should sit under a short list of specific opportunities, with dates, that the relationship exists to win. A development plan with no pursuit attached becomes a reporting exercise within two quarters.
- Put the protégé on live work inside ninety days. Small, scoped, real, and paid. The purpose is not the revenue. It is the performance record and the working relationship between engineers who will otherwise meet for the first time during a proposal sprint.
- Assign a named program person, not just the small business office. Development assistance delivered by the compliance function stays compliance. Assistance delivered by a program manager who needs the protégé to succeed on their own program produces capability.
- Give the protégé the capture calendar. What is coming, when the draft solicitation is expected, and what technical evidence the mentor will need. A protégé with eighteen months of notice can build the exact thing.
- Write the workshare percentage down before the bid. Ambiguity here is the most reliable way to end a partnership. Percentages, which scope elements, and what happens if the government cuts the award value.
- Decide the intellectual property and data-rights position at the start. What the protégé brings as background, what is developed under the subcontract, what markings apply, and what the government receives. Settle it once and reuse it.
- Review quarterly against outcomes, not activities. Bids submitted together, dollars flowed, capability added, evidence produced. Hours of training delivered is an input measure and it flatters everyone.
How we work inside a mentor's organization
Precision Federal is a small business engineering firm. We build AI systems, data platforms, cloud infrastructure, and full-stack web and mobile software, and we deliver them into production inside federal agencies. We work as a specialist subcontractor, a teaming partner, a protégé, and a nontraditional partner on other transaction agreements. We are not a staffing supplier, and the distinction matters to a mentor: we take a scope and answer for it.
What a mentor gets in the first weeks is concrete. In week one we read the solicitation or the program's statement of work and return a written technical position on the piece we would own, with the risks named. In the first two to three weeks we produce a scoped, priced statement of work with acceptance criteria written as tests rather than adjectives, and, where the pursuit is live, draft technical volume text the mentor's proposal manager can edit directly rather than rewrite. Where the work has started, our first increment lands in the mentor's environment on the mentor's pipeline, in the mentor's deliverable format.
What the mentor keeps is everything that matters to a prime. The customer relationship is the mentor's, and we do not go around it. The code, the models, the pipelines, and the documentation are delivered under the assignment terms in the subcontract, with our pre-existing tooling named and carved out and licensed back so nothing is stranded. Data-rights markings and the assertions table are settled before the first delivery, not argued about at closeout. On a proposal, we are visible or not visible as the capture strategy requires; where a named subcontractor with scored scope strengthens the technical volume, we will be named and stand behind the resumes, and where the mentor wants a single face to the customer, that is fine too.
Pricing takes one of two shapes. Fixed-price milestones against written acceptance criteria, where the scope is definable and the mentor wants cost certainty. Or a committed team at an agreed allocation for a stated period, where the program needs sustained capacity and the work will change shape. Both are quoted against a rate structure that supports the flow-downs the mentor's prime contract carries.
The first step is one email with a one-page brief: the program or pursuit, the technical scope in question, the environment the result must live in, the security destination, the date that matters, and the contract instrument. We return a scoped, priced statement of work. If the fit is not there, we say so in the same reply.
What goes wrong, and the tell for each
The agreement is signed and nothing follows. The tell is that no pursuit is named in the development plan. Fix it by attaching the relationship to two specific opportunities with dates.
The protégé cannot carry the workshare it was given. The tell is that the selection ran on category and rapport without a technical review by an engineer who would have to work with them. Fix it by putting the chief engineer in the selection.
The instrument does not do what the capture team assumed. The tell is a DoD agreement in the file and a set-aside on the bid board. Fix it by writing the intended pursuit type first and choosing the instrument second.
The relationship lives entirely in the small business office. The tell is that no program manager can name the protégé's engineers. Fix it by assigning a program owner who has skin in the protégé's performance.
Workshare is renegotiated after award. The tell is a teaming agreement with percentages described as approximate. Fix it before submission, in writing, including what happens when the award comes in smaller than the proposal.
Bottom line
A mentor-protégé relationship is a capture instrument, and it produces returns in proportion to how deliberately it is aimed. Pick the program that matches the pursuits: SBA when the point is joint venture eligibility for set-aside work, the Department of Defense program when the point is funded or credited development of a supplier the mentor needs. Pick the protégé the way a chief engineer would, on deployed systems and engineering practice rather than category and rapport. Give real work inside ninety days, share the pipeline eighteen months out, write the workshare and the data rights down before the bid, and review on outcomes. Done that way, the program stops being an obligation the prime reports on and becomes a channel into work the prime cannot otherwise bid, staffed by engineers the prime helped build.
Frequently asked questions
Three things. Eligibility to pursue small-business set-aside work through an SBA-approved mentor-protégé joint venture, which a large business cannot bid on its own. Substantive performance against the subcontracting commitments the prime already carries, delivered through a partner doing real technical work rather than pass-through volume. And a specialist engineering team the mentor shapes over years, working on the mentor's tooling and process, so the technical partner on a pursuit is qualified before the proposal rather than during it.
They do different jobs. The SBA All Small Mentor-Protégé Program is what makes a mentor-protégé joint venture eligible to compete for contracts set aside for the protégé's small business category, despite the mentor's size. The Department of Defense program is a development instrument centered on the mentor providing assistance to the protégé, with mechanisms for reimbursement or credit; it does not by itself create set-aside bidding eligibility. Choose the instrument after naming the pursuits it must reach.
Choose the way a chief engineer would. Ask what the firm has deployed into a federal environment, who sponsored it, what it does now, and what broke on the way. Look at the repository, the tests, the deployment from a clean checkout, and the infrastructure as code. Confirm how they handle controlled unclassified information today, what their data-rights position is, whether their business systems can carry a cost-type subcontract, and whether they actually want to be a managing venturer on a joint venture.
Developmental assistance, and the assistance that produces returns is specific. A scoped piece of live, paid work inside the first ninety days. The mentor's process, pipelines and deliverable formats. Help lifting the protégé's environment and compliance posture to what flows down on the mentor's programs. Business systems maturity so the firm can carry a cost-reimbursement subcontract. And pipeline visibility far enough ahead that the protégé can build the exact capability an upcoming pursuit needs.
In writing, before the proposal is submitted. State the percentage, name which scope elements the protégé owns, and state what happens if the government awards at a lower value or descopes after award. Joint ventures carry their own performance-of-work requirements that the arrangement has to satisfy, so confirm the split against those rules rather than negotiating it as a commercial matter. Ambiguity here ends more partnerships than performance problems do.