The problem a pipeline solves
By the time a requirement appears on SAM.gov, the government has usually finished its market research under FAR Part 10, spoken with somewhere between three and a dozen companies, and written an acquisition strategy shaped by what those companies said they could do. A firm that first learns about a requirement from the posted solicitation is bidding into a field where somebody else already helped set the shape of the work. That is the gap a capture pipeline closes. What follows is how we run one without a business-development department, and what the numbers honestly look like when you do.
Capture is often taught as a heavy discipline with color-team reviews, gate meetings, and a full-time capture manager per pursuit. That model is built for a firm chasing three $40M recompetes a year. A small technical firm chasing $150K to $3M actions has a different problem: too many candidate opportunities, too little time per opportunity, and no tolerance for spending forty hours on something that was never winnable. The pipeline has to be cheap to run and ruthless about killing pursuits early.
So the design goal is not a bigger process. It is a smaller decision cost. Every pursuit should be killable in twenty minutes with public information, and every pursuit that survives should get a specific, budgeted amount of pre-solicitation work assigned to a named person on the team. Everything below serves those two rules.
Sourcing channels, ranked by what they actually produce
Signal quality by sourcing channel
Editorial weighting of channel signal quality from public sources and practitioner reading. Illustrative, not a measured statistic.
SAM.gov Contract Opportunities. Under FAR 5.201(b)(1), contracting officers must transmit a notice to the governmentwide point of entry for each proposed contract action expected to exceed $25,000, so SAM.gov is the floor of any pipeline. Saved searches with email alerts, filtered by NAICS (541511, 541512, 541519, 541715, 518210 for software and data work), set-aside type, and place of performance, cost nothing to maintain. The Contract Opportunities API supports the same filters if you would rather pull JSON into your own tracker than read email. Treat these hits as raw material, not as a pipeline. Most of what arrives this way is already shaped around an incumbent.
Agency acquisition forecasts. Federal agencies are required to publish forecasts of expected contract opportunities that small businesses are capable of performing, a requirement that traces to Section 501 of Public Law 100-656. The forecasts vary in quality, but they routinely give you a description, an estimated dollar range, an anticipated set-aside, a target quarter, and a point of contact, six to eighteen months before anything posts. This is the single highest-value public data source for a small firm, and almost nobody reads it. Our team pulls the forecasts for the agencies we serve every quarter and diffs them against the prior pull, because changes tell you what moved, slipped, or got re-scoped.
Small business offices. Every agency has an Office of Small and Disadvantaged Business Utilization or Office of Small Business Programs under 15 U.S.C. 644(k). Their job is literally to connect capable small firms to requirements. A two-paragraph email with a capability statement, a specific NAICS, and a specific mission problem you already work on gets answered far more often than a cold note to a contracting officer.
Research-program pre-release windows. The SBIR and STTR programs are unusual in that they publish the technical requirement weeks before the proposal window opens, and during that pre-release period proposers may contact the topic author directly. Once the solicitation opens, questions move to a public written channel and the informal conversation ends. That pre-release window is the cleanest legal opportunity in federal contracting to talk with the person who wrote the requirement, and it is the reason research-program work belongs in a small firm's pipeline even when the dollars per action are modest.
Primes. Any prime contract expected to exceed $750,000 with subcontracting possibilities requires a subcontracting plan under FAR 19.702, with goals the prime reports against in eSRS. That reporting obligation is a standing reason for primes to look for capable small subcontractors. Subcontracting opportunities that used to live on SBA's SubNet now appear inside SAM.gov, and most large primes also run their own supplier portals. Inbound teaming requests convert better than almost anything else in the pipeline because the prime has already decided it needs what you do.
What the public feeds do not show you
Two large categories of federal spending never appear as a fresh public solicitation, and a pipeline built only on SAM.gov alerts will be blind to both.
The first is task orders under existing multiple-award vehicles. Once an IDIQ, GWAC, or agency vehicle is in place, individual orders are competed among the holders under the fair-opportunity procedures of FAR 16.505(b). The notice goes to the vehicle's holders. If you are not on the vehicle, you learn about the work when the award shows up in the public award data, which is too late to bid and exactly right for planning the next cycle. The second is GSA Multiple Award Schedule buys, which are quoted through GSA eBuy rather than the general opportunity feed, with orders above the simplified acquisition threshold posted there under FAR 8.405-2(c).
The practical consequence for a small firm is that vehicle access and teaming are not separate from capture. They are capture. A realistic pipeline has a lane for "work we can bid today," a lane for "work we can only reach as a subcontractor," and a lane for "vehicle or qualification we need in order to reach a class of work next year." Confusing the three produces a pipeline that looks full and closes nothing.
Qualifying an opportunity in twenty minutes
The point of fast qualification is to spend your capture hours on the small number of pursuits that deserve them. Everything on this list is answerable from public records in under twenty minutes.
- Who holds the work now, and for how much? Look up the incumbent and the obligated dollars in the public award data. A single incumbent with eight years and clean option exercises is a different bid than a first-time requirement.
- Is it set aside, and for what? Acquisitions above the $10,000 micro-purchase threshold and at or below the $250,000 simplified acquisition threshold are reserved for small business under FAR 19.502-2(a) unless the contracting officer finds no reasonable expectation of two competitive offers.
- What NAICS code, and what size standard? The code drives the size standard, the small-business determination, and often the evaluation posture.
- Is there a named technical point of contact? A requirement with a reachable technical owner is worth ten times a requirement routed through a shared mailbox.
- Has a sources-sought or RFI already run? If it ran and you missed it, the acquisition strategy is largely set.
- What does the work actually require on day one? A clearance, a facility, an existing ATO, or a vehicle you do not hold is a structural blocker, not a stretch goal.
- What evidence would we show? If you cannot name the artifact that proves you can do the work, the proposal will not be able to either.
- What is the realistic close date? Anything with a proposal due inside ten business days that you have not been tracking is a low-probability bid.
A bid/no-bid rubric that holds up
Scoring is only useful if the scale is anchored to observable facts. The version below uses six weighted factors on a one-to-five scale. Multiply, sum, and compare against a fixed threshold you set once and do not move to justify a bid you already wanted to make.
| Factor | Weight | Scores 5 when | Scores 1 when |
|---|---|---|---|
| Customer contact | 20% | We have spoken with the technical owner about this requirement | We know no one and the notice is already open |
| Technical fit | 20% | The core of the work is something our engineers already build | We would be learning the domain on the government's clock |
| Evidence we can show | 15% | A working system, measured results, or a named reference applies directly | We can describe an approach and nothing more |
| Competitive field | 15% | Set-aside, few credible bidders, no entrenched incumbent | Full and open, incumbent on option year three, ten likely offerors |
| Contract mechanics | 15% | We can hold the prime position, or the teaming path is already agreed | Requires a vehicle, clearance, or facility we do not hold |
| Cost of the bid | 15% | Under 40 hours of effort against a multi-year ceiling | Over 120 hours against a one-year action below $200K |
Three hard floors sit above the arithmetic. A pursuit with a structural blocker in the contract mechanics row is a no-bid regardless of total score, because a good score on the other five factors cannot manufacture a clearance. A pursuit where we would score a one on technical fit is a no-bid even at attractive dollars, since the delivery risk lands on the customer. And a pursuit that arrives with fewer than ten business days to close and no prior tracking is a no-bid unless a prime brings us in with the solution already framed.
The rubric's real job is not the number. It is the record. Six months later, when you compare the scores of the bids you won against the ones you lost, the weights start to correct themselves against your own results instead of against generic advice.
How much work belongs before the solicitation drops
The honest answer is a budget, not a doctrine. For a $150K to $500K action, a small firm can afford roughly fifteen to thirty hours of pre-solicitation work spread across four months. For a multi-year action above $2M, sixty to a hundred hours is defensible. Past that, the pursuit needs to be one of very few you are running, because the hours have to come out of delivery.
The capture clock for a mid-size action
Step three carries more weight than its six hours suggest. When a contracting officer decides whether to set an acquisition aside under the rule of two in FAR 19.502-2(b), the responses to the sources-sought notice are the evidence. A specific response that names relevant work, staff, and capacity can change the competitive field you eventually bid into. A generic capability statement does nothing.
Step two has a boundary worth naming. FAR 15.201 encourages exchanges with industry before receipt of proposals, and lists industry days, one-on-one meetings, draft solicitations, and RFIs as accepted techniques. The Procurement Integrity Act restrictions at FAR 3.104 govern what may not be shared, chiefly contractor bid or proposal information and source selection information. Asking a program office what the mission problem costs them today is normal and welcome. Asking what a competitor proposed is not.
The win-rate math, without flattery
Assume a twelve percent win rate, which is a defensible planning number for competitive small-business technical work and roughly in line with the published selection rates agencies report for early-phase research awards. The probability of winning at least one out of ten bids is 72 percent. To reach 90 percent you need eighteen bids. To reach 95 percent you need twenty-four. Those are the real numbers, and they explain why a firm that submits four proposals a year and wins nothing has learned almost nothing about its proposals.
At a 12 percent per-bid win rate
10 bids gives a 72 percent chance of at least one award. 18 bids gives 90 percent. 24 bids gives 95 percent. A single loss carries almost no information; a run of twelve losses with consistent scoring feedback carries a great deal.
The arithmetic also sets the bid budget. At a $250,000 average action and a twelve percent win rate, the expected value of one bid is $30,000 of top-line revenue. If the proposal costs sixty hours of engineering time, the pursuit is comfortably positive on expectation. If it costs two hundred hours, it is not, and no amount of enthusiasm changes that. Run this calculation before the pursuit, not after.
What volume alone does not fix
The binomial math assumes independent draws. Real proposals are correlated, because the same firm writes them with the same evidence, the same team, and the same blind spots. If every proposal has a thin section on transition, or every one names a role without naming the person who fills it, then twenty bids are not twenty draws at twelve percent. They are closer to one draw repeated twenty times.
Three failure classes stay constant across volume. The first is weak evidence: a proposal that argues capability instead of demonstrating it will lose to one that shows a working system and measured results, every time, at any volume. The second is the personnel section: reviewers read named people with real credentials, and a staffing plan built on roles to be hired reads as risk. The third is customer knowledge, which is why the capture clock above exists at all. A proposal written by a firm that has spoken to the customer is legibly different from one that has not, and reviewers notice within two pages.
So the volume question has a precise answer. Volume is the right strategy once the underlying proposal quality is sound, because it converts a good per-bid rate into near-certain wins over a year. Volume is the wrong strategy before that, because it multiplies a systematic defect. Fix the correlated weakness first, then scale the count.
Tracking that a small team will actually maintain
A CRM is not required and usually goes stale. What is required is one table, one owner, and a weekly ten-minute pass. The fields that earn their place: opportunity name, agency and office, sourcing channel, NAICS, estimated value, set-aside, incumbent, our role (prime or sub), the rubric score, the named owner on our side, next action, next action date, and the expected posting date. Anything else is decoration.
Two derived numbers make the table useful. Weighted pipeline value multiplies each live pursuit's value by its win probability, which keeps a list of twelve long shots from looking like a forecast. Channel yield tracks how many qualified pursuits each sourcing channel produced and how many of those converted, which is how you find out within two quarters that your forecast reading is worth more than your keyword alerts, or the reverse.
The weekly pass has exactly three questions per row: has anything changed, what is the next action, and should this be killed. Killing is the productive outcome. A pipeline where nothing ever dies is a list, and a list does not tell you where to spend Thursday.
State, local, and commercial lanes
Federal is not the only lane, and for a firm building toward steady revenue it should not be the only lane. States run their own procurement portals with their own registration requirements, their own small-business and resident-business preferences, and dollar thresholds far below the federal equivalents. Cooperative purchasing arrangements let one government's competitively awarded contract serve many others, which turns a single win into a durable channel. The qualification questions are the same; the sourcing channels are different, and the competitive field is usually thinner.
Commercial buyers of AI and data engineering work belong in the same pipeline table, with the same rubric and the same weekly pass. The sales cycle is shorter, the evidence that persuades is the same working system, and the revenue smooths the gaps between federal award cycles. We run all of it through one table for exactly that reason.
Bottom line
A capture pipeline for a small firm is four things: a set of sourcing channels weighted by what they actually produce, a twenty-minute qualification that kills most of what arrives, a weighted rubric with hard floors that you do not renegotiate mid-pursuit, and a budgeted number of pre-solicitation hours assigned to a named person. Run that, keep the table honest, and volume will do its work. Skip it, and volume will only make the same loss faster.
Frequently asked questions
Agency acquisition forecasts, agency small business offices, research-program solicitations during their pre-release windows, GSA eBuy if you hold a schedule, prime supplier portals, and public award data for upcoming recompetes. SAM.gov is the floor of a pipeline, not the pipeline, because task orders under existing vehicles and schedule buys do not appear there as new opportunities.
Six months is a reasonable target for an action you found in a forecast. The high-value steps are early customer contact and a specific sources-sought response, both of which have to happen well before the solicitation exists. For a $150K to $500K action, budget fifteen to thirty hours total across that period.
Yes. FAR 15.201 encourages exchanges with industry before receipt of proposals and names industry days, one-on-one meetings, draft solicitations, and RFIs as accepted techniques. The limits come from the Procurement Integrity Act restrictions at FAR 3.104, which govern contractor bid or proposal information and source selection information. Discussing the mission problem is normal; asking about competitors is not.
At a twelve percent win rate, eighteen bids gives roughly a 90 percent chance of at least one award and ten bids gives about 72 percent. That is the arithmetic case for volume. It only holds if the proposals are independent in quality; a systematic weakness repeated across twenty bids does not behave like twenty independent draws.
Hard floors. A structural blocker such as a required clearance, facility, or contract vehicle you do not hold ends the pursuit regardless of total score. So does a poor technical fit, because the delivery risk transfers to the customer. Write the decision down either way, so the scores can be compared against outcomes later.