The email lands at 7 a.m. with forty matched opportunities in the subject line. Two of them are work your firm could actually perform. The other thirty-eight are a janitorial recompete three states away, an aircraft component your shop has never machined, and three dozen small variations on the same mismatch. You skim all forty anyway, because the two real ones are buried in there, and finding them costs the first half hour of the working day.
That half hour is the real price. The subscription fee on a bid matching service is visible and budgeted. The attention is not, and attention is the scarcest resource a small business development effort has. Most small contractors run capture with one person, often the owner, in the hours left over after delivering the work already won. A feed of government contract leads that burns those hours on notices nobody should have sent is not a lead service. It is a distraction engine with an invoice.
Volume is the cheapest thing to sell
Most government bid notification services run on keyword matching: literal string comparison against the titles and descriptions of public notices. The word “security” appears in your profile, so the feed sends everything containing “security,” which includes guard services, network penetration testing, alarm panel maintenance, and the janitorial contract at a secure facility whose statement of work mentions escort requirements. The software is not wrong. It did exactly what literal matching does. The word matched; the work did not.
The economics explain why the feeds stay this way. Volume costs the vendor nothing to generate. Once the scraper runs, sending you eighty matches costs the same as sending eight, and eighty looks better in a sales demo. A daily email with two entries reads like a service that is barely working, even on the day when two is the honest number. So the incentive runs toward recall and away from precision, and the cost of that trade lands entirely on your side of the ledger, one skimmed notice at a time.
The damage compounds quietly. When thirty-eight of forty matches are noise, you stop reading closely, because close reading has stopped paying. The thirty-ninth notice gets the same two-second glance as the rest, and one week it is a sources sought from an office you have courted for a year. Alarm fatigue is a known failure mode in every monitoring discipline, from radiology to network operations. Bid feeds are not exempt.
“In my years running procurements for an agency, I could name the vendors who responded to everything we posted, because their cover letters never changed. I cannot remember one of them winning. The responses that won read like the firm had been waiting for that exact notice, and the good ones usually had been.”
What SAM.gov gives you for free
Start with what the government already provides, because it is more than most paid feeds admit. SAM.gov is the governmentwide point of entry under FAR 5.201, federal notices above the publication thresholds flow through it, and its saved searches and email alerts are free with an account. You can scope a saved search by NAICS code, PSC, set-aside type, notice type, agency, and place of performance, and have new matches mailed daily. The same account lets you follow individual notices, so amendments, deadline extensions, and question-and-answer attachments reach you without re-running the search. For federal work, that is the floor any paid service has to clear, and knowing the floor is how you price whatever sits above it. Any service whose product amounts to a repackaged SAM.gov alert is charging for something you could own by Friday.
The honest case for paying is what the free tool does not do, and it is worth stating exactly. SAM.gov covers federal opportunities only; state, county, municipal, and school district work lives across hundreds of separate portals that never touch it. Its keyword logic is literal, so the “security” problem above is fully present. It applies no judgment: a saved search cannot know that a solicitation technically inside your NAICS code is wired for an incumbent, sized wrong for your bonding, or in a state where you have no past performance. And it starts at publication. A notice reaches your alert the day it posts, which is the day the requirement stops being shapeable. Everything before that day happened where alerts do not reach.
What a defensible search strategy contains
The difference between a keyword list and a search strategy is that a strategy encodes who buys from you, not which words a clerk might have typed into a title. Built properly, it carries at least six layers.
- NAICS and PSC codes together. NAICS classifies the seller’s industry; the Product or Service Code classifies the thing being bought. Contracting offices filter on both, and either one alone leaks. A grounds crew watching only NAICS 561730 misses the requirement an office coded under facilities support.
- Target agencies and buying offices, drawn from award history. Not the agencies you happen to recognize; the offices with awards in your codes in each of the last three fiscal years. We published a full method for building that list in finding the offices that already buy what you sell.
- Set-aside posture. If you hold SDVOSB, 8(a), WOSB, or HUBZone status, reserved solicitations are a stream worth watching on their own. If you hold nothing, plain small business set-asides and full-and-open competition are your lanes, and everything reserved for a certification you lack is noise by definition.
- Contract vehicles. Task orders under an IDIQ you do not hold are not opportunities. They are notes for the next on-ramp. A vehicle holder has the opposite problem and needs fair-opportunity notices under their vehicles surfaced fast.
- Geography where it matters. A firm that self-performs with local crews should never see work outside driving range. A software firm should never filter on it at all.
- Negative filters. The layer almost nobody builds, and the one that kills the most noise. Excluded words, excluded offices, excluded notice types, dollar floors and ceilings. “Security” minus “guard” minus “custodial” is a different feed from “security.”
None of this is exotic. It is the difference between matching strings and modeling a market, and anyone willing to spend the setup hours can build it.
A strategy also decays. Offices reorganize, incumbents change, a requirement you relied on comes back recompeted under a different code, and last year’s negative filter starts eating a notice you now want. The filters need the same quarterly pruning as the target list they came from. Left alone, a layered strategy drifts back toward the keyword list it was built to replace.
Backtest the feed before you trust it
There is a test for any feed, including one you built yourself, and it runs on public data. Award records live in SAM.gov Contract Data, the system that absorbed the retired FPDS site, behind a free Login.gov account. Before trusting a feed with a year of your attention, run its logic backward against the year that already happened:
- Pull the last twelve months of awards at your target offices in your codes. These are the awards you should have seen coming.
- For each award, find the pre-award notices behind it in SAM.gov’s opportunity archive, and check whether the feed’s filters would have surfaced them when they posted.
- Count the misses: awards you would have wanted to chase that the feed would have skipped.
- Count the noise: everything the feed would have sent that connects to no award you would have touched.
The misses teach you more than the score does, so read where they cluster. Misses coded under a neighboring PSC mean the code layer is drawn too narrow. Misses that surfaced only as a presolicitation or a sources sought mean the notice-type filter is discarding the early signal. Misses with no federal notice behind them at all usually mean the award ran through a state portal or an existing vehicle, and no SAM.gov alert was ever going to catch it. Each cluster names the layer that needs work.
A feed that fails the backtest fails the year. Vendors will not run this test for you, because the result prices their product. Run it yourself, and run it against your own saved searches too, which fail it more often than their builders expect.
The race starts before the notice posts
There is a harder limit under all of this, and no amount of filtering moves it. FAR 10.002 has agencies conduct market research before soliciting, through sources sought notices, requests for information, and direct exchanges with industry. By the time the solicitation your alert catches is published, that research is documented, the requirement is written, and the set-aside decision is made. The vendors who answered the sources sought are inside the government’s market research file, and their capabilities influenced how the requirement reads. Some of that shaping ends in a set-aside sized around the firms that responded, and some ends further along still, in a justified sole source that was earned years earlier.
A notification service tells you the race has started. Market research tells you where the starting line forms. Both are useful, and only one of them can be bought as an alert. Reading an office’s award rhythm and answering its early notices is a different discipline from watching feeds, which is why it is a separate engagement with a different cadence.
Three honest ways to watch
Most advice about how to find government contracts sells one tool as the answer for every firm. Different firms need different machinery, and the honest version of this industry would say so up front.
Run your own SAM.gov saved searches. Free, official, federal only, literal. The right answer for a firm with a tight set of codes, one or two target agencies, and a person who will open the alert every morning and prune the filters every quarter. The tool is not the weakness in this tier. The discipline is, and once the alerts have sat unread for two weeks the tier has already failed.
Real-time filtered monitoring, with a person to call. This is what our Contract Recon subscription is: continuous monitoring across federal, state, and local sources, filtered on the layered profile described above instead of bare keywords, with specialist reviews available when something bites. It fits firms with someone who can work a live feed, firms whose revenue sits in state and local portals SAM.gov never touches, and vehicle holders who cannot afford to learn about a task order on day ten of a fifteen-day window.
An analyst-curated weekly briefing. This is Contract Pulse: analysts run the search strategy, read what it returns, and send a short list a human has already judged for fit, tuned by your feedback over time. It fits the owner-led firm with no bandwidth for a feed of any kind, where the entire watching function has to fit inside one readable email. Curation carries a delay that a live feed does not, which is the honest trade: hours of reading exchanged for days of latency, and for most requirements the response window makes that trade cheap.
The wrong choice in either direction wastes the same resource. A briefing subscriber with a BD analyst on staff is paying to slow down information that person could work live. An owner running a raw feed alone is back at the 7 a.m. email, paying in attention.
Whichever tier fits, run the backtest before committing the year to it. If you want the live stream with the layered filters, Contract Recon monitors federal, state, and local releases in real time. If you want one reviewed list a week instead of a stream, Contract Pulse delivers an analyst-curated weekly briefing. And if the misses in your backtest cluster before publication, the fix sits upstream of any feed, in the market research that finds your buying offices before they post. The forty-notice email will keep arriving either way. Reading it is optional.