The average government contracting portal publishes hundreds of new tenders every week. A contractor working in IT services in Virginia might see 400 new federal postings on SAM.gov in a given week — and maybe 15 of them are genuinely worth pursuing. The challenge isn't finding tenders. It's finding the right ones without spending your entire week reading solicitations that were never going to be a fit. AI ICP scoring solves this problem by reading every tender so you don't have to.
- What is an Ideal Customer Profile (ICP) in government contracting?
- The problem with keyword search for tender monitoring
- How AI ICP scoring works — the four dimensions
- NAICS code matching — the foundation
- Contract value scoring — why range matters
- Jurisdiction scoring — geography and relationships
- Set-aside and certification matching
- How scores combine into a single 0–100 number
- What a 92 score means vs a 41 score
- Semantic embeddings — the next frontier
- How to configure your ICP for maximum accuracy
- FAQ
What is an Ideal Customer Profile (ICP) in government contracting?
In commercial B2B sales, an Ideal Customer Profile (ICP) describes the type of company most likely to buy from you — defined by industry, company size, geography, budget, and other characteristics. In government contracting, the concept is inverted: your ICP describes the type of contract most likely to be won by you — defined by NAICS codes, contract value range, target jurisdictions, set-aside eligibility, and agency relationships.
A well-defined government contracting ICP is specific. It's not 'construction contracts in the Southeast.' It's 'NAICS 236220 and 237310, contract values between $500K and $5M, Texas and Georgia state agencies plus Army Corps of Engineers federal, HUBZone set-asides preferred, interested in SDVOSB opportunities.' That level of specificity is what allows an AI scoring system to accurately evaluate each of the hundreds of tenders posted every day and surface only the ones worth your time.
A vague ICP produces vague scores. If your NAICS codes are too broad, your jurisdiction is set to 'all states,' and your value range spans $100K to $50M, you'll receive high scores on tenders that have almost nothing to do with your actual capabilities. The time spent configuring a precise ICP is paid back every week in hours of filtering you don't have to do manually.
The problem with keyword search for tender monitoring
Every government procurement portal has a keyword search function. Most contractors use it as their primary filtering tool — searching for 'cybersecurity,' 'road repair,' or 'healthcare staffing' and reviewing what comes up. This approach has three fundamental problems that keyword search cannot solve.
Problem 1: Synonym blindness
Government solicitations are written by different contracting officers at different agencies with different institutional vocabularies. A road maintenance contract might be titled 'Pavement Rehabilitation Services,' 'Roadway Surface Treatment,' 'Asphalt Overlay Program,' or 'Highway Maintenance Services.' A keyword search for 'road repair' returns none of these. NAICS-based scoring returns all of them, because they all share the same NAICS code regardless of how the title is worded.
Problem 2: False positives at scale
Broad keywords generate massive false positive rates. A search for 'IT services' on SAM.gov returns thousands of results — most of which are either too large, too specialized, in the wrong geography, or restricted to certifications you don't hold. Manually filtering thousands of keyword results to find the 15 that matter is exactly the 4+ hours of wasted time that ICP scoring eliminates.
Problem 3: No relevance ranking
Keyword search returns results in chronological or alphabetical order — it has no mechanism for telling you which of the 200 results is most relevant to your specific business. ICP scoring returns results in relevance order — the highest-scoring tenders appear first, the lowest last. The difference between navigating 200 unranked results and a ranked list where the top 10 are genuinely strong matches is measured in hours per week.
How AI ICP scoring works — the four dimensions
BidEdgeHQ's ICP scoring engine evaluates each tender against your profile across four dimensions, each contributing a weighted portion of the final 0–100 score. The weights are calibrated based on which factors most reliably predict whether a contractor will both qualify for and win a given tender.
NAICS code matching — the foundation
NAICS code matching is the highest-weighted dimension because it's the most reliable predictor of whether your business is actually capable of performing the work. Every federal contract on SAM.gov is assigned a primary NAICS code by the contracting officer. State contracts use NIGP codes, which we map to NAICS. Canadian contracts use GSIN codes, which we also map.
The scoring isn't binary — it's graduated based on match precision. An exact NAICS code match scores the full 30 points. A match at the 4-digit NAICS group level (same industry cluster, slightly different specialty) scores 20–25 points. A match at the 2-digit sector level (same broad sector, different industry) scores 10–15 points. No match scores 0 points in this dimension, capping the total possible score.
ICP scoring only finds matches for NAICS codes in your profile. If you can credibly perform work under NAICS 541512 (Computer Systems Design) and 541511 (Custom Programming), register both. A tender for custom programming work will score 0 on NAICS matching if 541511 isn't in your profile — even if it's exactly the type of work you do.
Contract value scoring — why range matters
Contract value matching is weighted at 25 points because contract size is one of the strongest predictors of whether a bid is winnable and worth pursuing. A $50M contract requires a proposal team, past performance at comparable scale, and bonding capacity that most small businesses don't have. A $75K contract might not justify the proposal investment for a firm billing $5M annually. Your sweet spot — the range where you've consistently won and can win again — is the core of your value scoring.
The scoring applies a curve rather than a cutoff. If your value range is $500K–$3M, a $600K contract scores full points. A $4M contract scores 18–22 points (slightly outside your range but not dramatically so). A $25M contract scores 0–5 points (far outside your range — even if you can theoretically bid, your win probability is low). A $50K contract also scores low — too small to justify your proposal effort.
Why both floors and ceilings matter
Most contractors focus only on contract ceilings — they don't want contracts that are too large. The floor matters equally. Small contracts with high proposal requirements (complex technical evaluations, extensive past performance documentation) generate negative ROI for many small businesses. Setting a realistic minimum value in your ICP filters out the long tail of micro-contracts that would consume proposal resources disproportionate to their revenue impact.
Jurisdiction scoring — geography and relationships
Jurisdiction matching is weighted at 25 points. For most government contractors, geography is a genuine constraint — particularly for construction, facilities management, field services, and any work requiring on-site presence. A construction company based in Houston can theoretically bid on a New York City facilities contract, but their competitive position is weaker than a local firm, their subcontractor network is less developed, and their past performance in that specific geography is thinner.
The jurisdiction scoring is also graduated. An exact state match (you target Texas, the tender is a Texas agency) scores full points. A federal contract scores full points if federal is in your jurisdiction list. An adjacent state scores partial points — useful if you're expanding your geographic footprint. A state outside your target geography scores 0, even if the NAICS and value range are perfect matches.
Federal contracts are available to any US-registered business regardless of geography — a contractor in Montana can bid on a federal IT contract based in Washington DC. For this reason, 'Federal' is a separate jurisdiction option in your ICP, independent of state selections. Most contractors should include Federal in their jurisdiction list alongside their target states.
Set-aside and certification matching
Certification matching contributes 20 points to the score, but its impact is more binary than the other dimensions. A tender designated as an 8(a) set-aside is only available to 8(a)-certified firms — if you're not 8(a)-certified, the score for this tender is automatically capped regardless of how well it matches on other dimensions, because you literally cannot win it.
Conversely, if you hold 8(a) certification and a tender is designated as an 8(a) set-aside, you receive bonus points that push the score above what NAICS and value alone would produce. The certification match is an eligibility signal — it tells you both whether you can compete and whether your competitive pool is dramatically reduced.
- ✓8(a) set-aside + 8(a) certified: +20 points, competitive pool typically 10–50 firms nationally
- ✓HUBZone set-aside + HUBZone certified: +20 points, competitive pool typically 20–100 firms
- ✓SDVOSB set-aside + SDVOSB certified: +20 points, strong advantage at VA and DoD
- ✓WOSB set-aside + WOSB certified: +20 points, 5% federal goal creates consistent demand
- ✓Full-and-open competition + no relevant certification: 0 certification points, but full score available on other dimensions
- ✓8(a) set-aside + not 8(a) certified: automatic score cap — you cannot win this contract regardless of other factors
How scores combine into a single 0–100 number
The four dimension scores are summed to produce the final 0–100 score. A perfect match — exact NAICS, ideal value range, target jurisdiction, matching certification — scores 100. But in practice, scores in the high 80s and 90s are more common for genuinely strong matches, because the scoring system applies partial credit and curves rather than rigid cutoffs.
What a 92 score means vs a 41 score
A score of 92 means the tender matches your ICP across all four dimensions with high precision: the NAICS code is an exact match, the contract value is within your sweet spot, the jurisdiction is a target geography, and the set-aside type either matches your certifications or is unrestricted. This tender deserves immediate attention — stop what you're doing and read it.
A score of 41 means there's partial alignment — maybe the NAICS is a partial match and the value is slightly outside your range, or the jurisdiction is adjacent rather than primary. It might be worth a 60-second scan during your weekly review, but it shouldn't interrupt your day. A score of 41 is the system telling you 'maybe, when you have time.'
A score of 12 means the tender is in your sector but has fundamental mismatches on most dimensions — wrong geography, wrong value range, restricted to a certification you don't hold. The system is telling you 'not for you.' Most contractors never need to see these.
Scores of 80+ are tenders worth interrupting your day for. Scores of 60–79 are worth reviewing in your weekly opportunity review session. Scores below 60 are low priority. Configure your alert threshold accordingly — you'll receive 2–10 WhatsApp notifications per week at the 80+ threshold, each of which genuinely deserves your attention.
Semantic embeddings — the next frontier
BidEdgeHQ's current scoring system is already significantly more accurate than keyword search or manual filtering. The next evolution — already in development — is semantic embedding scoring, which adds a fifth dimension: the actual text similarity between your capability statement and the tender's scope of work.
Semantic embeddings work by converting text into high-dimensional numerical vectors that capture meaning, not just keywords. Your capability statement ('We provide cloud infrastructure migration, DevSecOps implementation, and zero-trust security architecture for federal civilian agencies') is converted into a vector. Each tender's scope of work is converted into a vector. The cosine similarity between the two vectors produces a semantic match score that identifies relevant opportunities even when the vocabulary doesn't overlap.
This means a tender for 'cloud modernization services' scores high for a contractor whose profile mentions 'infrastructure migration' — even though neither phrase appears in the other's text. It means a facilities management tender that describes 'preventive maintenance, HVAC, and janitorial services' matches a contractor profile that describes 'building operations and O&M services' — because the concepts are semantically similar even if the words differ.
How to configure your ICP for maximum accuracy
The accuracy of your ICP scores is a direct function of the accuracy of your ICP profile. An overly broad profile produces mediocre scores for many tenders but few precise high scores. An overly narrow profile produces high scores for the few perfect matches but misses adjacent opportunities. The goal is precision calibrated to your actual win probability.
Start with your actual history, not your aspirations. The NAICS codes where you have past performance are your highest-probability codes — they drive your best scores. Add these first, then add adjacent codes for capabilities you can credibly demonstrate even without a specific contract win.
Look at your last 10 contract wins. What's the range? Set your minimum to 80% of your smallest win and your maximum to 120% of your largest win. This range reflects where you've proven competitive. Extend the ceiling only for contract types where you have a specific, credible plan to compete above your historical range.
Start with the states and agencies where you've worked before. Past performance in a jurisdiction is your strongest competitive differentiator in that jurisdiction. Add Federal if you have SAM.gov registration and any federal past performance. Add new jurisdictions you're actively expanding into — but be realistic about your win probability in markets where you have no relationships.
8(a), HUBZone, SDVOSB, WOSB, VOSB — add every current, active certification. The scoring system uses these to boost your score for matching set-asides and cap your score for set-asides you're excluded from. An unchecked certification is a missed opportunity to surface contracts restricted to your competitive pool.
After two weeks of monitoring, review the tenders that scored 80+ and ask: are these the right opportunities? If you're seeing too many irrelevant high scores, narrow your NAICS codes or tighten your value range. If you're missing opportunities you know exist in your market, check whether you've registered the right NAICS codes or included the relevant jurisdictions.
Frequently asked questions
How is ICP scoring different from keyword search?
Keyword search matches words. ICP scoring matches business fit. A keyword search for 'construction' returns every tender that contains the word 'construction' — including tenders that are the wrong size, in the wrong geography, restricted to certifications you don't hold, or in adjacent industries you don't serve. ICP scoring evaluates each tender against your specific NAICS codes, value range, jurisdictions, and certifications to produce a relevance score that reflects your actual probability of qualifying and winning.
How many NAICS codes should I add to my ICP?
Add every NAICS code that accurately reflects work you can perform and have performed. There's no practical limit. However, adding NAICS codes for work you can't credibly deliver degrades your score accuracy — tenders in those codes will score artificially high even though you're not genuinely competitive. A focused profile with 5–10 highly accurate NAICS codes outperforms a bloated profile with 40 aspirational ones.
What does a score of 100 mean?
A score of 100 means the tender is a perfect match on all four scoring dimensions: exact NAICS code, value within your ideal range, exact jurisdiction match, and a certification set-aside that matches a certification you hold. In practice, scores above 90 are uncommon because few tenders match all four dimensions perfectly. When you see a 95+ score, treat it as a priority — it's a rare, high-probability opportunity.
Can ICP scoring identify sole-source opportunities?
Yes. Sole-source justifications are posted on SAM.gov as J&A (Justification and Approval) notices before or shortly after the award. BidEdgeHQ ingests these and scores them. A sole-source notice for work in your NAICS code, at your value range, in your jurisdiction, restricted to a certification you hold scores very high — and signals that the agency is buying from a specific firm. This intelligence is valuable for understanding your competitive landscape even when you can't bid on the specific contract.
How often is the scoring model updated?
The scoring weights and logic are reviewed quarterly based on outcome data — which scores correlated with contract wins for contractors in our network. The scoring isn't static; it improves as we learn more about which profile factors most reliably predict win probability. The semantic embedding capability, when released, will be additive — it won't replace the existing four-dimension scoring but will enhance it with text-based similarity.
Does ICP scoring work for Canadian tenders?
Yes. BidEdgeHQ maps Canadian GSIN codes (federal) and provincial commodity codes to NAICS codes, so your NAICS-based ICP profile scores Canadian tenders on the same 0–100 scale as US federal and state tenders. Jurisdiction scoring treats Canadian provinces as distinct jurisdictions — adding 'Ontario' or 'British Columbia' to your ICP activates scoring for those markets.
The fundamental promise of AI ICP scoring is simple: you should spend your time evaluating opportunities that match your business, not filtering out the ones that don't. Every hour spent reading a tender that was never going to be a fit is an hour not spent on a proposal that could win. A well-configured ICP profile, with accurate NAICS codes, a realistic value range, and precise jurisdictions, turns a firehose of tender notifications into a curated shortlist of genuine opportunities. The contractors who will win disproportionately in the next decade of government contracting are not the ones who read the most tenders. They're the ones who read the right ones — fast enough to act on them before the competition has finished checking their email.