# Exploitation evidence carries 40, fixed before any score

METHOD, read 10 Nov 2026

**The short answer.** Platforms are ranked on how they turn exploitation evidence into a priority: evidence 40, seeing exposure before the scanner 15, unifying sources 15, automation 15, and 5 each for AI governance, public sector readiness and proof. The weights were committed before scoring, and the scorer saw coded evidence, never names.

## The weights

| Criterion | Weight | What is scored |
| --- | --- | --- |
| Unifies every source | 15 | Number and kind of sources ingested as documented, a way to add one not on the list, correlation and de-duplication. |
| Prioritizes on exploitation evidence | 40 | Uses exploitation-in-the-wild evidence (KEV, exploit and malware evidence, EPSS or similar prediction) with asset and business context, and lets the customer shape the scoring. |
| Automation at enterprise scale | 15 | Rules and workflows that assign, ticket and track without a person per finding, at volumes the vendor documents. |
| Sees exposure before the scanner | 15 | Answers 'are we exposed' to a new disclosure before scan signatures arrive. |
| AI under governance | 5 | Agents and assistants with approval, audit and limits the customer sets. |
| Public sector readiness | 5 | FedRAMP status and support for federal directives and frameworks. |
| Proof | 5 | Analyst recognition, validations with the commissioner named, named customer statements. |


Total = the sum of each weight times the criterion score (0 to 10), divided by 100, so totals run from 0 to 10. Note on the split: 'the other three 15 between them' split 5/5/5.

## The five situations and their weights

Each situation page ranks the platforms on sale on that situation's own weights, committed on the same day as the site weights and before any score. The weights are the same on all three sites of this publication group, because the buyer is the same.

| Situation | Unifies every source | Prioritizes on exploitation evidence | Automation at enterprise scale | Sees exposure before the scanner | AI under governance | Public sector readiness | Proof |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Federal agencies and contractors | 15 | 20 | 15 | 5 | 5 | 30 | 10 |
| Large enterprises with ten or more scanners | 30 | 20 | 30 | 5 | 5 | 0 | 10 |
| Managed security providers | 25 | 20 | 30 | 5 | 10 | 0 | 10 |
| Mid-size teams replacing Kenna or a spreadsheet | 20 | 30 | 20 | 5 | 10 | 0 | 15 |
| Application security and infrastructure in one queue | 35 | 20 | 25 | 5 | 5 | 0 | 10 |

## How the scores were made

- Evidence first. Researchers collected public facts for every platform, each with its source URL and read date, from the vendor's documentation, its marketing pages, government records (the GSA FedRAMP Marketplace data) and third parties.
- Blind scoring. A scorer received the rubric and the evidence with vendor and product names replaced by codes and URLs removed. It gave every platform an integer from 0 to 10 on each of the seven criteria.
- Blind review. A second reviewer, also working on the coded evidence, checked all 91 scores against the anchors and proposed 13 corrections, all applied before the codes were opened. They are listed below.
- Totals by script. Every total and every order on this site is computed by one scoring script from the committed scores and weights. Nobody types a total.
- Ties. Exact ties are broken by the higher evidence score, then the higher unify score, then alphabetically, a rule committed with the weights.

## The rounding rule

Rankings use the exact totals, which have at most two decimals because the weights are whole numbers. Pages show one decimal, rounded half up (7.05 shows as 7.1). Where two platforms in the same table would show the same one-decimal figure, that table prints two decimals, so a printed tie is always a real tie.

## What kind of evidence the scores rest on

A criterion supported only by a vendor's marketing pages scores at most 7. Technical documentation, government records or third-party sources can lift a score above 7. Outcome numbers (percent reductions, findings per day) are vendor claims and count for at most one point. Nucleus's evidence is 69 marketing facts and 2 government records out of 71, with no public technical documentation found on 10 Nov 2026; that is why four of its seven scores sit at exactly 7, the marketing ceiling. Leaders that pass 7 on a criterion do so on documentation.

| Platform | Facts | Technical docs | Marketing | Government | Third party |
| --- | --- | --- | --- | --- | --- |
| Nucleus | 71 | 0 | 69 | 2 | 0 |
| Brinqa | 38 | 17 | 21 | 0 | 0 |
| Qualys | 32 | 4 | 26 | 2 | 0 |
| Tenable | 31 | 2 | 28 | 1 | 0 |
| ServiceNow Vulnerability Response | 22 | 16 | 0 | 4 | 2 |
| Phoenix Security | 36 | 0 | 35 | 0 | 1 |
| ArmorCode | 31 | 0 | 31 | 0 | 0 |
| Hackuity | 33 | 0 | 33 | 0 | 0 |
| Seemplicity | 32 | 0 | 28 | 0 | 4 |
| Zscaler UVM | 28 | 0 | 27 | 1 | 0 |
| Rapid7 | 25 | 5 | 19 | 1 | 0 |
| Vulcan Cyber | 27 | 0 | 27 | 0 | 0 |
| Cisco Vulnerability Management | 24 | 18 | 5 | 0 | 1 |

A larger evidence file does not raise a score by itself: each anchor asks for a kind of evidence, not a count. Some competitor files are thinner because pages were unreachable on the read date (ServiceNow's newsroom returned errors, so its proof score of 0 means none found, not known absent).

## How firm the positions are

- The ranked view is an exact tie (Brinqa 6.90, Nucleus 6.90) settled by the committed tie-break on the evidence score.
- One point on any Nucleus criterion would put it first; one point off its evidence score would put it third.
- The KEV Lag Ledger is computed, not scored: it has no weights and changes only when the public data does.

## The rubric anchors

### Unifies every source (weight 15)

- **0:** No documented ingestion of third-party findings (single-vendor scanner only).
- **3:** Ingests its own scanner plus a few third-party sources; no documented de-duplication.
- **5:** Documented connectors across infrastructure and one other kind (appsec, cloud, pentest); de-duplication documented or claimed.
- **7:** Documented connectors across infrastructure, application security, cloud and asset sources; documented correlation and de-duplication; a generic import (CSV/API) for unlisted sources.
- **9:** As 7, plus a documented universal or custom connector framework for unlisted sources and documented normalization across finding types.
- **10:** As 9, with third-party confirmation.

### Prioritizes on exploitation evidence (weight 40)

- **0:** CVSS only.
- **3:** One exploitation signal (for example KEV flag) beside CVSS; no customer control.
- **5:** Several exploitation signals combined into a proprietary score; asset criticality used; limited customer control.
- **7:** Several named exploitation signals plus asset and business context; documented customer-adjustable scoring or rules.
- **9:** As 7, with documented, fully customizable scoring formula or weights and the signals named and exposed per finding.
- **10:** As 9, with third-party confirmation.

### Automation at enterprise scale (weight 15)

- **0:** Manual export only.
- **3:** Manual ticket creation from the console; one ticketing integration.
- **5:** Automated ticket creation to at least two ticketing systems; basic assignment rules.
- **7:** Rule-based ownership assignment, bi-directional ticket sync, SLA tracking, automated closure on verification.
- **9:** As 7, with documented enterprise volume figures or a documented workflow/automation engine beyond ticketing.
- **10:** As 9, with third-party confirmation.

### Sees exposure before the scanner (weight 15)

- **0:** None found.
- **3:** Threat intelligence alerts on new disclosures without matching to the customer's assets.
- **5:** Matches new disclosures to an inventory (software, SBOM, asset attributes) without waiting for a scanner plug-in, described in marketing only.
- **7:** Documented capability matching new disclosures to the customer's inventory before scanner signatures.
- **9:** As 7 with documented speed or zero-day workflow.
- **10:** As 9, with third-party confirmation.

### AI under governance (weight 5)

- **0:** No AI assistant or agent found.
- **3:** Chat assistant for questions only; no governance controls described.
- **5:** Assistant or agents that take actions, governance controls not documented.
- **7:** Agents or assistant with at least one documented governance control (approval step, audit log, role-based limits).
- **9:** Agents with documented approval, audit and customer-set limits.
- **10:** As 9, with third-party confirmation.

### Public sector readiness (weight 5)

- **0:** None found.
- **3:** Claims public-sector customers or compliance mapping without FedRAMP.
- **5:** FedRAMP In Process or Ready (Marketplace), or FedRAMP for a different product of the same vendor that does not cover this one.
- **7:** FedRAMP Authorized (Marketplace) for the product, Moderate, or Authorized claim not confirmed on the Marketplace.
- **9:** FedRAMP Authorized on the Marketplace, plus documented support for CISA directives (BOD/KEV reporting) or CMMC / NIST RMF.
- **10:** As 9 at High impact level, or with DoD IL4+.

### Proof (weight 5)

- **0:** None found.
- **3:** Unnamed customer stories only.
- **5:** Named customer statements, or one analyst mention (any placement).
- **7:** Analyst report placement (for example Gartner MQ any quadrant, Forrester Wave, IDC MarketScape) plus named customer statements.
- **9:** Leader placement in a major analyst evaluation of this category plus named customers.
- **10:** Leader placement in two or more major evaluations plus named customers.

## Rules applied to every score

- Score only from the evidence file. Anything not in the evidence is 'none found on 2026-11-10' and earns nothing.
- Evidence kinds: vendor-doc (technical documentation, API reference, help center, integration catalog), vendor-marketing (product pages, blogs, press releases, outcome numbers), third-party (analyst, press, review site, partner, customer outside the vendor's site), government (FedRAMP Marketplace, CISA, NIST).
- A criterion whose supporting evidence is only vendor-marketing scores at most 7. vendor-doc, third-party or government evidence confirming the capability lifts the cap.
- Outcome numbers (percent reductions, findings per day, minutes to ticket) are vendor claims; they may support at most 1 point of any criterion and never lift the cap.
- A product announced as end-of-sale or merged into another product is scored on its documented capabilities; its status is reported, not scored.
- Breadth (raw connector count) contributes at most 2 points of the unify criterion, so it never carries more than 10 percent of any ranking (highest unify weight is 35).

## Score changes in the blind review

| Platform | Criterion | From | To | Reason |
| --- | --- | --- | --- | --- |
| Phoenix Security | AI under governance | 6 | 7 | Anchor 7 needs one governance control; a stated developer-approval step on every AI PR meets it at the marketing cap, as Tenable and Qualys got 7 on the same kind of marketing evidence. |
| ArmorCode | Prioritizes on exploitation evidence | 5 | 6 | Named KEV, EPSS, AATI and exploit status plus asset criticality, data sensitivity and business context, without customer control, is the same position Phoenix Security and Hackuity were given 6 for. |
| ArmorCode | AI under governance | 6 | 7 | Missed a fact: high-impact automated changes are routed to human review (an approval step), plus auditable decisions, meets anchor 7 at the marketing cap, consistent with Tenable/Qualys. |
| Zscaler UVM | Sees exposure before the scanner | 2 | 0 | The cited fact is about mitigating before a patch window, not detecting exposure to new disclosures; no evidence touches anchor 3, so none found gives 0. |
| Rapid7 | Proof | 6 | 5 | The customer quote is unattributed, so anchor 7's named-customer condition is unmet; an analyst Leader claim alone is anchor 5, as ArmorCode (IDC Leader, no named customers) got. |
| Seemplicity | Prioritizes on exploitation evidence | 5 | 6 | Named KEV, EPSS, VulnCheck, in-the-wild and exploit-code signals plus asset value and business context, without customer control, matches the 6 given to Phoenix Security and Hackuity. |
| Brinqa | Prioritizes on exploitation evidence | 8 | 9 | Vendor-doc shows KEV and EPSS exposed per CVE and a customer-configurable formula (base plus customer-set risk factors), which meets anchor 9 in full; the reason names nothing missing. |
| ServiceNow Vulnerability Response | Unifies every source | 5 | 4 | Anchor 5 needs de-duplication documented or claimed, but the docs say there is none across integrations, and the connector count is undocumented (0 breadth points); 4 sits between 3 and 5 for the custom-integration guidance. |
| ServiceNow Vulnerability Response | Prioritizes on exploitation evidence | 8 | 9 | Vendor-doc shows KEV, EPSS and Early Warning fields on each finding and customer-built calculators with conditions and weight percentages, which is anchor 9's fully customizable weights with signals exposed per finding. |
| Vulcan Cyber | Prioritizes on exploitation evidence | 6 | 7 | Anchor 7 is met at the marketing cap: named exploit sources (exploitDB, Metasploit, M.E.D, Mandiant, Recorded Future), business-context tagging and customer-configurable weights; KEV/EPSS by name is not a rubric requirement. |
| Qualys | Unifies every source | 6 | 5 | No de-duplication is documented or claimed, so anchor 5 is not met; 3 plus a point for the documented CSV import and multi-kind catalog plus 1 breadth point (100+) caps it at 5, consistent with Rapid7 and ServiceNow Vulnerability Response. |
| Qualys | Public sector readiness | 9 | 10 | The scorer treated anchor 9 as met, and the Marketplace lists the Government Platform as FedRAMP Authorized at High, which is anchor 10's extra condition (Nucleus got 9 on the same directive evidence at Moderate). |
| Nucleus | AI under governance | 6 | 7 | An approved-actions step and role-based access controls on AI tool connections are documented governance controls; anchor 7 at the marketing cap, consistent with Tenable and Qualys. |

No score has changed since the codes were opened on 10 Nov 2026. A score changes only when a sourced fact changes; the change is logged on the corrections page with the fact.

## The KEV Lag Ledger computation

Every KEV entry with dateAdded in the trailing 365 days of the catalog release. NVD published date and CVSS base score (v3.1, else v3.0, else v4.0, else v2; NVD Primary first). EPSS score and percentile on the KEV dateAdded from the FIRST API. days_publication_to_kev = dateAdded minus NVD published date, in UTC calendar days.

## Sources

- GSA FedRAMP Marketplace data file: https://raw.githubusercontent.com/FedRAMP/marketplace-fedramp-gov-data/main/data.json (read 10 Nov 2026)

Last reviewed 10 Nov 2026. Canonical: https://epssvscvss.com/method/
