Public documentation is the starting point.
The first edition uses primary product documentation, available release information and disclosed evaluation material. Every scored dimension identifies the specific offering, edition, date, rationale and sources. Most sources are vendor-authored. We have not independently measured these offerings' effectiveness, uptime, operating effort, total cost or outcomes.
The maps label this basis as provisional editorial research. A published feature can support an assessment of documented capability. It cannot by itself prove that the feature is effective, easy to run, or appropriate for your organization.
Six dimensions, kept separate.
- Operational maturity
- Documented deployment, lifecycle, recovery and administration safeguards appropriate to the segment. This does not establish observed uptime or reliability.
- Shipped innovation
- Available workflows that meaningfully extend the stated segment baseline. Announcements, company size, AI branding and roadmap promises do not establish innovation.
- Capability breadth
- Coverage of the jobs named in this comparison group. Adjacent products and separately licensed modules must be identified; breadth is not effectiveness.
- Ecosystem & integration
- Documented data exchange, supported integrations and usable interfaces. A connector count alone does not establish integration quality.
- Governance & control
- Documented permissions, approval boundaries, auditing, policy and recovery controls. Actual behavior still needs verification in the selected edition.
- Operator enablement
- Documented deployment, tuning, troubleshooting and maintenance support. This is not a measurement of staffing requirements, cost or ease of use.
Each segment publishes its own criteria and six stage anchors for each dimension. A stage in endpoint security is not interchangeable with the same stage in identity management. The dimension inspector exposes the exact criterion and the source-backed reasoning for an offering.
Tenths with a traceable calculation.
Research scores run from 0.0 to 5.0 in tenths. Whole-number anchors retain their published meanings. A fractional score credits documented requirements toward the next anchor, using the same weighted criteria for every offering in that dimension and comparison group. These are editorial assessments, not measured percentages: 4.2 does not mean 40 percent better than 3.0, and a small difference is not proof of superior performance.
Each refinement interval divides the next anchor's new requirements into criteria whose weights total ten tenths. The score is the last fully supported anchor plus the credited weights. For example, a hypothetical 3.0 anchor with criteria worth 0.3, 0.4 and 0.3 earns 3.6 when the two 0.3 requirements are established. Meeting all three earns 4.0. The inspector shows the calculation, credited evidence, uncredited requirements and sources. Criteria already required by the base anchor earn no extra credit.
Weights express a published editorial rule, not a statistical model. Missing evidence earns no refinement credit and does not prove a feature is absent. Confidence is reported separately and never multiplied into a score. We retain ties when the evidence supports them; decimal places are not used as visual offsets.
- Unknown: evidence is insufficient to place the offering on that criterion. It stays null in the download, leaves a gap in Unfold, and cannot supply an axis coordinate.
- Stage 0: affirmative evidence supports the segment's lowest anchor. Lack of public documentation is not evidence of absence.
- Stages 1–5: progressively stronger documented scope under the published anchors. A rationale must explain the placement and its limits.
No overall score averages away a material weakness. Overview shows maturity and innovation directly. Unfold shows all six dimensions. The third axis in 3D reveals a chosen criterion; it does not change the underlying assessment. Every assessed offering is visible by name, with consistent numbered symbols across views. Overlapping marks separate with lines back to their exact positions; turn off “Separate overlapping marks” to see the original overlap. This spacing does not change scores or break a tie. Optional pair focus narrows the chart, while the full score table remains available.
Confidence is separate from the score.
Low means the conclusion depends on material inference, narrow evidence or unresolved scope. Medium means the cited material directly supports the scoped judgment, while independent operating evidence remains limited. Public-documentation-only assessments cannot receive high confidence. Source volume alone earns no confidence. Missing or gated documentation is an evidence gap, not a negative finding about the product.
Confidence labels are editorial evidence judgments, not statistical probabilities. Missing evidence and conflicting evidence should remain visible in the rationale and constraints.
Compare the same job.
Every offering belongs to an explicit comparison group. Workforce identity providers, privileged-access tools, SOAR software and staffed MDR services address different jobs. The explorer keeps those groups separate. A product's edition, optional modules, deployment model and current lifecycle determine what belongs in its assessment.
Any highest-score callout applies only to assessed offerings in the selected group and preserves ties. It does not identify a market-wide winner. Products outside the researched sample remain linked as unassessed. Selection for this first edition is editorial coverage, not evidence of leadership.
Gartner or Forrester positions are not inputs. We do not reuse their axes, ratings, research text or visual classifications. Read our guide to analyst research for context on interpreting external research.
Your environment changes the questions.
A scenario highlights relevant criteria and names constraints to validate. It does not alter research scores or certify eligibility. Mandatory requirements such as residency, supported systems, approval controls or staffing coverage need their own evidence; a high score elsewhere cannot compensate for an unmet requirement.
Direction needs comparable history.
Momentum keeps two readings separate for each dimension. Own movement is the latest score minus the earlier score. Relative movement subtracts the median score change of matched peers, excluding the selected offering. A vendor gaining 0.1 points while peers gain 0.2 is improving but losing 0.1 points of relative ground. This is movement in editorial scores, not measured product performance or statistical significance.
Every comparison uses the same start and end dates, dimension rubric, comparison-group scope and offering edition. Each offering must be re-reviewed at every date in the period. New entrants, stale reviews, unknown scores and scope changes are excluded. The peer count shows the assessed sample, not the whole market. At least two comparable peers are required for relative movement; otherwise only the offering’s own comparable movement is shown. With an even number of peers, a median may be a half-tenth: 0.15 is arithmetic, not extra scoring precision.
Dated snapshots preserve scores, rubrics, scope and sources. A changed score needs an explanation and a classification: product change, research correction, or scope change. Product changes need a supported effective date within the review period and evidence observed on or after that date. Research corrections and method changes break the comparison; newly discovered documentation is not a product release. We do not infer a regression simply because documentation disappears.
The original September 21, 2026 baseline incorporates a review of all six dimensions for the initial 28 assessed offerings. New segments start with their own dated baseline. Earlier decimal calibration and research corrections do not form a historical trend. “Building history” means a second comparable review is still needed, not zero movement. Our intended review cadence is quarterly; each edition requires an editorial evidence review before publication.
In Unfold and Overview, the selected offering’s hollow mark shows its earlier score and the solid mark its current score. A dashed connector shows the dated movement. The normal lines between Unfold dimensions are profiles, not time. Relative values remain separate from score coordinates. Select a review date and open “Review history & what changed” for the evidence. Numbered marks identify offerings, not rank.
Inspection and correction are part of the method.
Grok subagents assist with research, implementation and review. Codex performs integration and editorial checks. Neither model output nor a successful software test establishes product performance. Review includes exact offering scope, source relevance, score rationale, lifecycle status and unresolved uncertainty.
Each segment provides the research as JSON, including sources, dates, confidence and null values. To propose a correction, identify the offering and criterion, cite the current primary source, and explain the change. Open an issue in the public repository ↗.
Atlas Fold complements the Security Fit Map evaluation protocols. Those protocols still require representative testing before scored fit placement. An editorial documentation assessment does not satisfy that requirement.