One evidence graph.
Combine repository and scanner results with requirements, interface documents, defects, change history, service criticality and milestones.
Output · traceable architecture hotspots and evidence coverage
Private design-partner preview
The assessment is a date. Coupling is a rate.
An assessment tells you what architectural debt costs today. Radar tells you what it costs every month after — the same evidence graph and the same exposure model, kept current, with every assumption still visible.
Imports existing scanner evidenceDocumentation-first optionNo implementation contract
SPECIMEN-02 · portfolio viewModel ready
$3.1M exposure
Modeled range $2.4M–$4.1M · assumptions visible
18–31 wkschedule range
4fund-now candidates
Where Radar fits
Land on the number. Then keep it.
Radar is not sold on its own. Without a baseline it has nothing to measure against, and an alert with no reference point is noise. The assessment establishes the baseline; Radar keeps it from going stale.
Step one · 3 weeks, fixed fee
Architecture Debt & Cost Assessment
An independent measurement of what coupling costs this program in money and integration schedule, with a ranked remediation portfolio and a stated band. You can stop here, and most clients should.
Step two · annual, optional
Architecture Radar
The same model, re-run continuously against your evidence. It answers the question the assessment cannot: is the number moving, in which direction, and did the money you spent actually change it?
The decision layer
Not another issue counter.
Code-quality tools are good at finding unhealthy code. Radar answers the harder executive question: which structural problem is consuming the milestone, how confident are we, and what remediation sequence is worth funding?
A range finance can challenge.
Convert coupling and delivery friction into cost and schedule exposure. Change an assumption and see whether the decision still holds.
Output · base, conservative and stressed exposure casesA portfolio, not a backlog.
Rank remediation by expected exposure retired per dollar or engineering day—and make “leave it alone” an explicit decision.
Output · fund, monitor or accept with owner and decision dateInteractive preview
Change the assumptions. Test the decision.
The sample below is a functional scenario model using illustrative data at portfolio scale. Select architecture nodes, vary the investment and confidence assumptions, then export the sample portfolio.
Illustrative sample only · not a client estimate or validated production output
Architecture exposure overview
Model v0.8 · calibrated 06 Aug 2026 · 14 evidence sources
Modeled exposure
$3.1M
Range $2.4M–$4.1M
Schedule exposure
18–31 wk
P50 / P85 confidence
Evidence coverage
68%
+12 pts after review
Fund-now candidates
4
of 17 findings
Architecture evidence graph
Select a nodeScenario assumptions
EditableTargeted remediation case
IllustrativeExposure retired$1.86M3.1× investment
Payback range4–7 moafter implementation
Schedule recovered18 wkmodeled midpoint
Ranked remediation portfolio
Expected exposure retired / effort| Priority | Candidate | Owner | Effort | Exposure retired | Confidence | Decision |
|---|---|---|---|---|---|---|
| 1Gateway boundary | Partition shared interface state | Platform | 42 d | $0.52M | 86% | Fund |
| 2Verification harness | Decouple end-to-end fixtures | V&V | 31 d | $0.35M | 78% | Fund |
| 3Supplier adapter | Stabilize ownership contract | Integration | 18 d | $0.21M | 72% | Fund |
| 4Control services | Split change-heavy services | Controls | 64 d | $0.30M | 61% | Stage |
How Radar works
Evidence first. Arithmetic second.
Generative models can help classify documents, but they do not decide the exposure number. The production scoring path is deterministic, versioned and auditable. The method is published in full in the methodology brief.
Connect
Import scanner, repository, work-tracker and engineering-document evidence without forcing a tool replacement.
Normalize
Build a typed graph linking modules, interfaces, requirements, teams, defects and milestones—with provenance.
Model
Estimate exposure using change, coupling, defect, cost and schedule evidence. Show ranges and sensitivity.
Decide
Rank actions, compare scenarios and export the evidence pack for executive and technical review.
Fits the existing estate
Keep the scanner. Add the decision.
Radar is designed to sit above continuous analysis products and client evidence. Replacing those platforms would slow the decision and erase useful history.
Tools you may already own
- Detect code smells, vulnerabilities and architecture violations
- Map application or portfolio structure
- Enforce developer workflow and quality gates
- Benchmark software quality
What Radar adds
- Client-calibrated cost and schedule exposure
- Visible assumptions, confidence and sensitivity
- Remediation return and decision sequencing
- Evidence pack finance and engineering can review together
Three tiers
Scoped to what you actually own.
Tiers differ by boundary and by evidence obligation, not by feature gating. Every tier produces the same artifact: a current exposure figure with its assumptions attached.
Program
One program, one integration boundary.
- Single evidence graph, re-run each integration cycle
- Exposure and schedule band, tracked against the assessment baseline
- Remediation portfolio refreshed as the program changes
- Decision record with owners and review dates
Portfolio
Several programs sharing platforms, suppliers or teams.
- Everything in Program, across multiple boundaries
- Cross-program coupling — the debt that belongs to nobody
- Comparative ranking, so capital goes where return is highest
- Shared-supplier and shared-platform exposure surfaced explicitly
Regulated
Programs carrying certification obligations.
- Everything in Portfolio
- Scored against DO-178C, AS9100, MIL-STD-882, ISO 26262 or IEC 62304 as applicable
- Evidence provenance retained to audit standard
- Handling, residency and retention agreed before any material moves
What Radar does not claim
Radar is in private preview and has no deployment benchmark of its own. It therefore makes no quantitative performance claim — no accuracy figure, no percentage improvement, no customer count. The research figures behind the classification method describe the method's reliability, not this product's performance, and are reported as such in the methodology brief. Radar scores against the certification regimes named above; it is not accredited by any of them and does not replace a DER, auditor or safety assessor.
Trust architecture
Built for evidence you cannot casually share.
The design-partner release begins documentation-first where required, agrees retention before transfer and keeps a complete record of sources, corrections, assumptions and exports.
Customer-defined retention
Retention and destruction dates are agreed per project, with deletion recorded in the audit trail.
Tenant boundaries
Role-based access, encryption in transit and at rest, and an enterprise SSO/private-deployment path.
Evidence provenance
Every score and decision preserves its source, model version, assumption set and correction history.
No silent certainty
Low evidence coverage lowers confidence. The product can refuse to produce a decision-grade estimate.
Private design-partner release
Start with one decision that matters.
Radar follows an assessment. Bring one system, one deadline and the evidence you already have; the assessment establishes the baseline, and Radar is how you find out whether it moved. Design partners shape the first commercial release.
Start a scoping conversationZOYA Architecture Radar is a product preview and is not currently evaluated, endorsed or recognized by Gartner. Product packaging and terms remain subject to validation.