Nishi PM & ROI Dashboard — honest bench vs SOTA

Nishi vs Jellyfish · LinearB · Swarmia · DX (getdx) · Cortex — engineering-intelligence & ROI platforms · 12 axes · presence not depth · liar-killed · snapshot 2026-07-18

~520‰

coverage (presence)

sovereign exceeds

honest gaps → all FILED as F-rungs

ungrounded claims (liar-kill)

Axis matrix

Axis	Nishi	SOTA (Jellyfish/LinearB/Swarmia/DX)	Verdict
Work-liveness / real-time WIP	ws_sync heartbeats + /standup ACTIVE-STALE, human AND machine	poll git/Jira, batch daily	EXCEED
Autonomous agent as tracked worker	zero-Claude cadences ARE tracked work; 328‰ autonomy MEASURED	none track autonomous agents as first-class	EXCEED
ROI honesty tagging	MEASURED / ESTIMATED / ASPIRATIONAL contract built-in	single confident numbers, no provenance tag	EXCEED
Real-time derive vs batch	every gauge derived live from sovereign stores	nightly ETL / batch	EXCEED
Sovereignty / zero-per-seat	self-host bits-up, own TLS, $0/seat	SaaS $/dev/mo, data leaves org	EXCEED
EVM / CPI / SPI / burndown	nishi_project engine EXISTS, currently UNFED (9/31)	LinearB/Jellyfish HAVE	PARTIAL
Investment allocation (new/KTLO/debt)	frontier lanes + debt plane give it structurally	Jellyfish core feature	PARTIAL
Critical-path / CPM schedule	live CPM board (float, ES/EF, crit) MEASURED	Gantt in some; not deploy-linked	HAVE
DORA metrics (deploy-freq/lead/CFR/MTTR)	deploy watchdog only; no git-based instrumentation	FULL DORA is table-stakes	GAP → F740
SPACE / DevEx (satisfaction, flow)	no developer-experience survey signal	DX/Swarmia core	GAP → F741
Cost-per-initiative / R&D capitalization	no timesheet-to-dollar rollup yet	Jellyfish HAVE (finance-grade)	GAP → F742
External source ingest (git/CI/Jira/incident)	own planes only; no third-party connectors	many integrations = the moat	GAP → F743

The 5 sovereign exceeds (genuinely beyond SOTA)

Autonomous-agent-as-tracked-worker — no incumbent tracks machine/agent work as first-class labor; Nishi measures 328‰ zero-Claude autonomy directly. A new category.
ROI honesty-tagging — every number carries MEASURED/ESTIMATED/ASPIRATIONAL provenance; incumbents present un-tagged confident figures.
Work-liveness heartbeats — real-time ACTIVE/STALE from beats, not git-poll inference.
Real-time derive — live from event stores, no nightly batch.
Zero-per-seat sovereign — bits-up self-host, data never leaves.

Beyond-SOTA research radar + fetch spec

Autonomous-worker productivity metrics — define the DORA/SPACE analogue for agent fleets (throughput-per-token, zero-touch-ratio, escalation-rate). Unclaimed frontier.
Real-time EVM from a live event store (vs monthly finance EVM).
FETCH: Jellyfish · LinearB · Swarmia · DX getdx · Cortex — current pricing + feature matrices; the DORA report, the SPACE paper (Forsgren et al.), the DX Core 4. What to look for: which metrics are instrumented from which sources, and the honesty/uncertainty treatment (none tag provenance = our exceed).

Method: capability-presence census across 12 category axes, depth honestly discounted, gaps filed as owned frontier rungs (F740–F743) per the less-than-SOTA-is-never-design law. Live gauges from the pmdash- plane + CPM board + ecomat rollup. Evidence: knowledge/status/roi_rate_research.log (displaced-SaaS, triangulated). Deviations allowed only as measured exceed bets — every gap above is a climb, not a design choice.