The jinflow

The jinflow is the principal mental model of jinflow — the sentence, the five names, and the one shape that connects them.

The sentence

Jinflow fosters Progress by gaining Understanding through Observations, Knowledge, and Data.

The sentence is the product, said once. It pairs with the surface tagline (“talk to your data, so it speaks to you”). The analytical model and the marketing surface say the same thing.

The five names

Five names, in the order the sentence reads them:

Name	Role
Observations	what we notice
Knowledge	what we know
Data	what the data says
Understanding	the resolution — findings + verdicts
Progress	what we’ll do

These are not layers, areas, aspects, or pillars. They are simply the five names of jinflow. The product’s shape is the relationship between them.

The jinflow

Understanding sits at the centre. Three names feed it (Observations, Knowledge, Data), one name draws from it (Progress):

Three feeders converging on Understanding. One forward arm to Progress. Drop any one of the five and the shape collapses — the model is held by all five.

How Data works

Data unfolds into three tiers, each producing a distinct artifact:

Tier	Asks	Produces
Signal	What can we detect?	signal_findings__<id> rows — many per tenant
Thesis	Is this happening?	thesis_verdicts rows — one per (tenant × thesis), status ∈ {confirmed, plausible, not_observed, insufficient}
Verdict	Why is this happening?	verdict_findings rows — many per tenant, one per matched root-cause case

Vocabulary trap: a thesis outputs a verdict (a status); the verdict artifact outputs findings (rows). Same word, two layers.

Findings and verdicts both land in Understanding. Findings are atomic — one row, one observation. Verdicts are aggregate — one row per (tenant × thesis), summarising whether the thesis holds. Together they are the truth-bearing band.

Theses and verdicts are pack-shipped activatable patterns — see Inside Knowledge below. Their dbt models run inside the tenant build (so the artifacts live under Data), but the patterns themselves are inherited Knowledge that the tenant engages.

The cascade gate

Verdict findings are gated. Two conditions must hold for a row to fire:

1. Thesis gate — the thesis the verdict is bound to must be confirmed. If plausible, not_observed, or insufficient, the verdict never fires — even if signals scream.
2. Condition gate — the verdict’s own conditions (signal references with field thresholds) must match.

Result: staged disclosure. The audience sees facts before interpretation, interpretation before explanation. The data has to earn its narrative, one tier at a time.

How Observations work

The Observation is the entity. It has state, gets validated, gets explained. Two facets travel with it:

Facet	What it is	Authored by
Explanation	Why the observation is true	a human; composed of one or more Contributing Factors
Contributing Factor	Which Understanding backs the explanation	a human; cites individual findings or verdicts

An Observation enters as a guess. It seeks ratification from Understanding: do supporting findings exist? Does the relevant thesis confirm? Does a verdict whose explanation aligns with the proposed Explanation actually fire?

A Contributing Factor is exactly a citation of a Finding or a Verdict from Understanding. That’s the seam: the narrative grounds itself in analytical evidence by referencing specific Understanding rows.

Inside Knowledge

Knowledge holds two kinds of inhabitants that behave differently:

Knowledge
├── Subject Matter — atomic SME knowledge (always-on)
│   ├── Wisdom (pack)        — facts the pack ships
│   └── 50cents (tenant)     — facts the tenant adds
└── Activatable patterns (pack-shipped, opt-in)
    ├── Theses              — question-shape: "is X happening?"
    └── Verdicts            — root-cause-shape: "why X is happening?"

Wisdom is just wisdom — it’s there, it informs, you don’t “check it out.” You either hold it or you don’t.

Activatable patterns are browseable and opt-in — the pack ships 12 theses for numetrix; a tenant might engage 6 of them. The unactivated ones aren’t wrong; they don’t apply.

Provenance

The structural feature shared across every artifact in Knowledge (and elsewhere in the tenant AFS) is provenance — which of three states the artifact lives in:

• 📦 Pack-original — exactly as shipped.
• ✏️ Pack-modified — tenant edited a pack-shipped artifact (overrode a threshold, rewrote interpretation prose).
• 🆕 Tenant-original — written from scratch in the tenant AFS.

This applies horizontally across artifact types — theses, verdicts, perspectives, signals, dossiers — not just Knowledge items. Every artifact detail page makes its provenance state visible.

How Progress takes the model forward

Progress is downstream of Understanding — directed change, not drift; the engagement surface for jin evolve. Once a thesis is confirmed and its verdicts have fired, Progress has two shapes — one inside the system, one outside it.

In-system Progress — the Sense 19 chain. Truth gets tested before it acts in the world:

Tier	Asks	Produces
Suggestion	What might we change?	The human-readable recommendation that travels in a verdict’s prose.
Intervention	What would that change look like as a deterministic edit?	A diff against the AFS — typed, scoped, reversible.
Scenario	What does the data look like under that change?	A branched AFS with the Intervention applied.
Simulation	What is the delta from the baseline?	A side-by-side comparison of verdict findings, money at risk, severity bands.

A Suggestion is what could be done; a Simulation is what would happen if we did it. The chain is how a leap of faith becomes a measured step.

Out-of-system Progress — governance moves. Interventions in the world (update a contract, file a credential renewal, stage a hiring decision). These don’t live in the data; they produce the next deliveries that Data will see, closing the cycle.

Direction carries the steering of forward motion — the attitude, goals, intent that recommendations / suggestions / scenarios / interventions are validated against. The chicken-and-egg is real: direction can be chosen up-front to filter options, or it can emerge from exploring options. Both readings are honest. The internal anatomy of Progress (which of suggestions and recommendations are peers vs synonyms, where Direction sits relative to them) is its own working session, held open here.

Why this matters

The five-name model solves three problems that analytical platforms historically conflate:

1. No premature explanation. Without thesis confirmation, verdict findings never fire. The audience never sees root-cause speculation about theses the data doesn’t yet support. Understanding is staged.

2. No unfounded narrative. Without analytical ratification, an Observation can’t progress past suspected. The hypothesis stays a hypothesis until evidence arrives. Observations ratify against Data.

3. No siloed knowledge. Knowledge sits as a peer feeder of Understanding and is queryable from any of the other four names. Subject Matter modifies what verdicts can defensibly say AND what observations can defensibly say. Knowledge that lives only in someone’s head doesn’t.

A jinflow pack ships the rules for all five names. A jinflow tenant produces the rows that those rules emit when run against its data. The narrative on top is authored by humans who care about the outcomes — and is verifiable against the analytical evidence underneath.

That is the whole product, in one diagram.