● the published method

The Six‑Layer Decode

How the studio reads an audience: from the decision on the table, down into the rooms where the audience actually lives, and back up to a move you can hold us to.

published 2026-07-23 · about an 8 minute read · report contract locked at v1.0 on 2026-07-21 · nothing on this page is a trade secret; the moat is the discipline

01 · the premise

Every audience runs on rules nobody wrote down

What earns status in one room reads as trying too hard in the next. A word that signals belonging in one crowd marks you as a tourist two tabs over. These rules are invisible to the people inside them and expensive to violate from outside, and most audience research never touches them: it describes an audience from the outside in, as age bands, interest categories and sentiment scores.

The decode works from the inside out. A persona, here, is a rule set: the small grammar of invisible rules a specific crowd actually runs on, with evidence attached to every rule. The gaps and collisions between rule sets are tensions, and tensions are where demand lives: naming the tension names the need. The whole method exists to surface both, and to say plainly how sure we are about each.

named lineage, not vibes

The fieldwork is netnography. The reading of signs is commercial semiotics. The business bridge is jobs-to-be-done. The quantitative layer descends from generative social science (Epstein and Axtell, Sugarscape, 1996) and from the threshold and contagion models the empirical literature says actually drive adoption (Granovetter, 1978; Centola and Macy, 2007). Full citations ship inside every report's evidence section, not in the sales copy.

02 · the framework

Six observational layers, one codebook

Every persona on the map is read through the same six layers, in order: from the surface of the language down to where the crowd is headed. Each layer condenses into a specific part of the finished map, so nothing in the read is decorative.

Lexicon

How they talk: signature words with in-group meaning, in-jokes, shibboleths, the tells that mark an outsider. Language is the cheapest authenticity test, so this layer ships as four acceptance bands per persona (accept, conditional, emerging, reject) plus a collision table for words that land in one room and burn in the next.

condenses intorule sets · outsider tells

Canon

What they cite: the creators, objects and reference texts that carry weight in the room, and the brands that are earned-in versus cringe.

condenses intorule sets

Values & status

What earns standing, what reads as trying too hard, the sacred cows. The semiotics pass works hardest here: the status behavior is the sign; the invisible rule producing it is the code.

condenses intorule sets · tensions

Platforms & behaviors

Where and how the crowd gathers: each platform's role, native rituals and formats, who creates and who only watches. The layer that reads most directly from the studio's own live harvest.

condenses intodigital footprint

Boundaries

In versus out: recognition signals, gatekeeping, and the adjacent groups the crowd defines itself against. Boundaries are where the tensions surface first.

condenses intorule sets · outsider tells

Trajectory

What is rising and what is dying: codes still ascending, codes about to peak, codes already cringe, and where the crowd is migrating next.

condenses intocultural pull

fig 01 · the six layers and what each condenses into on the finished map

03 · the pipeline

From a filed decision to a shipped read

A decode is not a vibe check with charts. It is a fixed sequence, and every step leaves a trace that can be audited in the finished report.

0the brief→ 1immersion→ 2personas→ 3the six layers→ 4tension edges→ 5quantification→ 6the report

fig 02 · the decode pipeline; each step below is auditable in the shipped report

The brief

Every decode opens from a business decision, not a subculture safari. The client files the decision on the table, the wager behind it, and the bar a useful answer has to clear. The map is only ever interrogated through that decision.

Immersion in the micro-communities

A scoped scan of the rooms where the audience demonstrably gathers. It starts from a named anchor set: accounts, publications and rooms where the audience is known to collect. The graph grows outward from those anchors, and the studio's own daily harvest, running since 2026-07-09, is filtered to the same surfaces. Every anchor is named in the report, with why it qualifies.

Synthesis into personas

The micro-communities condense into personas: each one a rule set with evidence, never a demographic sketch. Memberships overlap by design, because real people hold several rule sets at once. Each persona is also marked held (already inside the client's audience) or prospect (the audience the decision needs), so the map answers both intake questions at once.

The six-layer decode, per persona

The six layers above, run on every persona. This is the step the market skips: the graph work tells you who travels together; the decode tells you why they behave.

Tension edges

Rule sets are collided pairwise, and the gaps between them become named, scored tensions. A sharp tension between large, strengthening personas is the most valuable unowned ground on the map, and the report ranks it as such.

Quantification

The map becomes numbers with provenance: shares, sensitivities, thresholds, channel weights. Every value carries its rung on the measurement ladder printed below, and a parameter without an evidence hook does not get invented to make the model feel complete.

The report

What ships follows the v1.0 report contract, locked 2026-07-21: five sections in fixed order (01 the brief, 02 the read with the recommended move as a boxed block, 03 the map, 04 the climate, 05 the crowds), and inside every crowd four moves in fixed order: the code, the language, the footprint, the evidence. Underneath runs the report's analytical spine, §0 through §8, locked 2026-07-20: decision first, answer up front, then the map, the tensions, the trajectory, the footprint, the commercial translation, and the full evidence log. Every recommendation closes with a falsification metric: the named observable that tells the client, within 4 to 12 weeks, whether the read was right.

04 · the named metrics

The numbers, defined in the open

The field names its numbers, so we name ours. Five metrics recur across the studio's surfaces. Here is what each one actually measures, where it lives, and what it is not allowed to claim.

Thread weight

home · the loom, readtheloom.live

impact = volume × resonance

Each thread on the weekly weave carries a weight. Volume is the count of items on that thread across the studio's tracked feeds, the daily harvest running since 2026-07-09 (51 registered sources, 49 active, at the 2026-07-23 registry count), with each source's daily contribution capped so the loudest feed cannot impersonate the culture. Resonance is how strongly the thread's domain amplifies the live tensions crossing it. A heavy thread means many rooms are carrying it and the talk lands on a live tension, not merely that one source is shouting.

Climate scores

home · the loom's monitor

deviation = (value − baseline mean) / baseline std

A climate score says how far a domain's conversation currently sits from that domain's own recent baseline, in standard deviations: a z-score, computed daily from the harvest over 1, 3 and 7 day windows, live since 2026-07-16. The claim is always deviation, never size: louder than its own normal, not big. A score is labeled thin until its baseline is at least 14 days deep, then graduates to measured automatically. Nothing graduates by hand.

Message Credibility Score

home · decode

per persona, −100 to +100 · lands ≥ +30 · clocked ≤ −30

A proposed message is scored against each persona's rule set: every rule rates the message from −2 (violates) to +2 (aligns strongly), weighted by the rule's salience, and the total is normalized to a scale from −100 to +100. What ships is the band, not the bare number: LANDS at +30 and above, INERT in between, CLOCKED at −30 and below. CLOCKED is the band that matters most, because a violated rule does not just miss; it marks the sender as an outsider, at any budget. Scores stay banded until rule-weight calibration is earned on the record.

Whitespace Score

home · decode

W = 100 × intensity × reach × pull

Every tension on the map gets a whitespace score: intensity is how sharp the tension is, reach is the combined share of the personas sitting on it, pull is whether those personas are strengthening or fading. The result ranks positioning territories numerically, so the sharpest unowned ground between large, strengthening crowds ranks first. It exists so the positioning answer is a ranking, not an adjective.

Cascade outputs

home · decode's simulation pass

many stochastic worlds · the spread, never one trajectory

Seed the quantified map with a message and a media plan and run it forward as many stochastic worlds, not one. Adoption travels along affinity edges once enough of a persona's trusted network has moved; hostility travels along tension edges. The deliverable is the spread across runs: which message, which plan, which crowds move first and which pile on. Never a single predicted trajectory, because identical worlds have been shown to produce different hits (the MusicLab experiment: Salganik, Dodds and Watts, 2006). Relative outcomes only, until calibration is earned.

05 · the standard

What a claim must survive before it ships

The method above is publishable because none of it is the moat. The moat is the discipline. Three commitments hold on every read, and each one is checkable in the sample report.

Every claim cites its evidence

A rule enters the map only with a primary-source link and a date. No source, no rule. The report's evidence section logs every filed link that was read and what it yielded, so the research can be re-walked, not taken on faith.

Every claim carries its confidence

Two separate axes, never collapsed into one. Numbers sit on the measurement ladder below, which says how each value was produced. Interpretive readings, the rules and the tensions, carry their own low, medium or high confidence and never auto-graduate on volume. In the visuals the same honesty is drawn, not footnoted: solid means measured, dashed means directional, red means a gap.

Thin evidence abstains

Where the record is thin, the read prints abstained instead of guessing; an honest empty cell outranks a confident wrong one. And every recommendation ships with a falsification metric: a named observable that shows, within 4 to 12 weeks, whether the read was right. Resolved metrics feed the calibration loop. A read you can check is the product.

rung	meaning
directional	analyst synthesis, dated and labeled as such; no live feed computes it yet
thin	computed from a live feed, but the baseline is under 14 days deep
measured	computed from a live feed with a baseline of 14 days or more
calibrated	survived backtesting against held-out episodes it was not tuned on
abstained	the evidence was not there, so no claim was made; printed, never hidden

fig 03 · the measurement ladder; every number on every studio surface carries one of these labels

One absence is deliberate. You will not find an accuracy percentage anywhere on this site. A vendor-claimed accuracy number is marketing. A per-claim confidence label, plus a falsification metric you can hold us to, is a method.

06 · see it working

The fastest way to judge a method is to read its output

A finished sample decode is public, structured exactly as this page describes: decision first, six layers per persona, banded credibility, evidence log, abstentions printed. When you have a decision worth a read, the desk is open.

read the sample decode → bring a decision to the desk →