Audience Decode

01 . The premise

Jane Jacobs ended The Death and Life of Great American Cities by borrowing a taxonomy from Warren Weaver. There are problems of simplicity, with two or three variables. There are problems of disorganized complexity, with so many variables that averages describe them well. And there are problems of organized complexity, with a moderate number of variables interrelated into an organic whole, where an average describes nothing. Her charge was that planners had been treating cities as one of the first two kinds, and that this was why their interventions kept failing.

Audience research has the same defect. The dominant instruments treat an audience as a problem of disorganized complexity: take a large population, average it, and report the mean as a segment. That produces a description no member of the audience would recognize, and it cannot answer the question actually being asked, which is what happens when something enters the group.

Decode treats an audience as a problem of organized complexity. A small number of interrelated groups, each carrying rules, with defined relationships between them, and a simulation that moves a message across those relationships. The unit is not a demographic. It is a rule set.

Naming the tension names the need. Boundaries are where the tensions live.

A persona is a set of rules a group runs on, mostly invisible to the people running them. A tension is the gap between two rule sets, either inside one group or between two. Cultural pull is the drift of those rules over time. Everything the instrument computes is computed over those three objects.

02 . The read

The map is built by reading six layers off public material. The layers are a collection protocol, not a theory: they fix what gets looked for, so that two reads of the same audience can be compared.

1. LexiconHow they talk. Signature words carrying in-group meaning, in-jokes, shibboleths, and the tells that mark an outsider using the words wrong.
2. CanonWhat they cite. Creators, objects, reference texts, and which brands have been earned in as against which read as trying.
3. Values and statusWhat earns standing in the group, what reads as trying too hard, and the gap between self-image and the outside view.
4. Platforms and behaviorsWhere and how. The role each platform plays, the native rituals, the formats that belong to the group rather than to the feed.
5. BoundariesIn against out. Recognition signals, gatekeeping, and the adjacent groups the audience defines itself against.
6. TrajectoryRising and dying. Ascending, peaking, or already faded, and how durable the rule is that is doing the moving.

fig 02 · six layers into three primitives · drawn from the sentence below it, and from the definition in §01 · the boundaries layer reaches up two rows because the figure keeps the protocol's numbering rather than regrouping it

The layers resolve into the primitives. Lexicon, canon and boundaries give rule sets and outsider tells. Values and status give rule sets and tensions. Platforms and behaviors give the digital footprint. Trajectory gives cultural pull. Two of those five are carried rather than computed over: the instrument's arithmetic runs on rule sets, tensions and cultural pull, and on nothing else.

The six layers in full, with the codebook and the questions each one asks, are at the six-layer method. It is the same protocol written for a reader rather than for a peer.

03 . What gets computed

Two scores carry most of the weight, and both are composite indices. The formulas are printed here because a score whose construction is hidden cannot be argued with.

Stated plainly, because it is the first thing a careful reader will ask: a_i and w_i are authored. An analyst assigns them against the rule sets. The scores are therefore an ordinal judgment carried through arithmetic, and they inherit every weakness of that construction. The band a persona falls into is more robust than the point score, because a band survives assumptions the score does not. Credible reach sits downstream of the +30 threshold, which is a chosen number and not a measured one.

The simulation pass seeds the quantified personas with a message and runs a cohort cascade: adoption spreads by threshold contagion across affinity edges, hostility spreads across tension edges. Its output is a spread of outcomes across runs, never a single trajectory, and it is read for relative comparison between two messages or two plans rather than as a forecast of either.

fig 03 · the cascade · the mechanism in three panels and its output in the fourth · identical worlds are known to produce different hits, which is why the deliverable is a width and not a line (Salganik, Dodds and Watts, 2006)

04 . The measurement ladder

Every parameter on the map carries a rung. The rung is not a confidence adjective, it is a statement about where the number came from and what would have to happen for it to move up.

The harness for the fourth rung is Hindsight, the instrument that reconstructs a cultural episode already at full size. A parameter graduates only against a held-out episode. The first pass of that method was run against a product of the author's own that failed in the market, with the calls sealed in a commit before the answers were opened. That is one self-administered trial, not a calibration, and it is described here as the former.

05 . What this does not claim

• It is not a panel. Share figures are estimates over a scoped audience. They are not sampled, memberships overlap, and they need not sum to one hundred.
• It does not observe individuals. Collection is organizational and public account level. Person level data does not enter the system.
• It does not forecast. The cascade returns a distribution across runs and is valid for comparing options, not for predicting an outcome.
• Coverage is partial and the frame is stated. A scan names the sources it ran and is silent on the rest. Silence in the harvest is not evidence of absence.
• The evidence never supplies the rules. The harvest supplies material. A person authors the rule sets, and that authorship is marked wherever it appears.
• Thin evidence abstains. Where the material will not carry a claim, the instrument declines to score rather than producing a number with a caveat attached.

06 . The layer under this one

Everything above states what the instrument assumes and what it declines to claim. The stage by stage mechanism sits at /decode/method/: the pipeline and its gates, the precondition each of the six layers needs before it can say anything, the arithmetic printed in full, the data contract and the rules a read has to pass to ship, the constants with the file that owns each one, and the stages that are not built with what each one blocks. A read rendered from that contract is at /decode/field/.