The Assay™  ·  A controlled measure of machine-mediated representation

Are you the answer? Is the reason yours?

The Assay measures Retrieval and Fidelity inside a frozen decision space, then shows where the record, evidence or underlying reality breaks.

What is measured

Rebel.Rebel measures two things, and calls them Retrieval and Fidelity.

Retrieval is whether your company becomes the answer a machine gives when a person is making a consequential decision. Fidelity is whether the reason that machine gives is specific to you, accurate, and traceable to something you actually published.

Both are scored by The Assay™, the instrument this practice runs, inside a decision space frozen before any answer is logged. A company can hold Retrieval and fail Fidelity: the machine names you, then explains you in language that would fit any competitor.

The event

The question stopped dressing up.

Search asked people to compress their lives into keywords. AI can accept the circumstance in full. The machine is no longer only retrieving pages — it is interpreting a need, assembling evidence and producing a recommendation in language nobody at the company wrote.

The Assay measures that representation.

Why it is called an assay

A coin can look like gold and contain less gold. A sentence can look like brand language and contain nothing that belongs to the brand.

An assay tested the purity of the thing carrying value through a marketplace. It existed because coin circulates on trust, and nobody weighs it twice.

The Assay™ tests the purity of the sentence carrying preference through the market.

Retrieval

Whether you arrive.

Fidelity

The purity of what arrives.

Not purity

Not whether the machine repeated your copy. That would be a weak test, and a brand can be described accurately in words that belong to no one.

The sentence  ·  a hypothetical circumstance

“I’m 42, getting back into running after knee surgery, heavier than I used to be — and I don’t want a shoe that feels orthopedic.”

Brand A — retrieved without fidelity

“Brand A is a popular option for runners looking for cushioning and support.”

Retrieval
named among options
Fidelity
category language. Nothing in the reason belongs to it.

Brand B — retrieved with fidelity

“Brand B’s trainer pairs a high stack with a wide, stable base — and its published fit notes address heavier runners directly.”

Retrieval
the first recommendation
Fidelity
specific and attributable. The reason is its own.

Why Brand A lost

The machine could retrieve the name. It could not retrieve a reason to choose it.

Illustration only — hypothetical brands. No wave data shown.

The scores

Retrieval

Are you the answer?

Whether you become the answer in the relevant consideration circumstance.

Fidelity

Is the reason yours?

Whether the reason given remains specific, accurate and attributable to you rather than collapsing into category language.

One answer is an anecdote. The Assay is an instrument.

The instrument

The battery is not written. It is found. Before anything is sealed, the house does the qualitative and ethnographic work with the people who actually make the decision: how they phrase the question, what they are afraid of, what they have already tried, and what they would need to hear to choose. The buyer profiles and the questions come out of that fieldwork, which is why no two batteries are alike and none of them can be prepared for.

The thinking behind the battery is published, and it is not a secret; the questions are. A battery is written for the second and third question, not the first, because the machine asks again and the record has to survive being asked twice (RR-P-2026-001). It is written in the buyer’s own words rather than the category’s, because the distance between how a person phrases a need and how a search box taught them to phrase it is where a brand goes missing (RR-P-2026-006). And it is written for the unguarded question, the one a person asks an assistant that they would never have typed where anyone could see it (RR-L-2026-009). Read the papers. The battery is what they look like applied.

  • A frozen decision space

    Questions built from research with your buyers, in the words they actually use, then locked before outcomes are observed. Never after. You shape who is asking; the house writes what is asked, and does not show it.

  • Multiple systems, repeated runs

    Controlled sessions across the machines people actually use.

  • Competitive controls

    A frozen peer set: proposed by you, confirmed by the house, sealed before the run. Movement is read peer-relative, never in a vacuum.

  • Independent scoring

    Trained readers, versioned rubrics, documented disagreement.

The instrument is versioned. The ruler does not move because the answer disappoints us.

The disposition

Every baseline ends with a disposition.

  • None

    No action is warranted.

  • Bounded

    A defined weakness warrants limited action.

  • Structural

    The failure spans record, evidence, authority or underlying reality and may require sustained work.

A diagnostic that cannot legitimately say “None” is not independent.

What you hold

The diagnosis, and the prescription. Not the treatment.

The scores

Retrieval and Fidelity, banded, with the readers’ agreement rate printed beside them.

The disposition

None, Bounded or Structural. The engagement can end here, and sometimes should.

The editorial prescription

Three things, in order: the gaps the instrument found, the answers the record already contains, and how we recommend you close the distance. It names the failure and the evidence that has to travel, and it does not name a format, a channel or a campaign, because the format comes after the gap. It ends by setting out the routes open to you, ours among them and yours first.

The ledger and the run log

The one-page ledger, and every answer the machines gave with every code the readers assigned, yours permanently. The questions themselves are not in it. The battery and the method stay with the house, because an instrument anyone can prepare for is not an instrument.

What a baseline does not include is authored work of any kind. Not a page, not a line of copy, not a position. Editorial is commissioned separately, by retainer or by the piece, and a prescription is just as valid handed to the agency you already use, or to no one at all.

The routes it sets out are these. Your own internal teams or agency partners can act on it as written — that is what portable means. Standing Wave™ re-runs the same battery so the repair is measured rather than assumed, which is why acting on the prescription is the condition of buying one: a re-measurement of a record nobody changed is a measurement of the weather. Editorial can dig out a position and hold governance and a circulation plan for whoever executes. Fractional leadership or project work is available where a team needs hands rather than a standard.

A prescription written by someone who is paid to fill it is not a prescription. That is why the two are priced apart and invoiced apart, always, and why the first route on that list is the one that pays this house nothing.

The instrument diagnoses and prescribes. It does not treat.

Standing Wave™

Run the sentence again.

Standing Wave reruns the same decision space after the repair. The market can change. Models can change. The instrument records those changes rather than pretending time stood still.

The question is not whether a dashboard moved. It is whether the representation moved in the decision space that mattered.

The delta

Two readings of the same frozen battery, with the category’s tide subtracted before any boat claims to have risen. The number between them is what the wave produces, and it is the number you can carry upstairs.

Observed, not surveyed

A net promoter score asks people who already bought how they feel about it. The wave watches the machine decide, in the sentence it gives to someone who has not bought yet.

Leading, not lagging

Return on ad spend reports what happened after money moved. The wave reads the consideration before the money does, which is the only place a brand can still be changed.

Placed where the decision now happens

More buying begins with a question to an assistant than with any other single act, and the share grows every quarter. The wave is the only reading taken at that moment, on a frozen instrument, against control.

The delta claims that the record moved, measured against control. It never claims that revenue followed. That restraint is what makes it a number a chief financial officer can sit beside their own without either contaminating the other.

The obvious question

“Can’t we just ask the machines ourselves?”

Yes. You should. One answer is useful evidence. It is not an instrument.

An instrument freezes the decision space before outcomes are observed, repeats controlled runs across multiple systems, scores against a frozen peer set, reads every answer independently, logs everything — and preserves the same ruler so movement can be measured instead of remembered.

One answer is an anecdote. The Assay is an instrument.

What we publish, and what we do not claim

A frozen decision space. Multiple AI systems. Repeated controlled runs. Competitive controls. Independent scoring. When the record changes, Standing Wave runs the same decision space again. The exact questions remain sealed; controlled method inspection is available where a mandate requires it.

The claims stop where the evidence stops.

No revenue causality without commercial evidence and design. No precision claims before a declared agreement statistic. No universal battery before the evidence earns it. No moved ruler — version changes are disclosed, or the break is declared.

The sequence

A scout reads the ground before the survey.

Most companies should not begin with research. They should begin by finding out whether research is warranted — which is a different question, answered a different way, at a different cost.

Field Read
judgment

A bounded, paid first read of a company, brand, category or strategic question, from the outside in. It answers one thing: is there anything here worth measuring? It makes no claim to Retrieval or Fidelity scoring, statistical authority or longitudinal measurement — and it can conclude that no material problem exists. That conclusion is the product working, not the product failing.

The Assay™
measurement

The instrument run. A frozen decision space, repeated controlled runs across multiple systems, competitive controls, independent scoring. This is the baseline, and it is the only thing here that produces one. It ends in a disposition and an editorial prescription.

Editorial
commissioned separately

Repair what the prescription warrants. A missing record can be authored. A missing fact cannot. Editorial is the House’s authorship arm and is bought on its own terms, by retainer or by the piece — or the prescription goes to whoever already writes for you. Where the evidence needs to travel, Editorial builds the owned, earned and paid circulation plan and works beside the people responsible for putting it there.

Standing Wave™
proof

The same decision space, run again, so movement is measured rather than remembered.

The Field Read is judgment. The Dig is evidence. The Assay is measurement. Neither is dressed as the other.

The Field Read  ·  $2,500 The Dig  ·  $7,500 Request the rate card How the instrument works

Different layer  ·  different question

Three disciplines meet the same machine. Only one of them asks whether the answer is yours.

Dimension RAG GEO / AI visibility The Assay™
What is it? A technical retrieval architecture. A visibility / optimization discipline. A representation measurement and diagnostic.
Primary question What information should the system retrieve? Are we appearing, cited or visible? Are we the answer — and is the reason ours?
Unit of concern Sources, chunks, retrieved context. Mentions, citations, visibility surfaces. Sentence, representation, decision circumstance.
Starts with System architecture. Discoverability and content visibility. A consequential decision.
Reality gap Not its primary job. Not usually its primary job. Can identify when the desired sentence outruns the underlying evidence.
After the finding Improve the retrieval system. Improve visibility / discoverability. Change the business, evidence, authority, editorial or circulation — depending on what failed.
Longitudinal proof System evaluation. Visibility monitoring. Standing Wave reruns the same decision space.

These are not mutually exclusive. RAG helps systems retrieve. GEO helps content become visible. The Assay asks whether the representation produced under real consideration is actually yours.

The objection

The purity of metal is a physical fact. This is a judgment. It is.

The assay office never answered that by pretending the test was mechanical. It answered it by making the test independent, repeated, and its verdict public.

Independent

Read by three people, not scored by a model. Two of the three do not know which house they are reading, or why. Where they disagree, a third settles it, and the rate at which they disagreed is reported with the result.

Repeated

A frozen battery, written once and never edited between runs. Change the record, not the test.

Public

An index published rather than sold. The architecture is disclosed. The questions are not, because an instrument anyone can prepare for is not an instrument.

The answer is institutional, not mechanical, and it always was.

Questions

What exactly do you measure?

Retrieval: whether you become the answer in the relevant circumstance. Fidelity: whether the reason given remains specific, accurate and attributable to you.

Is The Assay GEO?

No. GEO and AI-visibility work generally focus on appearing, being cited or becoming more visible in generative answers. The Assay starts with the decision and measures whether you become the answer — and whether the rationale remains yours. GEO may become useful after the diagnosis.

Is this RAG?

No. RAG is a technical architecture for retrieving external information into a model’s generation process. The Assay measures the market-facing representation produced by the systems people already use.

How reliable is a human-scored instrument?

Meaning requires judgment, so the method does not pretend humans can be removed. The instrument uses trained independent readers, versioned rubrics, calibration and documented disagreement. Reliability is a method requirement, not a marketing claim.

What if our scores are already good?

We say so. Every baseline can conclude None. If no action is warranted, the engagement ends there.

What if the problem is not our content?

Then we do not prescribe content. Sometimes the record is weak. Sometimes the evidence is missing. Sometimes the business has not earned the desired sentence yet. The repair follows the failure.

Does this prove revenue impact?

No. The Assay measures representation. Standing Wave measures whether that representation moved under the same decision space. Revenue or valuation causality requires separate commercial evidence.

Do you publish client findings?

No, not by default. Client work is private. Touchstone™ is the independent public research program. Rebel.Rebel Press is the House’s publishing imprint.

We hold a portfolio. Where does this fit?

Funding is one of the decisions machines now sit inside. When the decision is diligence, a hold, or an exit, the object is the same one we always measure: the representation a machine produces about a company under real consideration, and whether the reason given is the company’s own. A thin record is a thin record whether you are buying the company or buying from it. That route is private rather than a published product — bring the decision through the investor line and we will tell you whether the instrument fits.

How much of the method do you disclose?

The architecture in full, the battery never. You shape the instrument at two points: the buyer profiles the questions are written to, and the peer set you are measured against, which you propose and the house confirms and seals. You do not see the questions, before the run or after it, because an instrument anyone can prepare for is not an instrument. You keep every answer and every code. Deeper controlled inspection is available when diligence requires it, under terms.

Begin

Bring the decision. Measuring it is our job.

We will tell you whether the instrument fits, and if it does not, we will say so.

The Field Read  ·  $2,500 Contact