See what is working in your market

A loud example is still only an example

A feed can make repetition look like proof. This method turns a market question into five inspectable records: a research contract, a source register, a claim ledger, a contradiction register, and a dated synthesis. The source, the observation, and the decision stay separate.

The result is a bounded evidence set that another person can audit. It tells you what the sample supports, what remains disputed, and when each conclusion needs another look.

Raw input and tangible output

Raw input

• One decision the research must inform.
• One market and observable behavior stated in plain language.
• A fixed date window and a fixed collection frame.
• Public sources the researcher is permitted to inspect.
• An evidence threshold chosen before collection begins.

Tangible output

The method produces five records:

1. A research contract that fixes the question and stop rule.
2. A source register with dates, origin, locators, and limits.
3. A claim ledger with supporting and opposing source IDs.
4. A contradiction register with an explicit disposition.
5. A dated synthesis with a freshness queue and next decision.

Run the method

1. Write the decision and the narrow question it depends on.
2. Freeze the collection frame before opening the first source.
3. Capture one source record at a time and trace each item to its origin.
4. Split every observation into an atomic claim that can be supported or opposed.
5. Classify the evidence and assign one of four claim states.
6. Open a contradiction record whenever credible sources disagree.
7. Write the synthesis from the ledger, with source IDs beside every material statement.
8. Set review dates and remove stale claims from current decisions.

Each step leaves a record. A researcher should be able to stop after any step and resume from the files without reconstructing intent from memory.

Set a bounded collection frame

A useful research question names five things: the market, the behavior, the channel, the time window, and the decision. Write it before searching.

For [market], what evidence shows [observable behavior]
in [channel] during [date window], so we can decide [decision]?

Define the collection contract beside the question:

• Inclusion rule: the exact conditions a source must meet.
• Exclusion rule: the conditions that disqualify a source.
• Source families: the distinct kinds of evidence required.
• Quota: the number of slots available in each source family.
• Stop rule: the date or filled-slot condition that ends collection.
• Evidence threshold: what a claim must show before it can enter the synthesis as supported.

A usable starter frame has twelve slots across four source families:

1. Three primary market artifacts, such as public offer pages, public pricing pages, or published campaign artifacts.
2. Three pieces of public buyer language, with personal data left out of the record.
3. Three independent analyses or datasets that state their method.
4. Three deliberate counterexample sources chosen to attack the starting assumption.

Twelve is a control default for this manual starter. It does not make the sample representative. Leave an unfilled slot visible when a qualifying source cannot be found. Do not widen the market, window, or inclusion rule during collection.

Record provenance before interpretation

Give every source a stable ID. The source register holds:

• Source ID.
• Title, publisher, and public locator.
• Source family.
• Original source or derivative source, plus the root source ID.
• Publication or update date, when available.
• Capture date.
• Exact page, section, timestamp, or short excerpt that carries the evidence.
• Direct observation.
• Claim made by the source.
• Scope limit, incentive, or method note.

Trace reposts and summaries back to the earliest available origin. Sources that repeat one press release, dataset, or study share one evidence lineage. Record that lineage instead of counting each link as independent confirmation.

Use evidence types rather than a hidden score:

• Direct artifact: the researcher can inspect the market object itself.
• Published measurement: a source states its population, method, and result.
• Self-report: a party describes its own result or practice.
• Interpretation: an author draws a conclusion from other material.

The evidence type controls the claim you may make. A direct page supports a claim about what appears on that page. It does not support a revenue claim. A self-report stays labelled as self-report unless independent evidence supports the same outcome.

Turn evidence into a claim ledger

Write one atomic claim per row. If a sentence contains two facts, split it into two claim rows.

Each claim row holds:

• Claim ID and exact wording.
• Market, channel, and date scope.
• Evidence type.
• Supporting source IDs.
• Opposing source IDs.
• Claim state.
• Reason for the state.
• Checked date, review date, and refresh trigger.
• Decision that may use the claim.

Use four states:

• Supported: the claim meets the threshold fixed in the research contract.
• Contested: credible evidence materially disagrees.
• Unconfirmed: the available evidence is indirect, incomplete, or below the threshold.
• Stale: the review date passed or a named condition changed.

Claim state governs what may enter the brief. It is not a declaration of permanent truth. Outcome claims require outcome evidence that matches the population, measure, and time window. Repeated messaging supports a messaging observation only.

Keep contradictions open until the evidence resolves them

Create a contradiction record as soon as two credible sources cannot support the same claim. Record:

• Contradiction ID and claim ID.
• Sources on each side.
• The exact point of disagreement.
• Definition, population, channel, date, method, and incentive differences.
• Current disposition.
• Evidence that would resolve the conflict.
• Review date.

Test the conflict in this order:

1. Normalize the terms. Two sources may use the same word for different measures.
2. Match the population and channel. Evidence from one segment may fail to answer a claim about another.
3. Align the dates. A real change over time can create an honest conflict.
4. Trace shared origins. Several apparent sources may descend from one record.
5. Compare methods. An artifact, a survey, and a self-report answer different questions.
6. Inspect incentives and missing data.

Close the contradiction by narrowing the claim, choosing the better-matched evidence with a written reason, or leaving the claim contested. Never average incompatible values to make the disagreement disappear.

Write a synthesis that a decision can use

Build the synthesis from current ledger rows. Use this order:

Question:
Decision:
Collection frame:
Sources collected and slots left open:
Supported observations with source IDs:
Contested claims with contradiction IDs:
Unconfirmed claims:
Exclusions and known gaps:
Decision implication:
Next test or collection action:
Checked on:
Review on:

State counts with their denominator. Use wording such as, “Within this frame, 5 of 12 collected artifacts showed the pattern.” Reserve the word working for an outcome claim backed by matching outcome evidence. The brief may still recommend a test when the evidence supports a plausible candidate.

The decision implication should be conditional. It states what the evidence justifies doing next, the limit on that decision, and the observation that would reverse it.

Apply freshness rules to claims

Freshness belongs to the claim row because one source can support claims that change at different speeds. Every current claim needs a checked date, a review date, and an event that forces an earlier review.

Use these starter control intervals:

• Review before every use when the claim affects legal, compliance, safety, or irreversible spending decisions.
• Review within thirty days for live offers, prices, platform behavior, and current campaign artifacts.
• Review within ninety days for current messaging patterns and buyer-language collections.
• Review within one hundred eighty days for stable definitions and methods.

These intervals are operating defaults. They make no claim about how fast a particular market changes. Shorten the interval when the source has no publication date, the market changes quickly, or the decision carries a high cost of error.

Trigger an immediate review when a source is revised or removed, a method or population changes, a credible contradiction appears, or the decision scope changes. Mark the claim stale until it is rechecked. Keep its history in the ledger so later readers can see what changed.

Manual starter

Create one spreadsheet or plain document with five sections. Copy the fields below.

The reader supplies permitted public sources and creates and maintains all five records by hand. This page supplies the method and the fields.

1. Research contract

Decision:
Research question:
Market:
Observable behavior:
Channel:
Date window:
Inclusion rule:
Exclusion rule:
Source families and slot count:
Evidence threshold:
Stop rule:

2. Source register

Source ID:
Title and publisher:
Public locator:
Source family:
Original or derivative:
Root source ID:
Published or updated:
Captured:
Evidence locator or short excerpt:
Direct observation:
Source claim:
Scope or method limit:

3. Claim ledger

Claim ID:
Exact claim:
Scope:
Evidence type:
Supports:
Opposes:
State:
State reason:
Checked on:
Review on:
Refresh trigger:
Decision use:

4. Contradiction register

Contradiction ID:
Claim ID:
Sources for:
Sources against:
Exact disagreement:
Mismatch check:
Current disposition:
Evidence needed:
Review on:

5. Dated synthesis

Question and decision:
Frame and stop rule:
Source count and open slots:
Supported observations:
Contested claims:
Unconfirmed claims:
Exclusions and gaps:
Decision implication:
Next test or collection action:
Claims due for review:
Checked on:
Review on:

Use one source row per source and one claim row per claim. Cite source IDs in the claim ledger and claim IDs in the synthesis. That chain is the audit trail.

The freshness queue is the claim ledger sorted by Review on. Copy the due claim IDs into the dated synthesis.

Worked example: synthetic structure only

All names, dates, observations, and counts in this example are invented to demonstrate the record structure. They carry no market evidence or outcome claim.

Contract

• Decision: choose one proof format for a controlled service-page test.
• Market: fictional scheduling consultancies serving independent repair shops.
• Observable behavior: public use of response-time proof or process-visibility proof.
• Channel: public service pages and public buyer discussions.
• Window: fictional dates from May 1 to June 15, 2026.
• Frame: eight slots, with four provider pages, two buyer discussions, and two independent field notes.
• Stop rule: June 15 or all eight slots filled.
• Threshold: a messaging observation needs the collected artifact count; a buyer-preference claim needs a matched method and population.

Source register excerpt

• S-01, S-02, and S-03 are fictional provider pages that display a response-time promise.
• S-04 is a fictional provider page that displays a process diagram and no response-time promise.
• S-05 is a fictional public buyer discussion in which speed is described as important.
• S-06 is a fictional field note in which process visibility is described as more useful than speed.
• Two frame slots remain empty and stay recorded as gaps.

Claim and contradiction records

• C-01: “Response-time language appears on three of four collected provider pages.” State: supported as an observation about this invented frame. Sources: S-01 through S-04.
• C-02: “A response-time promise increases enquiries.” State: unconfirmed. The fictional records contain no conversion evidence.
• C-03: “Buyers prefer speed over process visibility.” State: contested. S-05 supports the claim and S-06 opposes it.
• X-01 records the C-03 conflict. The populations and collection methods do not match, so the disposition stays contested. A matched buyer study would be needed to resolve it.

Synthetic synthesis

Within this invented frame, response-time language appears on three of four provider pages. That count supports a messaging observation only. Buyer preference remains contested, two collection slots remain open, and the frame contains no outcome evidence.

The bounded decision is to test a defined response-time proof against a process-visibility panel. The test needs its own measurement plan before either format can carry a performance claim. Claims drawn from the provider pages receive a thirty-day review date, while claims drawn from the discussion and field note receive a ninety-day review date.

Close the research only when it passes these checks

• The decision, question, frame, and stop rule were written before collection.
• Every synthesis statement traces to claim IDs and source IDs.
• Derivative sources share one lineage and do not inflate the source count.
• Observation, self-report, interpretation, and outcome evidence remain distinct.
• Every material contradiction has a record and disposition.
• Open slots and excluded sources remain visible.
• Every current claim has a checked date, review date, and refresh trigger.
• Stale claims are absent from current decision implications.
• A second reader can reconstruct the conclusion without asking what the researcher meant.

Use only sources you are permitted to inspect. Keep personal data and private material out of the public ledger. The method ends with a dated, reviewable decision brief.