# LinkedIn Apex Research

Research date: 2026-09-21

Status: `VERIFIED` for bounded public first-party platform guidance and signed competitor discovery; `PARTIAL` for usable candidate observations; `NOT EXECUTED` for Saffron account research and matched comparative scoring; `VERIFIED LOCAL` for the preregistered method and matrix completeness.

## Arena

The comparison arena is international LinkedIn company-page and executive-led publishing that helps defined audiences understand, trust, remember, discuss, and act on an organization’s work.

It includes global reach, B2B thought leadership, brand storytelling, recruiting, executive ecosystems, industrial and technology communication, consulting, design, automation, AI, public benefit, regional leadership, criterion specialists, founder-led publishing, owned newsletters, communities, and emerging approaches.

Follower count alone cannot establish leadership.

## Method registered before scoring

Fifteen criteria are defined in `../data/competitors.json`:

1. Brand clarity and memory.
2. Audience relevance and content usefulness.
3. Evidence density and trustworthiness.
4. Editorial originality and recognizable voice.
5. Hooks, narrative, and calls to action.
6. Format excellence.
7. Cadence quality and sustainability.
8. Executive and employee amplification.
9. Community response and comment quality.
10. Localization and cultural relevance.
11. Accessibility and inclusive communication.
12. Conversion architecture and measurement.
13. Governance, review speed, and claim safety.
14. Production efficiency for Mojay.
15. Strategic defensibility and learning velocity.

Each criterion specifies user value, measurement method, direction, unit, matched conditions, minimum meaningful advantage, uncertainty rule, evidence class, and refresh trigger. A WIN requires the stated margin under matched conditions. Missing, inaccessible, incomparable, or stale evidence yields `UNKNOWN`.

## Current route result

A signed Saffron Browser research package completed a bounded first-party LinkedIn platform-guidance pass through `browser_scrape` on `POST /v1/scrape`. The package contains 34 retained entries: 27 `VERIFIED` operations and 7 `ERROR` attempts. The verified operations ran from `2026-09-21T07:05:35.975Z` through `2026-09-21T07:12:15.161Z` and bind each source to a non-secret operation ID, result SHA-256, receipt SHA-256, and S11 policy SHA-256 `958a5b563ef4b5ce99f2bf30190f4a2a2b8534bcbff517d2b06b0f3051904c83`.

Package bindings:

- Evidence index SHA-256: `a32b439a2dd20c5d705a44684194f8b63fe75d6d938574b3b24b0ad210acb6f5`.
- Platform guidance SHA-256: `38ba5428465862d53d6f4c95c2163845a27964756e4085353188410fe2a4a399`.
- Route verification SHA-256: `4f31be8a0b8c507c0b354b6681dc4f2ef6389db5ce8deafe1facc7d446852216`.
- Package manifest SHA-256: `d822d4dab8624b543cfe512d33a48618f48645fbcef9fa8006350d90b15b290b`.

The pass verified public rules, recommendations, definitions, and LinkedIn-reported associations for Page administration, content formats, scheduling, deletion, alt text, captions, video, newsletters, analytics, exports, native competitor analytics, supported languages, targeted posts, and role boundaries. Every integrated statement is identified as `PG-001` through `PG-021` in `../data/platform-guidance.json`, where it resolves to the exact E-ID and operation receipt.

The platform-guidance pass did not inspect any Saffron account, private analytics, published post, candidate competitor Page, matched competitor post sample, audience comment, or regional performance. It therefore does not score a single competitor cell.

A separate signed competitor-discovery pass then attempted all 21 concrete candidates through the approved Saffron Browser route. It retained 60 unique signed `LIVE` operations: 8 substantive public discovery results, 50 challenge or rate-limit responses, and 2 empty shells. Ten tool or client error records, one expected invalid-target rejection, and nine local diagnostics were retained outside the accepted receipt set. Substantive discovery was available for LinkedIn, Microsoft, Salesforce, and HubSpot. It established only dated indexed references and limited surface signals. It did not establish post quality, cadence, accessibility, governance, conversion performance, causal advantage, or a matched comparison. The two substitute candidates require a different evidence design and received no entity-target request.

Competitor package bindings:

- Package manifest SHA-256: `dd3297110eb2c02d094b7eb6fb35476c03cf5fde5e468babea53495a407bd6c9`.
- Structured evidence SHA-256: `6367b7e60e0d59b7c6607819c06c73640deda15d40cd73e0fa99325a20b54d16`.
- Evidence index SHA-256: `6fd81320bb49b207b7ca432342169ed25b804d9f212215ab0d8ee926be5f85f0`.
- Current S11 policy SHA-256: `958a5b563ef4b5ce99f2bf30190f4a2a2b8534bcbff517d2b06b0f3051904c83`.
- Retained package: `output/linkedin-s11-competitor-research/`.

The result is deliberate:

- 23 comparison candidates are retained as a research queue, not declared leaders.
- 345 competitor-by-criterion cells exist.
- All 345 results are `UNKNOWN`.
- Indexed Page references are retained only where the signed public discovery returned them. No sample window, inspected-post count, cadence, growth rate, engagement rate, native-market quality, or causal mechanism is asserted.
- No average, rank, or headline score is calculated.

## Retained local evidence

| Evidence | Date | What it supports | What it does not support |
|---|---|---|---|
| `production/formats.json`, LinkedIn personal cover entry | 2026-09-19 | The local registry recorded 1584 by 396 as the first-party recommendation on that date, with responsive-crop limitations. | Current platform acceptance, universal safe area, or upload success. |
| `production/formats.json`, LinkedIn company Page cover entry | 2026-09-19 | The local registry recorded 1512 by 256 as the first-party recommendation on that date, with responsive-crop limitations. | Current platform acceptance, universal safe area, or upload success. |
| Signed Saffron Browser platform package, `E-003` through `E-032` as applicable | 2026-09-21 | Current public first-party platform rules and LinkedIn-reported associations enumerated in `platform-guidance.json`. | Saffron account state, entitlement, performance, competitor performance, or causal distribution advantage. |
| Signed Saffron Browser competitor package, 60 accepted receipts | 2026-09-21 | Dated route execution, limited indexed references, challenges, empty shells, and explicit candidate-by-criterion UNKNOWN states. | Comparative performance, operating quality, direct post inspection, native-market quality, or leadership. |

No retained local source supports comparative performance or live Saffron page state.

## Comparison set

The queue spans LinkedIn, Microsoft, Salesforce, HubSpot, Adobe, Canva, IBM, Siemens, Accenture, Deloitte, McKinsey & Company, Patagonia, NVIDIA, Anthropic, Aramco, stc, Careem, Zain, The LEGO Group, GitLab, IDEO, founder-led publishing, and owned newsletter or community programs.

Inclusion means only that the candidate could challenge a preregistered criterion. The signed discovery pass verified indexed Page references for four candidates and left the remaining identities or exact qualification `UNKNOWN`. No inclusion establishes leadership.

## Ranked mechanisms to test

These are hypotheses, not findings:

1. **Evidence-rich recurring franchises.** A recognizable structure may reduce production decisions and build expectation. It fails when evidence supply is thin or repetition becomes performative.
2. **Accountable human amplification.** Named practitioners may add trust and distribution. It fails when participation is coerced or every voice sounds ghostwritten.
3. **Platform-native teaching formats.** Format selected for the idea may improve comprehension. It fails when novelty displaces substance or accessibility.
4. **Local relevance from one canonical truth source.** Transcreation may preserve identity while earning regional usefulness. It fails under weak reviewer capacity or entity-status drift.
5. **Closed-loop measurement and reuse.** Preregistered tests may compound learning. It fails when denominators, attribution, or decision rules are weak.

## Proposed Saffron differentiators

Two user-valued systems are proposed for testing:

- One evidence-gated international publishing flow built around a shared claim ledger.
- One command center connecting strategy, assets, locale review, accessibility, governance, and measurement.

No claim is made that these are absent from every rival. Their distinctiveness remains `UNKNOWN` until the dated comparison set is inspected.

## Five future head-to-head milestones

1. Mojay completes the six representative tasks at least 25 percent faster than the retained baseline, with zero added critical errors.
2. A matched 20-post sample exceeds the strongest evidenced rival’s supported-claim rate by at least 15 percentage points, with no critical escape.
3. Saudi Arabic and French native-reviewed samples exceed the strongest evidenced rival by at least 0.5 rubric point in both locales.
4. A matched accessibility suite has zero critical failures and at least a 10 percentage-point pass-rate advantage.
5. A six-month learning-velocity review exceeds the strongest evidenced rival by at least 0.5 rubric point with source-complete decisions.

These milestones do not become WINs until samples, rivals, conditions, and evidence are current.

## Next matched research route

The coverage ledger now records both completed Saffron Browser passes. The next pass must change mechanism rather than repeat challenged search requests: fixed dates, equal post counts, direct retained post and media captures, the same viewports and languages, disclosed denominators, native review for priority markets, blinded evaluation, and independent adjudication.

Each operation must retain UTC time, target or query, approved tool, non-secret operation ID, status, result digest, current policy digest, URL, access date, and limitations.

## Stopping rationale

The competitor pass stopped after every concrete candidate had a signed public discovery attempt and the route repeatedly returned challenge, rate-limit, or empty content. Only 8 operations yielded substantive discovery. Repeating the same mechanism would add request volume without satisfying the matched method. Saffron-account state remains uninspected, and every comparative result remains `UNKNOWN`.

## Refresh trigger

Run a matched candidate pass when direct retained samples, fixed windows, native reviewers, denominators, and independent adjudication are available through the approved route. Refresh an individual platform claim when its source changes, its operating decision depends on unverified entitlement, or the evidence package exceeds the accepted freshness window. Fill the required sample record before changing any matrix result. Preserve every LOSS, ERROR, and UNKNOWN.
