Affiliate conflicts & editorial governance

One IndexFair founder is a co-founder of UFFILIATES.me, an affiliate network for licensed gambling and betting operators. Editorial independence is a structural commitment, not an editorial promise. This page sets out the four standing commitments that govern the conflict — at the category level, with no brand lists.

We disclose at the level of verticals and markets which categories of IndexFair’s catalogue overlap with UFFILIATES commercial activity. We do not publish brand-by-brand affiliate-partner lists on this page. The structural firewall — which brands have the affiliate-tagged outbound CTA suppressed in code — is enumerated on the separate editorial firewall page, where the list is queried from the database at render time.

Category disclosure (to be approved by the founder before publication; intentionally left blank in this draft because the exact wording is the founder’s call, not the executor’s): the verticals and country markets where UFFILIATES partner relationships may overlap with the IndexFair catalogue will be enumerated here.

Methodology changes are versioned, dated, and accompanied by a rationale. A new version applies forward from its release; we do not retroactively change parameters (weights, half-lives, thresholds) inside an already-published version. Every substantive change carries a version bump and a public changelog entry at /changelog.

This rule is the structural commitment: it makes any bias detectable by outsiders from published artefacts because the old version is dated, the new version is dated, and the rationale is in the diff. A reader can always go back and check whether the change moved the affiliate-overlap cohort differently from the non-overlapping cohort.

Every published methodology version is accompanied by an auto-generated differential-impact report comparing how brands in the affiliate-overlap set score versus the non-overlapping cohort, per vertical and country market. The overlap set is sourced from the editorial-firewall table — the same rows the firewall page renders. We do not maintain a separate manual list.

The generator runs as a read-only CLI (pnpm --filter @bdr/workers cli:generate-differential-impact-report). For each cell it reports mean and median scores per cohort, the delta, sample sizes, and a two-sided Mann–Whitney U test at α = 0.05. The standing claim is absence of systematic tilt: partners are not systematically higher and non-partners are not systematically lower. Committed artefacts live at docs/_archive/differential-impact/<methodology_version>.md (and matching JSON) in the repository; regenerate when methodology versions ship.

Setting up an independent oversight body (separate entity or named non-executive board) is deferred to a pre-announced trigger so the precondition is verifiable rather than discretionary. The trigger fires on the first of these events:

• first B2B revenue earned by IndexFair (any operator-facing paid product);
• first material public attack on IndexFair’s independence that is not resolved by published artefacts alone;
• first external capital raise.

Until a trigger fires, governance rests on the firewall, the no-retroactive rule, the differential-impact-report commitment, and the founder’s personal disclosure at /founders. When the trigger fires we publish the constitution of the oversight body before the next methodology release.

• Not a brand-by-brand affiliate-partner list. The firewall page enumerates brands with suppressed CTAs; this page governs the conflict at category level.
• Not a recusal claim. A solo founder cannot recuse from himself; the commitments above are designed so any bias is detectable by outsiders from published artefacts.
• Not the conflict register — that lives at /policies/conflicts (append-only, currently empty).

Founder disclosure: /founders. Firewall enforcement surface: /policies/firewall. Public conflict register: /policies/conflicts.