How we research, review, and correct.

Sourcing

Every regulatory or numerical claim on RateStack is sourced from a primary authority — the FFIEC for area median income, the FHFA for conforming loan limits, GSE selling guides for program rules and LLPA grids, the CFPB for rules and enforcement actions, MISMO for data-standard specifics, the Mortgage Bankers Association for industry surveys. Where we paraphrase a source, we link the source inline. Where we summarize several, we link each.

Operational claims about the platform — "every quote ships with a per-rule trace," "the audit log is hash-chained" — are grounded in code that ships in the platform. If a claim describes a feature, that feature exists in production at the time of publication. When a feature changes, the page changes.

Author attribution

We default to a collective byline, "RateStack Team," because our writing is genuinely collaborative — engineers, product, and operations all sit in the same room. As individual contributors consent to be named, their bylines appear on /team.

We do not list affiliations or certifications that have not been confirmed in writing by the individual. Where credentials are surfaced (an NMLS license number, a public conference talk, a published paper), the page provides a verifiable anchor.

Dates

Long-form content carries two dates: published (when the article first went live) and reviewed (when it was last reread end-to-end for accuracy). The review date updates even when the words don't change — the act of rereading is the signal.

Evergreen content (guides, glossary, cookbook recipes, feature pages) is on a quarterly review cadence. Time-sensitive content (changelog entries, news posts) is reviewed when the underlying state changes.

AI tooling

We use AI tooling for research drafts and copy editing — never as the final author of published content. Every published page has been read, edited, and signed off by a human team member. We do not publish AI-generated lists of items we have not vetted, AI-summarized regulatory content we have not crosschecked against the primary source, or AI-written code samples we have not run.

We will continue to evolve this policy as the tooling does. The non-negotiable: a human reviewed it before publication.

Customer references

We do not publish customer logos without written permission. Customer quotes are paraphrased and anonymized by default; identifiable attributions appear only with explicit consent and the customer's opportunity to review the quote. We do not invent customer testimonials.

Competitive content

On comparison pages, we describe competitor products by their generally-known category posture and direct readers to verify specifics with the vendor. We do not assert that a competitor lacks a feature unless we have verified its absence. Where a competitor has a feature we lack, we say so.

We use competitor names, product names, and trademarks descriptively for comparison and interoperability discussion — nominative fair use. We do not imply endorsement.

Corrections

When we get something wrong, we fix it and log the correction at /corrections. The correction notes the date, the page affected, what was wrong, and what we changed. Significant corrections also trigger a notice on the changelog.

If you spot something incorrect, email editorial@ratestack.com. We respond within one business day and aim to resolve within five.

Independence

Editorial decisions are made by the team. We do not accept undisclosed promotional content, paid placements, or affiliate links. Where a piece of content references a partner, the relationship is disclosed inline.

Reader feedback

Constructive criticism is welcome and read. The fastest way to influence the editorial direction is to email a specific suggestion to editorial@ratestack.com or to /company/contact.