← Back to BaddEatsBADDEATS

Our method

How We Rank

A useful selection of places worth knowing, presented alphabetically rather than from first to last.

Current policy: v3.6 Revision 44 · Effective October 28, 2026

Specific dishes. Supported choices.

BaddEats researches the exact dish at the listed restaurant branch. We look for a real, independent positive judgment, verify that the place and offering are current, and consider unfavorable evidence with the same care as praise. Our guides express editorial judgment; they do not establish an objective winner or a numerical food score.

An ordinary guide needs at least 3 distinct supported restaurant concepts. We work toward 8–10 useful choices, with a maximum of 15 cards. We never add a weak candidate to reach a number. A guide with three to seven supported choices can publish after the required expansion and review; a completed result below three follows a separate nonpublication decision.

For a combination dish, each main food component needs positive quality support. An ordinary accompaniment can instead have verified composition and current availability when the recommendation makes no separate claim about its quality. We decide those roles before selecting places; an unresolved material problem with an accompaniment still blocks the affected claim.

Three ways to support a recommendation

01

An independent editorial recommendation

A credible food writer or qualified creator deliberately recommends the exact dish at the listed branch. One real recommendation can be enough. Authorized BaddEats human tasting has its own calibration and confirmation requirements.

02

A screened pattern of firsthand diner reports

At least three distinct authors report on the exact dish and branch across at least two dates, with at least two clear positive judgments. We inspect the whole relevant sample and recurring adverse patterns. A single platform can be used with its limitations recorded.

03

Direct dish evidence with relevant specialty support

A positive firsthand judgment of the exact dish is supported by a separately authored, independently controlled qualitative assessment of relevant specialty or craft. A shared-production component inference requires a documented current connection to the listed branch.

For diner consensus, the usable sample normally falls within the previous 36 months and includes at least one report within 18 months. An exception needs documented recipe or operator continuity and recent independent relevant specialty support. Dates, rejected reports, uncertainty and limitations stay in the research record. Star averages, review totals, queues and virality do not qualify a restaurant.

Two current criticisms prompt closer review rather than automatic rejection. Positive dish reports must outnumber unresolved comparable adverse reports, and we assess the criticism’s severity, specificity, independence and relevance. An unresolved material problem can block a recommendation even when other accounts are favorable.

Coverage gaps deserve more research

We search across the Maryland suburbs, Washington DC and Northern Virginia, including relevant languages, neighborhoods and service formats. We check credible omitted challengers and reconcile named candidates from useful outside guides. Geography helps guide discovery; it does not guarantee a place on a list.

When fewer than eight qualify, we review the shortfall and pursue useful additional evidence for any plausible contender whose evidence gap can be addressed. We check that the scope fits the requested dish, then work through applicable evaluator searches, screened diner evidence, component or specialty corroboration, and authorized human tasting feasibility. We document why further routes are exhausted, infeasible or uninformative. Reaching eight does not replace the final omitted-candidate checks or independent review.

Two independent evaluations may together describe a dish more fully. An independent judgment of a shared crust, sauce or production process can support specialty corroboration when we verify that the same component or process is used in the listed branch’s current dish and assess any differences in finishing or handling. We keep those limited inferences separate from direct observations and never count the same source twice. Shared ownership, equipment or a menu description alone supplies no quality judgment.

An optional coverage-lag designation adds safeguards for a mature, undercovered restaurant: enhanced screened consumer evidence, relevant structural support, Moderate confidence and a review within 180 days. Designated places can make up no more than half a guide. Every place must still qualify through one of the three evidence routes above.

Current relevance matters

We retain original evaluation dates and check whether the dish, branch, operator and preparation remain comparable. An older review is neither a fresh tasting nor automatically worthless. Material recipe changes, relaunches, closures and current adverse patterns trigger closer review. We set review dates according to production risk and the evidence actually used.

Offering verification can use a whole menu, PDF, verified merchant ordering page or dated menu photograph when its source, exact branch and current relevance are established. A restaurant does not need an item-specific web address. We record dining mode and handling when known, preserve unknown details honestly, and assess favorable and unfavorable food observations in their actual context.

Rapid execution, batch or artisanal production, and standardized production have respective review-planning benchmarks of 18, 30 and 36 months. These are planning aids, not automatic approval or expiry dates. The specific diner-sample date requirements still apply.

Clear claims and independent review

A recommendation does not mean BaddEats has personally tasted every dish. BaddEats tasting used for selection requires real, paid human observations under an authorized, calibrated and comparable protocol, with notes locked before synthesis. AI can help organize research; it cannot taste food or become an independent source.

For a narrowly defined coverage-repair addition below eight qualifying concepts, two separately locked observations—two reviewers or paid visits on different days—can support a Sufficient recommendation with Moderate confidence without an external quality judgment. That path applies when adding the place does not displace another qualifying choice or decide a material comparative boundary; full comparative tasting and corroboration requirements still apply to those decisions.

Before release, the frozen research artifact receives a separate review, independent editorial adjudication and the repairs those findings warrant. We reconcile the final selected records, public claims and source links. Restaurant participation, supplied photographs, advertising and commercial relationships never determine inclusion or placement.

This policy governs new research and reviewed updates. Existing guides retain their source dates and historical research records until a genuine refresh; changing the methodology does not claim a new visit or re-review of every restaurant.

Report Inaccuracy