How this desk decides what passes

What a judging point is

A judging point is one observation that can be decided on its own. It has to satisfy four conditions before it is published here. It has to be identifiable, which means a reader can say in one sentence which part or which form is being looked at. It has to have an evidence requirement, which means the page states which photographs or which pieces of information are needed to decide it.

It has to carry three thresholds — pass, query and return — expressed as observable features rather than impressions. And it has to carry a stop condition: the finding that should end the assessment rather than send you to the next step.

Anything that cannot meet all four is not published as a judging point. It may still appear as a question page, where the standard is explanation rather than decision.

Why the object is a rule and not a product

The unit of publication here is a criterion, not an item. An item page rots: the listing changes, the link dies, the seller relists under a new title. A criterion about how midsole alignment reads in a side-on photograph does not rot in the same way. It changes only when the way the part is made changes.

That choice is what makes a dated commitment possible. This desk can promise a check date on a criterion because a criterion has a stable identity. It could not make the same promise about a product row.

Evidence levels

E1 is a seller-supplied photograph or video taken before dispatch, at the angles a step requires. It is the only level that is useful both for a decision and for a claim, because it exists before money changes hands and can be attached to a message thread.

E2 is a photograph you take after the item arrives. It is useful for a claim and useless for prevention. E3 is platform public information: the listing, the storefront, the public reviews. It is enough to shortlist and never enough to judge workmanship, because the seller chooses every part of it. E4 is a third-party reference — a community post, a video, a second-hand account. It is a lead and not a criterion, and it is the highest level any roster record can reach.

The three verdicts and their boundaries

Pass means the observation sits inside normal manufacturing variation for that part, and no other step has flagged a query. Query means the observation could be normal or could not, and that the answer depends on the batch or on how the photograph was taken. A query never carries a verdict by itself, and it never turns into a pass because other points looked fine.

Return means the observation falls inside the fatal set for that part, or cannot be told apart from it. The distinction between query and return is drawn so that two people reading the same photograph arrive at the same answer, which is why the criteria are written as counts, distances and named features rather than as quality adjectives.

Reading order and stop conditions

The sequence runs from what cannot be undone to what can be negotiated. Deal credibility comes first, because money handed to the wrong object is not recoverable. Structural defects come second, because a lasting error cannot be repaired by conversation. Surface and finish come third, because most variation there is negotiable. Handover and completeness come last, because those failures are usually settled with a partial refund rather than a return.

Each step may carry a stop condition. When one is met, the assessment ends there: the remaining steps are not read, and the page links to the route that applies instead. This is deliberate. Continuing to read a sequence after an irreversible finding produces false precision.

Review windows

Parameters that depend on a rule that changes — duty bands, freight, carrier policy, platform thresholds — carry the shortest window at thirty days. Criteria that depend on how a part is made carry ninety days. Terminology, carrier-shape explanations and process descriptions carry one hundred and eighty.

When a page passes its window, it says so. On a judging point, the stop logic is disabled and the page refuses to issue a verdict until it has been re-checked. On a gauge, the figures are shown as placeholders rather than withheld entirely, with the stale label attached.

Adding a record to the roster

A record describes a direction rather than a shop. It states what that direction is known for, the type of source it rests on, how many observations were read to produce it, whether any dispute type has been publicly reported, and when it was last checked.

The evidence level for a record is never higher than E3, and usually E4. This desk does not buy from these directions, does not inspect their output and does not hold transaction data. A record is a summary of what is publicly observable, with the observation count attached so a reader can weigh it.

Withdrawing a record

A record is withdrawn when its source can no longer be verified, when the direction it describes stops being identifiable, when a dispute type that has been publicly reported changes the picture materially, or when a review window passes twice without a check being possible.

Withdrawal does not delete the record. It moves the record to retired, the page states why, and the change is logged with a date. A reader who saw the earlier version should be able to find out what changed.

How corrections are handled

Any reader can report a criterion that looks wrong, a record that looks out of date, or a link that no longer resolves. Reports are read in the order they arrive. A report that changes a page produces two things: the change itself, and an entry in the change archive naming the date, the page and the reason.

Reports that do not change a page are still answered with the reasoning, because a criterion that cannot be explained to a reader who disagrees with it is usually a criterion that was written badly.

Judging points, questions and gaps

Not everything belongs on a judging point. A question that has a stable answer but no threshold — why a link stops resolving, how a claim window works — is published as a question page instead. The distinction is whether the reader can be handed a decision rule, or whether they are being given context for a decision they will make somewhere else.

The reverse case matters too. A topic that sounds like a question but hides a threshold belongs on a judging point, because a question page has no room for three criteria and a stop condition side by side. Where both forms would work, the threshold wins.

What is left over is gaps. A failure mode that has been reported but has no stable criteria yet does not get a page with a placeholder in it. It is recorded as a report through the correction channel and stays off the site until a criterion can be written that two people would apply the same way.

Why the criteria are counted rather than described

Every threshold here is written as a count, a distance, a named feature or a condition that can be checked in one look. Nothing is written as an adjective about quality. That constraint is what makes a verdict reproducible, and reproducibility is the point: a criterion that two readers apply differently is not a criterion, it is an opinion with a heading.

The cost of the constraint is that some real differences cannot be written down this way. How a zip feels through its travel, how a garment smells out of the bag, whether hardware has the right heft in the hand — none of those reduce to a count, so none of them appear as thresholds. They are named on the pages as things the evidence cannot reach, rather than smuggled in as subjective criteria.

What this desk cannot do

There is no laboratory here, no physical inspection, no test purchasing and no access to transaction data. Nothing on this site can establish authenticity, and no page issues an opinion on it.

The gauges model costs from recorded parameters; they do not quote a price. The roster summarises public observation; it does not vouch for anyone. Where a question needs a fact this desk does not have, the page says so rather than filling the gap.