Built on the four families
Face, table, camera, and path observations match the surfaces current community videos document — and the official badges confirm two of the interactions (feed the Table Monster, bat the mimic).
Select only the clues you noticed. This observation calculator helps you slow down and recheck the scene; it never makes the serve-or-reject decision for you. Every prompt is grounded in the four documented anomaly families and their badge-confirmed responses.
Choose only the signs you personally saw in the current scene.
Select only what you actually noticed, or continue observing the scene before making your own in-game decision.
Face, table, camera, and path observations match the surfaces current community videos document — and the official badges confirm two of the interactions (feed the Table Monster, bat the mimic).
The checker reads only the boxes you select and cannot see your Roblox session. Nothing you enter leaves your browser.
Use visible prompts, the full scene, and your judgment before acting in the game. The result is a prompt to recheck, not a verdict.
The checker's prompts follow the same order every time. Learn the routine here and it works inside the game without the tool.
Where: The customer in front of you
Detect: Compare the whole customer against a normal guest — proportions, expression, posture — rather than one feature. A slightly odd single feature can be fine while a mismatch across the full model is the real tell.
Respond: Serve customers who read normal top to bottom; refuse the ones that do not.
Where: The table area
Detect: A named table entity that behaves unlike a regular guest occupies table space.
Respond: Feed it — do not serve it like a customer and do not bat it. Feeding it correctly pays the Yummy badge.
Where: The security-camera monitor feeds
Detect: A feed that looks disturbed, altered, or simply wrong signals an active camera anomaly. Cycle through every feed; one glance is not enough.
Respond: Scan all feeds habitually during service lulls instead of only reacting to obvious changes.
Where: Walkways between counter, tables, and kitchen
Detect: Anything obstructing a route you would normally walk.
Respond: Clear or handle the obstruction before it costs service time and reviews.
The most commonly documented cause of a bad review is acting on one detail instead of scanning the whole scene.
Community guides agree on the failure pattern: a player notices one odd feature — a slightly wrong face, one disturbed camera frame — and acts immediately. Sometimes that feature is fine on its own; the real anomaly was a mismatch somewhere else in the scene. The checker therefore asks about every surface before summarizing, and its summary always ends with a recheck prompt rather than an instruction.
The same logic explains what the tool will not do. It will not declare a customer safe, rank anomalies by danger (no verified power or spawn-rate data exists), or guess at hidden thresholds. The official description publishes exactly one failure metric — bad reviews — and no formula behind it, so a tool that claimed to compute safety would be inventing data.
Follow the multi-surface scan.
Track suspected mistakes cautiously.
Review checked dates and unknowns.
Compare reported observation surfaces.
No. It only summarizes the observations you enter.
No. It has no connection to your Roblox session.
Select only the face, body, table, path, or camera signs you actually noticed.
The result asks you to recheck multiple surfaces before making your own in-game decision.
Community compilations and the official badge list support four: the customer's appearance (face), the table area (Table Monster), the camera monitor feeds (the largest reported family with four chapters), and the walkways (path-blockers). Scanning all four in order is the routine this checker prompts.
The badge-confirmed interactions are: feed the Table Monster (Yummy badge), refuse customers whose whole appearance reads wrong, hit physical threats like the mimic with a bat (Time to Fight Back badge), and clear path obstructions before they cost service time. Save-a-customer moments are confirmed by the Savior badge.
Because no source — official or community — documents a complete rule set that would justify a guarantee. Single-clue certainty is the most commonly documented cause of bad reviews, so the tool is built to slow the decision down, not to automate it.
The logic itself does not read the game, but the guidance around it is reviewed after every named update. Anomaly behavior demonstrably changed between versions in mid-July 2026, which is why every supporting page carries a checked date.