Frequently asked questions

Use Brand portal on the home page or Brand company portal in the workspace. Your invited account and organization membership determine access. Admins, executives and regional users can review their portfolio and screen conversion opportunities; admins manage company playbooks and directory listings. Hilton and Marriott sandbox modes use fictional hotel candidates and illustrative criteria, with no access to those companies’ actual data.

No. Fit is an explainable screening score across key count, room footprint, rate positioning, amenities, condition and renovation budget. Evidence coverage measures supplied inputs, not accuracy or probability of approval. Sandbox numerical criteria are invented. Confirm official standards, franchise terms, territory availability, owner interest and an on-site property improvement plan before making a decision.

Hotelium helps hotel teams collect inspection evidence, review AI observations, track physical assets and corrective work, compare purchasing options, and prepare capital budgets. Original evidence and human review decisions stay connected to the work.

Choose Try it out — Demo sandbox on the home page. You will enter your own editable sandbox with five fictional properties and synthetic findings. Use dummy data only. Sandbox access lasts 24 hours; expiration does not mean immediate deletion of stored data.

Open App Flow in the workspace sidebar for an interactive, step-by-step walkthrough from property setup and capture to review, corrective work and reporting. The whitepaper provides an overview you can download and share.

Use a sample button to populate a form with example values. Review and edit those values before submitting. Filling a sample does not save a record or approve a decision; the form’s submit action is still required.

Set up a property and its spaces, then open Inspections and choose Create inspection. Select the applicable scope and capture requirements. Available creation and review actions depend on your assigned role and property access.

Each tracker labels its own measure, such as accepted required shots, completed inspections or reviewed findings. Capture progress is not the same as a condition score or compliance result. Review the tracker’s insight bubble for its calculation and scope.

Follow the assignment’s retake instructions when evidence is unusable. Record restrictions and missing evidence honestly. Do not enter a restricted or unsafe area to increase completion. Use Not Applicable only when the shot genuinely does not apply and record the rationale.

Authorized teams can create versioned standards, define rules and import structured CSV rules from sources they are permitted to use. A rule’s verification method distinguishes visual evidence from document, manual, equipment-test and service checks. A photo cannot verify every requirement.

No. Model observations and findings are candidates for human review. Reviewers compare them with the original evidence and applicable requirements, then approve, reject or leave them unresolved. AI cannot establish hidden conditions or replace a qualified inspection.

It is the confidence recorded by the model for an individual observation or finding. It is not calibrated proof of accuracy, a hotel quality score or a human approval. Missing or invalid values show Not provided. DEMO confidence comes from synthetic examples, not measured predictions.

AI Intelligence cards apply explained workflow rules to your loaded records. Pending findings with missing confidence or confidence below 70% are prioritized for review. That cutoff is a triage heuristic, not a validated accuracy threshold. Open Why this suggestion? to see the reason; suggestions never change records automatically.

Scoring depends on the recorded inspection, accepted evidence, reviewed observations and applicable scoring rules. Missing coverage or insufficient evidence can qualify or suppress a score. Open the property’s scoring details to see its reasons. A missing score is not zero, and scores should be compared only with their dates and scoring versions in view.

No. The demo uses synthetic analysis and the landing page animation runs in your browser. Real analysis in a live deployment requires a separately configured provider. The app does not silently substitute demo results when live analysis fails.

An assigned action can collect resolution evidence and move to Awaiting Verification. An authorized reviewer must verify the correction before it is treated as verified. Uploading evidence or reaching a target date alone does not close the issue.

Treat a visual identity as a candidate until it is verified against labels or documentation. Check the catalog’s verification status, product suitability and any required brand approval before using a model or manufacturer name for purchasing.

No. Procurement comparisons depend on entered quotes and costs, including freight, installation and switching assumptions. Capital budgets depend on sourced quantities, unit costs and contingency. Review sources and approvals before acting; potential savings are not realized savings. The landing-page calculator is an editable estimate of capacity value.

Each visitor sandbox has its own organization and access session. Your changes stay within that sandbox. Refreshing the same browser resumes it while access is valid; Exit sandbox revokes that sandbox session. Demo properties and findings are fictional.

No. The Intelligence page lists separate data-purpose permissions, including internal processing, reporting, benchmarking, training and raw-image licensing. Uploading evidence is not blanket permission for these uses. An authorized administrator manages these settings.

Your organization, assigned role and property access determine the records and actions available to you. Check that you have selected the intended organization, then ask your organization administrator to review access. Sandbox users cannot invite users or join real organizations.

An administrator can inspect recent Processing jobs in Administration and review the recorded error. Check provider configuration, processing permission and evidence readiness before retrying. A failed job is not a successful analysis and does not establish that the image contains no issues.