Key points
- Involve front desk, housekeeping supervision and frontline roles.
- Separate a verbal answer from a result actually demonstrated.
- Keep critical blockers outside the average score.
Make the demonstrations comparable
Send the script before the session so each supplier can prepare a suitable environment. Use the same fictional rooms, roles and starting conditions. Ask which version, modules and configuration are being demonstrated. If a capability belongs to another offer, identify that difference in the record before evaluating it.
Choose a facilitator to introduce events and an observer to record facts. Department representatives assess the consequences for their work. Do not spend the session simply counting clicks. Focus on missing information, ambiguous decisions and additional interventions required to complete the workflow. These observations are more useful than a general impression that the interface looks polished.
Run four defined events
| Fictional scenario | Requested action | Decisive observation |
|---|---|---|
| Late departure | Change departure for a room already in the work plan. | Who sees the change, and what happens to the planned task? |
| Blocked room | Report a defect that prevents release. | Can front desk distinguish cleaned from available for release? |
| Reassignment | Transfer a task during the shift. | Who owns it now, and what does the previous assignee see? |
| Operational incident | Create a report and request further information. | Do the original context, response and next action remain connected? |
Inspect each result from the role that needs to use it. A change on an administrator screen does not prove that a room attendant or receptionist can understand it in their own workflow. If the PMS exchange is simulated, explicitly label that evidence as simulated.
Score evidence rather than a promise
The proposed scale runs from 0 to 3: 0 means the need is unsupported; 1 means it is described verbally; 2 means the result is demonstrated with a substantial workaround; 3 means the agreed workflow is executed and understood by the relevant roles. This is an evaluation convention, not a scientific measure of software quality.
Add a separate “not tested” category. A supplier may have a capability that cannot be shown in the session, but it remains unverified. For every score, record the observed step, its limitation and the evidence still required. A screenshot can document a screen; it cannot, by itself, demonstrate an entire working process.
Keep disqualifying conditions visible
Define in advance the situations that would prevent operation: overly broad access, missing critical information or a task that appears to belong to two people. Strong scores on less important workflows should not cancel these blockers. CNIL recommends access appropriate to each role, so include a restricted role in the demonstration rather than testing only administrator access.
In a fictional example, the four scenarios score 3, 2, 3 and 1: 9 points out of 12, or 75%. If the incident workflow fails an essential condition, that descriptive percentage does not make the system acceptable. The decision remains conditional until the supplier demonstrates the necessary correction with the relevant roles.
Introduce one controlled variation
After the prepared workflow, change a single condition: the new assignee becomes unavailable, departure changes again or the reported defect remains unresolved. This variation tests recovery without demanding an improvised demonstration of the entire product. In particular, check that outdated information does not silently become the current reference after the second change.
A usability difficulty deserves a question. Is an explanation missing, is configuration incomplete, or is the capability unavailable? Do not immediately classify every difficulty as a product defect. Ask the supplier to identify the cause and repeat the workflow after correction. Retain both observations so the comparison remains accurate and fair.
Close with a record that supports the next decision
Produce a record of the evaluated offer, completed scenarios, scores, blockers, evidence and open questions. Have department representatives review it, then invite the supplier to correct factual errors. Keep unresolved disagreements visible instead of replacing the original observation with an unsupported assurance.
The session is complete when each question has either an outcome or a named next step. Important capabilities that were not demonstrated move into the acceptance test plan. Also record the work required from the hotel, including data preparation, configuration and training. A convincing demonstration does not remove those practical deployment conditions.
To place the evaluation within the first improvement project, see where to start hotel digitalisation.
Frequently asked questions
Can suppliers be compared without weighted scores?
Yes, when the scenarios have similar importance. If some workflows matter more, agree their weights before the demonstrations. Always keep disqualifying conditions separate: even a weighted average can hide a critical operational failure.
Should the whole demonstration be recorded?
A recording may help when participants and applicable rules permit it. You still need structured notes to find the scenario, configuration and outcome quickly. Use fictional data and keep real guest information out of the demonstration environment.
