Set a period and a denominator
Start with a precise interval, such as a full month, and count every review in it. Without the total, “two guests mentioned a wait” can sound large or small depending on what readers imagine. The star distribution helps orient you but cannot explain the reason for each rating by itself.
Synthetic example: in August a restaurant receives eight reviews: three five-star, two four-star, one three-star, one two-star and one one-star. The average is 29 divided by 8, about 3.6. This describes the month; it is not enough to conclude that service is improving or worsening.
Separate observation from interpretation
In the same example, two reviews mention waits on different occasions. The observation is that this theme appears in two of eight reviews. One possible interpretation is that certain service periods deserve a closer look. We still do not know whether the issue affects every shift, how common it is among guests who do not review, or what caused it.
Two other reviews praise the staff’s hospitality. That is a strength worth preserving, without turning it into a satisfaction percentage for all guests. A single comment on another topic is an episode to read, not an established pattern.
Choose a check you can actually make
For waiting times, the useful next step is not to announce a new process immediately. The manager can compare the days and times mentioned with staffing, dining room load and how waits were communicated. If there is a plausible cause, they can decide on a limited change and watch later periods.
When comparing months, look at review counts and context as well. A different average based on a small number of comments may reflect one experience. Comparison raises questions; by itself it does not prove that an action caused a change.
- Observation: how many reviews and which passages support the theme?
- Interpretation: what explanation is plausible, and what is missing to confirm it?
- Action: what check or small change can the manager decide on?
What Repliku reports show
Repliku reports bring together ratings, positive themes, recurring issues, supporting excerpts and suggested actions for a period. The initial report covers the previous calendar month when reviews are available for analysis; later monthly reports cover complete months. A period with no reviews does not produce an empty customer report.
Suggested actions are ideas to assess, not decisions made by the system. The manager can return to the reviews, check the context and decide whether to act. The Examples page contains a static reconstruction of the same eight-review synthetic scenario. It shows report fields rather than proving a real outcome.