Insights can summarize large amounts of hotel data, but not every question is best answered in one broad request. When an AI-generated response says that detailed results were sampled, truncated, or only partially inspected, narrowing the analysis helps you get a result you can confidently use.
A complete headline and an incomplete detail list can appear in the same answer. For example, Insights may calculate a complete portfolio Revenue total from an aggregated result while using only part of a much larger property-level result to explain the strongest and weakest contributors.
This article explains:
What incomplete detail means
How to identify the type of limitation
How to narrow and rerun the analysis
When to use a saved query or export instead
Check what is incomplete
First identify which part of the answer appears wrong or incomplete:
Headline total: a portfolio or property total such as Revenue, Units, Occupancy, or ADR.
Named row: one property, stay date, segment, inventory, or reservation.
Ranking or exception list: a claim about the top, bottom, best, worst, or every qualifying result.
Comparison: Budget, Forecast, last year, Rate Shopping, or Demand360 information.
Do not assume the entire answer is incomplete because one detailed section is limited. Equally, do not assume a complete total proves that every underlying row was included in the written explanation.
Recognize the three common causes
The detailed result is too large
The underlying analysis may stop after a maximum number of rows. If the requested combination of properties, stay dates, segments, and inventory types produces more rows than the result contains, the Agent cannot analyze rows it did not receive.
Typical signs include:
expected properties or dates are absent;
the response mentions a row limit or truncation;
an exception list ends at a round number of rows.
Reduce the scope or split the work into several smaller analyses.
Only part of the returned detail was inspected
The detailed result may be available, but the Agent may inspect only a subset while writing the response.
Typical signs include:
the answer says the result was sampled or partially inspected;
a ranking is described as indicative;
the response discusses only selected rows from a larger result.
Treat any top, bottom, contributor, or exception list as potentially incomplete when the response discloses sampling or partial inspection. A statement that the result is complete is helpful context, but it is not independent proof that every expected row was available.
Comparison data is missing
A query can be complete while a comparison is unavailable for part of the scope. Budget, Forecast, prior-year, mapping, Rate Shopping, or Demand360 coverage may differ by property and period.
Typical signs include:
blank comparison values;
zero Budget or Forecast for some properties;
missing mappings;
a comparison that covers only part of the portfolio.
Narrowing the request will not create missing source data. Check which properties and periods have the required comparison before relying on a portfolio variance.
Narrow and rerun the analysis
Use one or more of these approaches:
Reduce the number of properties—for example, run one region or brand at a time.
Shorten the stay-date or booking-date window.
Remove dimensions that are not required for the decision.
Request one comparison at a time.
Request a defined ranking with an explicit sort, and compare the result count with the population you expect.
Separate the complete portfolio total from the detailed contributor analysis.
State what Insights should do when required comparison data is missing.
Split a known property list into smaller groups when you need to verify that every property was covered.
For example:
Calculate portfolio Units, ADR, and Revenue first. Then return the five properties with the largest Revenue decline, sorted from largest to smallest decline. State the number of property rows used for the ranking and disclose any sampling, truncation, or partial inspection. If the number does not match the property population I expect, do not describe the ranking as complete; tell me how to split the request into smaller groups.
Use the right output when every row matters
Use a saved query, workbook, dashboard, or scheduled export when you need:
every reservation or stay date that meets a condition;
a complete mapping or data-quality work queue;
an editable detailed table;
a fixed visual layout;
a CSV or spreadsheet for further work.
Use an AI-generated response to interpret a focused result that you have verified—for example, explaining the main drivers within a complete top-five list. Keep the detailed table or export as the source of complete evidence, and use the written response to explain a clearly defined part of it.
Add a completeness guardrail
Add this instruction whenever an AI-generated response produces rankings, contributor analysis, or exception lists:
Before summarizing, state the number of detailed rows used and disclose any sampling, truncation, or partial inspection. Do not claim a complete ranking or exhaustive exception list when the expected population cannot be reconciled with the result. Suggest a narrower scope or smaller groups; if that is not possible, return only conclusions supported by the available result.
When to contact us
Select ? Help in the lower-left corner of the side navigation panel to contact our support team when:
the same defined total or row differs after you confirm the filters and date range;
a property or period with known Budget, Forecast, or external data remains missing;
a smaller, more focused request still produces an unexplained discrepancy;
the response does not disclose that it used partial detailed results.
Include the analysis name, property scope, date range, and the warning or unexpected figure you received.
Send us the details through ? Help whenever something still looks wrong. We’ll help you work out whether the issue is scope, coverage, or the underlying data.