Insights in the MX8 Labs Research Platform are AI-generated or user-created answers to research questions. They are built from the survey's reporting data, so they are most useful when you ask a clear question and give enough context for the answer you need.
Unlike a general report summary, an insight responds to the question you ask. If you ask a broad question, you will get a broad answer. If you ask a specific question with a clear audience, decision, and format, the result will usually be more useful.
What insights use
Insights are based on the same reported survey data you see in MX8 Labs reports. That means:
- Results are weighted according to the survey's weighting setup.
- Statistical testing is respected.
- Insights use statistically significant results rather than treating every visible percentage difference as meaningful.
- Low-base cells and non-significant differences should not be overinterpreted.
This is important because a large-looking difference is not always a reliable finding. Insights are designed to help you focus on findings that are strong enough to support interpretation, not just the noisiest numbers in the table.
For more detail on how statistical significance is calculated and displayed in reports, see Interpreting Crosstabs and Understanding Stat Testing in MX8 Labs Reports.
When a survey is eligible for insights
A survey becomes eligible for survey-level insights when the survey is complete and reporting data has refreshed. In practice, check that:
- The survey has completed fieldwork.
- The survey has reportable responses.
- The dataset you are viewing has responses.
- The reporting cache has finished updating after fieldwork, edits, or data-prep changes.
If a survey is still in field, has no complete respondents, or has not finished refreshing its reports, insights may be missing, pending, or based only on setup data.
Dataset eligibility
Insights are generated against a selected dataset. Live insights require live, public audience data with responses. If you only have simulated data, you may be able to use it to check setup, question wording, report structure, and expected analysis shape, but simulated results should not be treated as final research findings.
When reviewing an insight, make sure the selected dataset matches the decision you are making. A simulated dataset can help you test whether the survey works. A live dataset is what you should use for real analysis.
Project insights
Project insights answer questions across surveys in the same project. They only use surveys that have been enabled for project insights.
Use project insights when you want to compare or combine related surveys, such as waves of a tracker, related audience segments, or multiple studies in the same research program. If a project insight is missing expected data, check that the relevant surveys are in the same project and are enabled for project insights.
How to ask better insight questions
Good insight questions include the decision you are trying to make, the audience you care about, and the format you want back. You can also tell the insight what to ignore.
Broad question:
What are the most important findings from this survey?
Better question:
What statistically significant differences in brand consideration should the marketing team act on, grouped by age and gender?
Format-specific question:
Write this as three executive bullets with one evidence point under each.
Strategy-focused question:
Identify the statistically significant results that should change our media plan, and ignore differences that are not statistically significant.
You can also include a detailed brief:
We are deciding whether to shift spend toward younger buyers. Focus on statistically significant differences in awareness, consideration, and purchase intent among 18-34s compared with older audiences. Use a direct executive tone. Return three recommendations, each with the supporting result and the business implication.
What detail to provide
Add detail when you need a specific kind of answer:
- Audience: who the answer is for, such as executives, researchers, marketers, or a client team.
- Business context: the campaign, brand, product, market, or decision behind the survey.
- Desired angle: what you want the insight to focus on, such as media planning, messaging, pricing, product fit, or audience segmentation.
- Output format: bullets, narrative summary, recommendation table, slide-ready headlines, or risks and opportunities.
- Tone: concise, executive, technical, client-ready, skeptical, or exploratory.
- Exclusions: ask it to ignore non-significant differences, low-base cells, simulated data, or findings that do not affect the decision.
The more specific the brief, the less time you will spend rewriting the answer.
Troubleshooting missing or weak insights
If insights are not available or do not look useful, check these common causes:
- No complete respondents: wait for completes or use reports to inspect setup and in-progress data.
- Survey still in field: final insights are strongest after fieldwork is complete.
- Wrong dataset selected: switch from simulated data to the live dataset when live responses are available.
- No statistically significant results: the survey may not have strong enough differences to support the question you asked.
- Project has no enabled surveys: enable at least one relevant survey for project insights before creating a project insight.
- Refresh still pending: wait for report and insight data to finish updating after fieldwork, edits, weighting changes, or data prep.
If the question is valid but the answer is too generic, rewrite the prompt with a narrower question, clearer business context, and a stricter output format.