When you build a report, each question is summarized using one or more aggregation types. The aggregation type controls what number (or text) appears in a report cell — a count, an average, a grouped box, a coded category, and so on. You choose aggregations in the Analysis step of the report editor, and the chips are multi-select, so a single question can appear under several aggregations side by side.
The aggregation types fall into a few families: numeric summaries, multi-select summaries, open-end (text) handling, discrete-choice (conjoint and MaxDiff) outputs, and time series. Not every aggregation is available for every question — the platform only offers the ones that make sense for the question type and its data.
Numeric aggregations
Frequency — the count of each value, shown as a count and a percentage of the chosen percent base. This is the default for non-numeric questions and for any question with recodes, and it answers "how many respondents chose each option."
Mean — the average of the numeric values in the cell. Useful for ratings, scales, and ranks when you want a single comparable number rather than a full distribution.
Mean without outliers — the mean calculated after extreme values have been removed, so a small number of unusually high or low responses do not distort the average. A value is treated as an outlier and dropped when it falls more than two standard deviations from the mean. Outlier removal only applies once a cell has at least 30 responses; below that threshold every value is kept, so the result matches the plain Mean. Use it on open-numeric questions (for example spend, age, or quantity) where a few outliers would otherwise skew the result.
Top Box — recodes the question into Top, Neutral, and Bottom groups and reports the share in each. The most positive value forms the top box and the most negative forms the bottom box, with everything in between treated as Neutral. For rating and numeric questions the highest value is the top box; for ranking questions the lowest number (first-ranked) is the top box.
Top 2 Box — the same Top / Neutral / Bottom recoding as Top Box, but grouping the two most positive values into the top box and the two most negative into the bottom box. Top 2 Box is a common summary for satisfaction and agreement scales because it captures the combined positive sentiment (for example "agree" plus "strongly agree").
Multi-select aggregations
Selected count mean — the average number of options a respondent selected on a multi-select question. Rather than reporting how often each option was chosen, it reports how many options the typical respondent picked overall.
Open-end (text) aggregations
Verbatim — the raw, unedited text responses exactly as respondents wrote them. Use this to read open-ends in respondents' own words.
Coded — open-end responses that have been classified into coded categories, so they can be counted and compared like a closed question. See open-end coding in MX8 Labs for how coding is set up.
Uncoded response — reports the raw underlying response value instead of its recoded category. Where a categorical, multi-select, calculated, or quota question would normally show recoded buckets, this aggregation shows the original answer as captured, which is useful for inspecting the underlying values behind a recode. It is available for categorical and multi-select question types rather than free-text open-ends.
Discrete-choice aggregations
These apply to conjoint and MaxDiff questions, where respondents make trade-offs between concepts.
Utility Scores — the modeled preference weights (utilities) for each attribute level, estimated from respondents' choices. Higher utilities indicate more preferred levels. See utility and simulated share methodology for how the model produces these.
Simulated Share — the simulated preference share for each concept, derived from the utility scores. It estimates the proportion of respondents who would choose each option in a modeled scenario, which is the basis for "what-if" share simulations.
Time series
Time Series — applies only to dial testing. Respondents move a dial while watching video, and this aggregation tracks each reading second by second across the clip, so you can see how sentiment rises and falls moment by moment rather than as a single pooled number. It is not available for other question types.
Choosing an aggregation
Because aggregation chips are multi-select, you can show the same question several ways in one report — for example Frequency and Top 2 Box together on a rating scale. Start from what the reader needs to take away: use Frequency for the full distribution, Mean or Mean without outliers for a single comparable figure, Top Box and Top 2 Box for headline positive sentiment, and the discrete-choice and time-series aggregations for the specialized question types they belong to. For the full report-building workflow, see reporting with MX8 Labs.