Reporting with MX8 Labs

With MX8 Labs, you start to get data as soon as you have respondents; in fact, you can see simulated respondents as soon as you've set up the survey. Start by visiting the Reports tab:

The main elements you'll see here are:

• A drop-down containing the current reports. When you create a new survey, this will be populated by a set of AI-created reports, which will likely provide a good starting point for your analysis.
• A link to download the report in EXCEL format
• A link to edit the QUESTIONS included in the report
• A button to change ANALYSIS for the report
• Buttons to COPY reports from another survey, CREATE a new report, and run DATA PREP to modify the data.
• A drop-down to change the dataset. When you first create a survey, this will show the simulated data and switch to Live as soon as you have live data in the platform.

For AI-generated answers to specific research questions, see Insights in MX8 Labs. Insights use the survey's reporting data and statistically significant results.

The report editor

Click EDIT to open the report editor. It's a five-step wizard — Scope → Layout → Analysis → Recategorization → Statistical Significance — and you can move between steps using the stepper at the top or the BACK / NEXT buttons.

1. Scope

The Scope step controls what the report is about and which respondents it draws from.

• Name — the report name shown in the Reports drop-down.
• Questions — pick questions from the drop-down or type to search. Derived questions (configured from the Questions tab) appear here alongside regular questions; remove a question with the X on its chip.
• Report window start / end — for multi-wave trackers or longitudinal studies, restrict the report to a specific date range. Leave blank to include everything.
• Respondent Statuses — by default a report is built from complete respondents only. You can instead include all unterminated respondents — people who are still in progress or who reached a reportable point without being terminated — to read directional results while a survey is still in field, or to analyze partial responses. You can also build and view a report for a survey with no complete respondents yet, to check question wiring and report layout against simulated or in-progress data before the first completes arrive. Switching between statuses changes base sizes, so keep this in mind when comparing reports or reading statistical tests.

2. Layout

The Layout step controls how each question and tag is arranged in the cross-tab.

Each question in scope appears as a row in the Question Layout table; any tags on those questions appear in a separate Tag Layout table below.

• Role — set each question or tag to Row, Column, or Filter.
  • Rows are the primary reporting element.
  • Columns group and cross-tabulate. Base sizes are reported per column, so percentages are shown as a percentage of the column.
  • Filters add a drop-down at the top of the report that filters the entire report by that variable.
• Order — the up / down arrows next to each item move it within the table. This is the order rows and columns appear in the rendered report.

3. Analysis

The Analysis step controls how each question is summarized.

For each question in the report, you choose:

• Calculation — one or more of:
  • Frequency — the count of each value. The default for non-numeric questions and any question with recodes.
  • Mean — the cell's mean. Useful for ranks and ratings when you want a single comparable number.
  • Top Box — groups the top and bottom numeric values. For rating and numeric questions the highest values go in the top box; for ranking questions the lowest number (first-ranked) goes in the top box.
  • Top 2 Box — same idea, grouping the top two and bottom two.
• Percent Base — the denominator used to turn the cell count into a percentage. The default is Question Respondents (everyone who answered that question). Set it explicitly when you need a different base — for example, all respondents in the report rather than respondents who saw a routed question.

Calculation chips are multi-select, so you can show the same question multiple times in the same report under different aggregations — for example Frequency and Top 2 Box side by side on a single rating question.

Frequency, Mean, Top Box, and Top 2 Box are the most common aggregations, but the platform offers several more depending on the question type — including Mean without outliers, Selected count mean, Coded and Verbatim open-ends, Utility Scores, Simulated Share, and Time Series. For the full list and guidance on when to use each, see Question aggregation types.

4. Recategorization

The Recategorization step lets you recode question options on the fly for this report, without changing the underlying survey data. Available for non-numeric questions in the report.

Edit the Category Name for any value to merge or rename it; rows that share a category name are grouped in the rendered report.

5. Statistical Significance

The Statistical Significance step configures how the report is tested.

• Significance Testing — pick one of three tests:
  1. A residual t-test highlights results that are statistically higher or lower than expected based on a linear distribution of values between rows and columns.
  2. A t-test on each row marks, in each cell, the row labels that are statistically lower than this cell.
  3. A t-test on each column marks, in each cell, the column labels that are statistically lower than this column.
• Confidence Level — typically 0.95. How confident the platform must be before highlighting a result as statistically significant.
• Minimum respondents — typically 20, defaulting to the value set on the project screen. Any cells in the report with fewer respondents are grayed out.

Interpreting the data

The results are displayed in a standard cross-tabulation:

Keep in mind the following:

• Results are always weighted to represent the quotas you have defined in your survey. This is typically Gen Pop unless you've set up custom quotas.
• Percentages are always calculated as the percentage of the base size on the column.
• Any cells starting n= represent the base size for that column where there is a range in the base size that indicates that different base sizes apply to each row.
• Any bars highlighted in green are statistically significantly higher than other cells.
• Any bars highlighted in red are statistically significantly lower than other cells.
• The blue bars are neither statistically significantly higher nor lower, and should be considered the baseline.
• The white bars are those with fewer respondents than the minimum set in the analysis tab.
• If you want to know more about the errors on each bar, there are two things to look at:
• There is a smaller range bar in each cell that shows the range of values for that cell that fit within the statistical significance threshold.

When you move the mouse cursor over each cell, you will see a tooltip with more information:

• n= shows the number of respondents within each cell.
• Value shows the value of the cell.
• MoE shows the margin of error on this result based on the specified confidence level.
• The residual value, if a residual test is used.

Row and column T-Tests

Row and column t-tests are marked in the TAB using letters to show which row or column the results are statistically significantly different from:

You can see the letters for the relevant columns in the headers

Recoding and other data prep tasks

You can make extensive post-field updates to the survey data by clicking the DATA PREP button. This functionality is documented here:

Recoding and Post-Field Data Wrangling