Mean Series Configuration

Overview

mean-series-settings

Configure which data series (measures) are used for mean (average) calculations, statistical properties, and significance testing in mean tables.

Edition Availability

Mean tables: Available in Pro and Premium editions
All series configuration available in both Pro and Premium

Understanding Mean Tables

Mean tables display average values rather than percentages or counts. They're used for:

Customer satisfaction scores
Performance metrics (ratings, scores, times)
Financial averages (revenue per customer, average order value)
Measurement data (temperature, weight, distance)

tip

Mean tables are ideal for continuous data where averages and variability matter. Continuous (or numeric) data are indicated within Power Bi with a sigma (Σ) symbol next to the measure name (here under L2 and L5).

alt text

Example Mean Table:

Mean table showing average score of customer recognition by media channel and survey cell

warning

The component itself DOES NOT calculate means directly. You must provide appropriate measures from your data model for each series (Average, Count, Std Dev, etc.) or a pre-calculated measure for weighted mean. See the Data Model Requirements section for details on creating these measures.

Core Series Configuration

Mean Series

Setting: Mean Series
Required: Yes
Type: Dropdown (lists available measures)

The primary data series containing the values to be averaged.

What It Contains:

Individual measurements or scores
Pre-calculated means (if aggregated in data model)
Weighted average (if using weights)

Usually, after having dropped the variable in the 'values' section, you will use the 'Average' function in the dropdown listed measures:
powerbi-functions

You can also create custom DAX measures for more complex mean calculations. Examples:

Example 1: Satisfaction Scores

Mean Series = AVERAGE(Survey[SatisfactionScore])

Data:
- Respondent A: 5
- Respondent B: 4
- Respondent C: 5
Mean = (5 + 4 + 5) / 3 = 4.67

Example 2: Revenue per Customer

Mean Series = DIVIDE(SUM(Sales[Revenue]), DISTINCTCOUNT(Sales[CustomerID]))

Calculates: Average revenue per unique customer

Common Sources:

AVERAGE(Table[Metric])
DIVIDE(SUM(Values), COUNT(Records))
Pre-calculated mean field
Weighted mean calculation

Count Series

Setting: Count Series
Required: For displaying sample size
Type: Dropdown (lists available measures)

The number of observations used to calculate show the sample size.

What It Shows:

Sample size (n)
Number of respondents/records
Reliability indicator

Why It Matters:

Product A: Mean = 4.2, Count = 1000  ✅ Reliable 
Product B: Mean = 4.8, Count = 5     ❌ Unreliable

Same difference (0.6), but Product B has tiny sample

Usually, you will use the 'Count' or 'Count(Distinct)' function in the dropdown listed measures:
powerbi-functions

You may also create custom DAX measures for specific counting logic. Examples:

Example 1: Simple Count

Count Series = COUNTROWS(Survey)
Counts: All rows in the table

warning

using COUNTROWS will count all rows, including one which may NOT take part in the calculation. If your variable contains missing or irrelevant data, use the 'Count Rows' function in the dropdown listed functions.

Example 2: Respondent Count

Count Series = DISTINCTCOUNT(Survey[RespondentID])

Counts: Unique respondents (not duplicate responses)

Example 3: Non-Missing Values

Count Series = COUNTROWS(FILTER(Survey, NOT(ISBLANK(Survey[Score]))))

Counts: Only records with actual scores

Display in Table:

                Product A   Product B   Product C        
Base              1250         980         45
Average            4.2         4.5        3.9 
                ■ Large sample          ▪ Small sample (less reliable)

Standard Deviation Series

Setting: Standard Deviation Series
Required: For significance testing and variability analysis
Type: Dropdown (lists available measures)

The measure of variability/spread in the data. This measure is required for significance testing of mean differences. It is never displayed directly in the table but used in calculations behind the scenes.

Specific Significance Testing Series

Significance Mean Series

Setting: Significance Mean Series
Required: Only if using significance tests
Type: Dropdown (lists available measures)
Available in: Pro, Premium

There is cases where you may need to provide a different mean series for significance testing than the one used for display, reason why the tool proposes this separate setting. Set here the mean values used for statistical significance testing.

When It Differs from Mean Series:

Display uses weighted means
Testing uses unweighted means
Different aggregation rules for testing

Common Scenarios:

Scenario 1: Same as Mean Series (Most Common)

Mean Series: AVERAGE(Survey[Score])
Significance Mean Series: AVERAGE(Survey[Score])

Use when: Testing and display use same data


Mean table showing same mean series for display and significance testing

Scenario 2: Different Series

tip

To Create a weighted mean, use "New Measure" in Power BI and entrer a DAX formula like:

L2_weighted = DIVIDE(
    SUMX(main, main[L2] * main[Weight]),   // Note the SUMX to multiply row by row each L2 value by its Weight
    SUM(main[Weight]),
    BLANK() // if divide by Zero
)

You will also need to create a weighted standard deviation measure for significance testing.

L2_WStdDev = 
VAR MeanValue = [L2_weighted]
RETURN  
    SQRT(
        DIVIDE(
            SUMX(
                main,
                main[Weight] * POWER(main[L2] - MeanValue, 2)
            ),
            SUM(main[Weight])
        )
    )

Example: Here under are three different settings

in the first one the 'Mean Series' is set to a raw mean measure with the 'Significance Mean Series' set the same as count Series.
in the second one the 'Mean Series' is still set to a raw mean measure with the 'Significance Mean Series' and 'Count Series' set to the Weighted measures. With these settings, significance use the weighted population and the weighted mean for testing while the display use the raw mean so values remains the same while significance tests are based differently.
in the third one the 'Mean Series' is set to a weighted mean measure with the 'Significance Mean Series' set to a raw mean measure and the 'Count Series' set to the raw count measure. With these settings, significance use the raw population and the raw mean for testing while the display use the weighted mean so values differs between display and testing.


Mean table showing the same mean series for display (unweighted) and significance testing (unweighted)


Mean table showing different mean series for display (weighted) and significance testing (unweighted) Note the difference in the 'Count series' and 'Mean Series for significance tests'


Mean table showing different mean series for display (unweighted) and significance testing (weighted) Note the difference in the 'Count series' and 'Mean Series for significance tests'

Why Separate?:

Research protocols may require unweighted testing
Different statistical requirements
Regulatory or methodological standards

Unweighted Base Series

Setting: Unweighted Base Series
Required: Only when using weighted data
Type: Dropdown (lists available measures)

The raw count of observations before weighting adjustments. Used for unweighted base row only, no calculation is made with it.

When to Use:

For research transparency
To show actual sample sizes

Data Model Requirements

Required Measures for Mean Tables

// Simple average
MeanScore = AVERAGE(Survey[Score])

// Or weighted average
WeightedMean = 
VAR FilteredData = FILTER(Survey, NOT(ISBLANK(Survey[Score])))  // Define the Sub-Set of Data where 'Score' is present (not necessary if variable is mandatory)
    DIVIDE(
        SUMX(FilteredData, Survey[Score] * Survey[Weight]),
        SUMX(FilteredData, Survey[Weight]),
        BLANK()
    )

// Simple count
SampleSize = COUNTROWS(Survey)

// Or unique respondents
RespondentCount = DISTINCTCOUNT(Survey[RespondentID])

// Or non-blank count
ValidResponses = COUNTROWS(FILTER(Survey, NOT(ISBLANK(Survey[Score]))))

// or Weighted count
WeightedCount = SUM(Survey[Weight])

warning

You SHOULD NOT Count another variable (like the record ID or respondent ID) as the Average or Std Deviation are calculated upon non-null values of the selected mean variable.

// Sample standard deviation
StdDeviation = STDEV.S(Survey[Score])

// Or Weighted standard deviation
WeightedStdDev =
VAR FilteredData = FILTER(Survey, NOT(ISBLANK(Survey[Score])))  // Define the Sub-Set of Data where 'Score' is present

VAR MeanValue = DIVIDE(SUMX(FilteredData, Survey[Score] * Survey[Weight]), SUMX(FilteredData, Survey[Weight]))
RETURN  
    SQRT(
        DIVIDE(
            SUMX(FilteredData,
                Survey[Weight] * POWER(Survey[Score] - MeanValue, 2)
            ),
            SUMX(FilteredData, Survey[Weight])
        )
    )

4. Unweighted Base : This shoud ALWAYS be a raw 'count' or 'Count distinct' function of non-null value of selected mean variable

UnweightedCount = DISTINCTCOUNT(Survey[Score])

warning

You SHOULD NOT Count another variable (like the record ID or respondent ID) as the Mean or Std Deviation are calculated upon non-null values of the selected mean variable.

Practical Examples

Example 1: Customer Satisfaction Scores

Question: "Rate your satisfaction (1-5)"

Configuration:
- Mean Series: "Average Satisfaction Score"
- Count Series: "Count of Respondents"
- StdDev Series: Not shown

Example 2: Performance Analysis with significance

Metric: Employee performance ratings

Configuration:
- Mean Series: "Average Rating"
- Count Series: "Employee Count"
- StdDev Series: "Rating Std Dev"

Example 3: Weighted Research Study

Survey: National satisfaction study (weighted)

Configuration:
- Mean Series: "Weighted Mean Satisfaction"
- Count Series: "Sum of Weights"
- StdDev Series: "Satisfaction Std Dev"
- Unweighted Base: "Count of Respondents"
- Significance Mean: "Unweighted Mean"

Example 4: Revenue Analysis

Metric: Average revenue per customer

Configuration:
- Mean Series: "Revenue per Customer"
- Count Series: "Customer Count"

Statistical Context

When to Use Mean Tables

Good for:

Continuous measurements (temperature, time, money)
Rating scales (1-5, 1-10)
Satisfaction scores (NPS, CSAT)
Performance metrics
Financial averages

Not Good for:

Categorical data (Yes/No, Male/Female)
Count data (number of purchases)
Percentages/proportions
Binary outcomes

Use percentage tables instead for: Frequency distributions, category breakdowns

Sample Size Considerations

Minimum Sample Sizes:

n < 30: Means unreliable, use caution
n = 30-100: Acceptable for most purposes
n > 100: Reliable means
n > 1000: Very stable estimates

Impact on Significance:

Large samples detect small differences
Small samples miss real differences
Always consider practical significance

Troubleshooting

Q: Mean values look incorrect

Verify Mean Series calculation in a simple visual first
Check for filter context issues in DAX
Ensure measure aggregates correctly across dimensions

Q: Count doesn't match expectations

Check if counting rows vs distinct IDs
Verify weighted vs unweighted count
Look for filter context in measure definition

Q: Standard deviation shows as blank

Ensure StdDev Series is mapped
Check that measure returns values
Verify sufficient data points (need n ≥ 2)

Q: Unweighted base same as count

A: Your data likely isn't weighted; you can hide unweighted row

Q: Mean tables not available

A: Mean tables require Pro or Premium edition

Q: Significance tests don't work with means

Verify all three series: Mean, Count, StdDev
Check Significance Mean Series is configured
Ensure adequate sample sizes (n ≥ 30 recommended)

Best Practices

Always Show Count: Sample size is critical for interpreting means
Consider StdDev: Helps understand data consistency
Test Measures First: Validate in simple visuals before using in cross-tab
Document Calculations: Note whether means are weighted/unweighted
Check for Outliers: Extreme values can skew means
Minimum Samples: Use thresholds to mask small sample means
Weighted Data: Always show both weighted and unweighted counts
Practical Significance: A statistically significant 0.1 difference may not matter

Table Contents — Set table type to "Mean"
Percentage Series — For percentage table configuration
Significance Testing — Configure t-tests for means
Thresholds — Mask means with small samples
Totals & Subtotals — How counts display

For more help, see the Quick Start Guide or contact support.

Overview​

Understanding Mean Tables​

Core Series Configuration​

Mean Series​

Count Series​

Standard Deviation Series​

Specific Significance Testing Series​

Significance Mean Series​

Unweighted Base Series​

Data Model Requirements​

Required Measures for Mean Tables​

1. Mean: Either select 'Average' in the dropdown selection of the continuous variable (raw only) or create a Mean Calculation​

2. Count: Either select 'Count' in the dropdown selection (raw), the 'Sum' function of the weight variable or a Size Calculation​

3. Standard Deviation: Either select 'Standard Deviation' in the dropdown selection of the variable or create a Standard Deviation Calculation​

4. Unweighted Base : This shoud ALWAYS be a raw 'count' or 'Count distinct' function of non-null value of selected mean variable​

Practical Examples​

Example 1: Customer Satisfaction Scores​

Example 2: Performance Analysis with significance​

Example 3: Weighted Research Study​

Example 4: Revenue Analysis​

Statistical Context​

When to Use Mean Tables​

Sample Size Considerations​

Troubleshooting​

Q: Mean values look incorrect​

Q: Count doesn't match expectations​

Q: Standard deviation shows as blank​

Q: Unweighted base same as count​

Q: Mean tables not available​

Q: Significance tests don't work with means​

Best Practices​

Related Settings​

Overview

Understanding Mean Tables

Core Series Configuration

Mean Series

Count Series

Standard Deviation Series

Specific Significance Testing Series

Significance Mean Series

Unweighted Base Series

Data Model Requirements

Required Measures for Mean Tables

1. Mean: Either select 'Average' in the dropdown selection of the continuous variable (raw only) or create a Mean Calculation

2. Count: Either select 'Count' in the dropdown selection (raw), the 'Sum' function of the weight variable or a Size Calculation

3. Standard Deviation: Either select 'Standard Deviation' in the dropdown selection of the variable or create a Standard Deviation Calculation

4. Unweighted Base : This shoud ALWAYS be a raw 'count' or 'Count distinct' function of non-null value of selected mean variable

Practical Examples

Example 1: Customer Satisfaction Scores

Example 2: Performance Analysis with significance

Example 3: Weighted Research Study

Example 4: Revenue Analysis

Statistical Context

When to Use Mean Tables

Sample Size Considerations

Troubleshooting

Q: Mean values look incorrect

Q: Count doesn't match expectations

Q: Standard deviation shows as blank

Q: Unweighted base same as count

Q: Mean tables not available

Q: Significance tests don't work with means

Best Practices

Related Settings