19 Measuring Diversity and Inclusion

19.1 Why Measurement Matters

A diversity number that the audience cannot interpret is worse than no number at all.

The previous chapter framed the three faces of EDI analytics. This chapter goes one level deeper into how diversity and inclusion are actually measured. Measurement is the place where most EDI programmes either earn the credibility to act or lose it before they begin. A diversity number that has not been computed carefully will be challenged in the meeting where it most matters; an inclusion measure built without segmentation, response-rate discipline, and longitudinal context will be read as marketing rather than as evidence.

The challenge is not lack of techniques. As Lisa H. Nishii et al. (2018) document in their multi-level review of diversity-practice effectiveness, the literature has accumulated a substantial toolkit for measuring composition, dispersion, equity gaps, and climate. The challenge is choosing the right measure for the question, computing it transparently, and rendering it on a page that the audience can interpret without needing a methodology footnote. As Quinetta M. Roberson et al. (2017) set out in the centennial review of diversity at work, the field’s main barrier to progress is no longer measurement availability. It is the discipline of using the available measures honestly, with their boundary conditions intact and their uncertainty rendered visibly.

The visualisation lens determines whether measurement reaches its audience. Diversity charts that show only composition flatten a multi-dimensional concept into a single bar; inclusion charts that show only an aggregate score erase the segmentation that gives the score its meaning. The dashboard that surfaces both with the discipline this chapter sets out is the dashboard the executive committee comes back to without prompting, because the page rewards careful reading rather than rewarding only a confident headline.

The diversity-and-inclusion-measurement contract

Every diversity measure surfaced on the dashboard names its index — representation, dispersion, similarity, or equity — so that the audience reads the chart at the level the index supports.
Inclusion measures are reported with response rate, segmentation, and longitudinal trend on the same page; a single aggregate score without those companions does not earn the dashboard.
Measurement is multi-level by default. The same metric is rendered for the individual, the team, and the organisation, because the same number can mean very different things at each level.

19.2 Diversity Indices

Diversity is not one measurement. It is a family of indices, each of which captures a different aspect of workforce composition. A scorecard that picks one index and ignores the others is a scorecard that hides at least as much as it shows. The four indices that recur across mature programmes are representation, dispersion, similarity, and equity.

Four Families of Diversity Indices

Index family	What it measures	Example	Visualisation
Representation	Share of the workforce by attribute	Percentage by gender, ethnicity, age band	Stacked bar by level, treemap
Dispersion	How spread the workforce is across attribute values	Blau index, Shannon entropy of role-family mix	Distribution chart with index value
Similarity	How similar a team is to a reference (firm, market, customer)	Index of dissimilarity from market base rate	Pair-test chart with reference line
Equity	Whether outcomes are equal across attributes	Pay-equity gap, promotion ratio, exit ratio	Distribution chart with target gap

Choosing the index for the question

The right index depends on the question. Are we hiring representatively from the labour market is a similarity question, answered by an index of dissimilarity. Is our leadership pipeline as diverse as our entry level is a representation question, answered by stacked bars across levels. Are some teams much more or less diverse than others is a dispersion question, answered by an entropy or Blau index. Are the people we hired equally likely to be promoted, paid, and retained is an equity question. Picking one index and using it for all four questions is the most common mistake in diversity measurement, and the dashboard inherits whatever confusion follows.

19.3 Inclusion Measurement

Inclusion is the experiential climate that determines whether a diverse workforce can contribute. It cannot be inferred from composition; it has to be measured directly, and the measurement requires more discipline than most other HR climate work because the audience that consumes the result is also the audience whose answers produced it.

The Three Channels of Inclusion Measurement

Channel	What it captures	Visualisation
Survey-based climate	Belonging, voice, fairness, manager support	Distribution gauge with response-rate disclosure
Behavioural signals	Promotion, mobility, voluntary exit by segment	Funnel and survival chart by segment
Listening signals	Open-text comments, listening sessions, employee networks	Theme heat map with longitudinal trend

What disciplined inclusion measurement looks like

A disciplined inclusion measurement programme has four properties. First, it is segmented: the aggregate score is paired with break-outs by demographic, tenure, manager, and role family. Second, it is response-rate-aware: the score is rendered alongside the response rate, and the audience is told when the rate is low enough that the result should be read with caution. Third, it is longitudinal: the score is shown across cycles, and direction matters as much as level. Fourth, it is multi-channel: the survey number is paired with behavioural and listening signals so that the audience can triangulate. As Lisa H. Nishii et al. (2018) argue, inclusion measurements that fail any one of these four properties produce results the function eventually has to defend without the data needed to defend them.

19.4 The Multi-Level Frame

Diversity and inclusion measurements have to be read at three levels simultaneously: the individual, the team, and the organisation. A score that looks healthy at one level can be hiding a serious issue at another, and a dashboard that operates at one level only is incomplete by design.

The Same Measure at Three Levels

flowchart LR
  A[Individual Level<br/>experience and outcomes] --> B[Team Level<br/>climate and dynamics]
  B --> C[Organisation Level<br/>aggregate composition and equity]
  C --> A
  style A fill:#E8F0FE,stroke:#1A73E8
  style B fill:#FEF7E0,stroke:#F9AB00
  style C fill:#E6F4EA,stroke:#137333

Reading the loop in either direction reveals different things. From individual to team to organisation, the question is whether the experience of the few aggregates into the climate of the many. From organisation back to individual, the question is whether the firm’s stated commitments translate into the lived experience of the people who work there. The dashboard that supports both reads is the dashboard that surfaces every level explicitly and does not let an aggregate hide a local problem.

Cell-size discipline at the individual level

The individual-level measurement always raises the cell-size question. A score for a department of three employees is not a department-level measure, it is a near-disclosure of the three individuals. A disciplined dashboard sets a minimum cell size, suppresses the cell when it falls below the threshold, declares the suppression openly, and offers an alternative roll-up the audience can use. As Quinetta M. Roberson et al. (2017) note, the firms that build cell-size discipline into their diversity dashboards earn the trust of the workforce whose data appears on those dashboards, and that trust is what makes longitudinal comparisons possible at all.

19.5 Visualising Diversity and Inclusion Together

The most credible diversity-and-inclusion dashboard is the one that surfaces composition, equity, and climate on the same page, with the disciplines this chapter has set out rendered visibly. Five design choices, applied consistently, turn the multi-index, multi-channel, multi-level discipline into a page the audience can act on.

Five Design Choices for a Combined Dashboard

Choice	What it does on the page
Index label on every chart	The audience knows whether they are reading representation, dispersion, similarity, or equity
Response rate next to climate scores	Climate scores never appear without the response rate that produced them
Longitudinal panel for every measure	Every chart shows the trajectory across at least four cycles
Cell-size suppression marker	Suppressed cells are visibly marked rather than silently empty
Action-tracking column	Each cycle records what was decided and what changed since

Reading the page as a triangulation

A combined dashboard is a triangulation. The composition charts answer one question, the equity charts answer another, the climate charts answer a third, and the page is credible because the three answers are visible together. When the three agree, the audience reads a clear pattern. When they disagree, the disagreement itself is the most informative finding on the page, because it tells the function which lever — composition, equity, or climate — is the binding constraint this cycle. The dashboard that lets the audience read the disagreement at a glance is the dashboard that earns its place in the strategic conversation.

19.6 Hands-On Exercise: Computing the Diversity Indices

Aim, Scenario, Dataset, Deliverable

Aim. Compute the four families of diversity indices — representation, dispersion, similarity, and equity — in Excel, then surface them on a Power BI page that respects cell-size suppression and segmentation discipline.

Scenario. You are running the diversity-and-inclusion analytics for an organisation with around 1,500 employees. The audience for the dashboard is the executive committee, and the page has to render composition by level, dispersion across business units, similarity to a market benchmark, and pay-equity gap, all on a single working surface.

Dataset. The IBM HR Analytics Employee Attrition dataset, available publicly on Kaggle at www.kaggle.com/datasets/pavansubhasht/ibm-hr-analytics-attrition-dataset. The dataset contains 1,470 employee records with relevant fields including Gender, Age, Department, EducationField, JobLevel, JobRole, MonthlyIncome, and YearsAtCompany.

Deliverable. A Diversity-Indices.xlsx workbook with the four index families computed by department, level, and tenure, and a Diversity-Indices.pbix Power BI file with a triangulation page rendering composition, equity, and climate proxies together.

19.6.1 Step 1 — Load the dataset and prepare the categorical fields

Download the IBM HR Attrition CSV. Open it in Excel, convert it to a Table named HR, and confirm the date and numeric fields type correctly. Add a TenureCohort column that bins YearsAtCompany into 0–2, 3–5, 6–10, and 10+.

19.6.2 Step 2 — Compute representation indices

Build representation by Gender, by Department, and by JobLevel using COUNTIFS.

Code

Excel Formula

Female Share by Department = COUNTIFS(HR[Gender], "Female", HR[Department], <dept>)
                           / COUNTIF(HR[Department], <dept>) * 100

Senior-Level Female Share  = COUNTIFS(HR[Gender], "Female", HR[JobLevel], 4)
                           / COUNTIF(HR[JobLevel], 4) * 100

Render representation as stacked bars by JobLevel and as a treemap by Department.

19.6.3 Step 3 — Compute dispersion indices

Compute the Blau index for gender within each Department on a Dispersion sheet.

Code

Excel Formula

Blau Index = 1 - SUMPRODUCT(p^2)

where p is the array of share-of-total values for each gender within the department. The Blau index runs from 0 (one group only) to 0.5 (perfect balance for two groups). Compute the same index by EducationField across the firm.

19.6.4 Step 4 — Compute similarity indices

Build a similarity index of dissimilarity comparing the firm’s gender mix at JobLevel 4 to a market reference share you specify on the Definition sheet (for example, 30 per cent female senior managers).

Code

Excel Formula

Similarity Gap = ABS([Senior-Level Female Share] - [Market Senior-Level Female Share])

19.6.5 Step 5 — Compute equity indices

Compute the gender pay-equity gap by Department.

Code

Excel Formula

Pay-Equity Gap (Pct) = (AVERAGEIFS(HR[MonthlyIncome], HR[Gender], "Male", HR[Department], <dept>)
                      - AVERAGEIFS(HR[MonthlyIncome], HR[Gender], "Female", HR[Department], <dept>))
                      / AVERAGEIFS(HR[MonthlyIncome], HR[Gender], "Male", HR[Department], <dept>) * 100

Surface the gap as a horizontal bar chart with department on the y-axis and the gap on the x-axis, with a target gap line drawn at the firm’s commitment level.

19.6.6 Step 6 — Apply cell-size suppression

For every chart, calculate the cell size and suppress any cell with fewer than ten employees, replacing it with a dash and a footnote. Document the suppression rule on the Definition sheet so that the discipline is auditable.

19.6.7 Step 7 — Promote to Power BI

Open Power BI Desktop and load the IBM HR table. Build the four index families as DAX measures mirroring the Excel formulas. Add a tooltip to every measure that names the index family and the cell-size threshold.

19.6.8 Step 8 — Build the triangulation page

Lay out one page with three regions: representation (stacked bars and treemap), equity (gap chart with target), and a climate-proxy panel using JobSatisfaction and EnvironmentSatisfaction distributions from the dataset. Add a longitudinal mock-up by treating YearsAtCompany cohorts as if they were measurement cycles, so the trajectory discipline becomes visible even on cross-sectional data.

19.6.9 Step 9 — Publish and instrument

Publish the page to a workspace, surface the cell-size suppression rule as a footnote on every chart, and turn on usage metrics so the equity panel’s open rate can be tracked.

Connect to the Visualisation Layer

This page extends the foundational scorecard from Chapters 7 and 8 with the diversity-and-inclusion lens. It pairs with the impact-test page from Chapter 20 and the segmentation page from Chapter 21 to form the diversity-and-workforce analytics block of Module 3.

Files and Screen Recordings

Diversity-Indices.xlsx, Diversity-Indices.pbix, and ch19-diversity-indices-walkthrough.mp4 will be attached at this point in the published edition. The screen recording walks through Steps 1 to 9 with the Excel index calculations and the Power BI triangulation page shown side by side.

Summary

Concept	Description
Why Measurement Matters
Measurement earns credibility	Measurement is where most EDI programmes either earn or lose credibility
Toolkit availability not the barrier	The literature has accumulated the techniques; honest use is the limiting factor
Honest use of measures	Use measures with boundary conditions intact and uncertainty rendered visibly
Multi-level by default	The same measure means different things at individual, team, and organisation levels
Visualisation determines reach	Diversity and inclusion measurements only matter when the audience can interpret them
Diversity Indices
Representation index	Share of the workforce by attribute drawn as stacked bar or treemap
Dispersion index	How spread the workforce is across attribute values, drawn with index value
Similarity index	How similar a team is to a reference, drawn with reference line
Equity index	Whether outcomes are equal across attributes, drawn as gap to target
Choosing the index for the question	The right index depends on the question being asked, not on convenience
Inclusion Measurement
Survey-based climate channel	Belonging, voice, fairness, manager support measured by survey
Behavioural signals channel	Promotion, mobility, voluntary exit by segment, read as funnel and survival
Listening signals channel	Open-text comments, listening sessions, employee networks read as themes
Segmentation discipline	Aggregate scores paired with break-outs by demographic, tenure, manager, role
Response-rate awareness	Climate scores rendered alongside the response rate that produced them
Longitudinal discipline	Scores shown across cycles so direction matters as much as level
Multi-channel triangulation	Survey numbers paired with behavioural and listening signals
The Multi-Level Frame
Individual level	The level of personal experience and individual outcomes
Team level	The level of climate and team dynamics
Organisation level	The level of aggregate composition and equity across the firm
Cell-size discipline	Minimum cell size, suppression, declaration, and alternative roll-up
Visualising Together
Index label on every chart	Audience knows whether they read representation, dispersion, similarity, or equity
Response rate next to climate scores	Climate scores never appear without the response rate that produced them
Longitudinal panel for every measure	Every chart shows trajectory across at least four cycles
Cell-size suppression marker	Suppressed cells are visibly marked rather than silently empty
Action-tracking column	Each cycle records what was decided and what changed since
Triangulation in Practice
Triangulation read	Composition, equity, and climate read together rather than separately
Composition-equity-climate disagreement	When the three faces disagree, the disagreement is the informative finding
Binding-constraint identification	Disagreement reveals which lever is the binding constraint this cycle
Disclosure of suppression	Suppression is declared openly to preserve trust in longitudinal comparisons