Data Roles
Unlike a typical Power BI visual where each field well has one specific meaning, the Funnel visual exposes four data roles, and one of them — Values — is a single bucket that holds every measure the visual might need (scores, bases, comparison values, logos, extra-column measures…). Once your measures are in that bucket, you tell the visual which measure plays which role from dropdowns inside the Format pane cards (Main Partition, Second/Third Partition, Group Header, Column N…).
| Data Role | Type | Description |
|---|---|---|
| Funnel Steps | Grouping | Categorical dimension whose distinct values become the funnel's steps (e.g. Awareness, Familiarity, Consideration, Purchase) |
| Group By | Grouping | Optional grouping that repeats the whole funnel per value (e.g. one funnel per Country) |
| Values | Measure (many) | All measures: scores, bases, comparison values, logos (base64)… assigned to a role from the Format pane |
| Tooltips | Measure (many) | Optional. Extra measures shown only on hover — see Tooltips below |
Funnel Steps
Drag a categorical column into Funnel Steps. Each distinct value becomes one bar in the funnel. The visual does not require the column to be pre-sorted: it automatically orders the steps by descending score, based on the Main Partition → Score measure of the first group returned by the data view (the first Group By value, or the only group if Group By is unbound). That order is then applied identically to every group, so all funnels in the visual stay aligned on the same step sequence.
If your groups don't all decrease monotonically in the same order, the bars in the other groups may not look strictly descending — only the reference group (the first one) is guaranteed to. This is expected: the step order must be identical across all funnels so comparisons line up.
Group By
Optional. Bind a column here (e.g. Country, Brand) to draw one funnel per distinct value, side by side in the same visual. Leave it unbound to draw a single funnel.
Values
Drag every measure you need into this single well — there is no need to bind them one at a time to specific roles. Once they're in the Values bucket, each Format pane card that needs a measure exposes its own dropdown (Score, Comparison measure, Comparison base measure, Logo…) listing everything available in Values. See:
- Main Partition — the primary funnel's Score / Base / Comparison / Comparison base
- Second Partition / Third Partition — competitor/benchmark comparison scores
- Group Header — optional per-group logo measure
- Columns — up to 8 extra per-step measures
Score, Base, Comparison, Comparison base — what they mean
These four measure roles power both the on-bar value and the statistical significance test:
| Role | Meaning |
|---|---|
| Score | The group's value at this step — typically a percentage of observations/customers who reached this step |
| Base | The group's observation/sample count at this step — the denominator behind Score |
| Comparison | The reference value to test the group's Score against. Despite the name, it doesn't have to be a "segment average" — it can be any comparable figure: another category, another year, another product line, a market benchmark… |
| Comparison base | The comparison's observation/sample count — the denominator behind Comparison |
Only Score is mandatory. Base + Comparison + Comparison base are optional, but all three are required together to unlock significance testing (colored leakage pills and column cells) for a given step — if any of the three is missing or non-positive, that step's significance falls back to neutral.
Score, Comparison and any per-column equivalents are expected to be numbers on a 0–100 scale (e.g. 42.5 for "42.5%"), because the significance formulas divide by 100 internally. Format these measures as a plain number in your model (e.g. 0.0) rather than Percentage — a Percentage format whose underlying value is already 0–100 would be displayed ×100 too high on the funnel bars.
Significance testing
A standard two-proportion z-test runs in two places, each with its own Significance level (90/95/99%) and Independence setting:
- Leakage pills — is the group's step-to-step drop-off significantly worse (red) or better (green) than the comparison's own drop-off between the same two steps? Configured on the Leakage card.
- Column cells — is the group's Score at this step significantly above (red) or below (green) its own Comparison? Configured per column, in the Additional Table dialog's Significance section.
Set Independence to something other than its default ("Main separate from comparison") when your Comparison measure isn't a genuinely separate sample from Main — see Leakage → Independence for the three options and when to use each. This matters for both the main funnel (via the Leakage card) and any column whose Score/Comparison roles are wired the opposite way round from Main Partition.
Tooltips
Optional. Measures dropped here are not drawn anywhere on the funnel — they only appear in the tooltip when the reader hovers a bar, appended after the rows the visual builds on its own (group, step, score and step-to-step leakage).
They are also deliberately absent from the Format pane dropdowns: a Tooltips measure can never be picked as a Score, Base or Comparison, so adding one can't disturb an existing configuration.
Because the visual declares canvas tooltip support, a whole report page can be used as its tooltip instead of the default list — the hovered bar's identity (group × step) is passed to that page so it filters down to the step under the cursor.
Logos
Both the Group Header logo and the Main/Second/Third Partition logos accept base64-encoded images, exactly like the MetricTile's Logo role: if the value does not start with data:, the visual automatically prepends data:image/png;base64,. Keep them small (a few KB, e.g. 24×24 to 40×40 px source) to avoid bloating the data model.
Extra columns
Beyond the funnel bars, up to 8 extra columns of data can be shown per step (e.g. NPS, satisfaction, repeat-purchase rate, a rank). Each column has its own Score type — Continuous (a numeric measure, with Base / Comparison / Comparison base / Precision / Display / colors, exactly like the main funnel) or Discrete (an arbitrary text/ordinal value such as "1st"/"2nd", shown as-is with no comparison or coloring). Continuous columns support six display modes (Value, Gap vs. the main score, Comparison value, Comparison Gap, and the "both" combinations of each). See Columns and Additional Table for the two ways to configure them.