Skip to main content

Leakage

The Leakage card is a composite card grouping two concerns: the Pills drawn at each junction between two consecutive bars (showing the step-to-step drop-off (%) for the main funnel), and an optional Legend explaining what the pill colors mean.

Pills

PropertyDescriptionDefault
Show leakage pillsMaster switchOn
Pill height (px)Height of the pill20
Font sizeText size inside the pill11
Significance levelConfidence level for the two-proportion z-test: 90%, 95% or 99%95%
IndependenceHow Main and Comparison overlap — see Independence belowMain separate from comparison
Neutral colorPill border/text color when the drop-off isn't significantly different from the comparison#888888
Above average color (bad)Pill border/text color when the group's leakage is significantly worse than the comparison's#cc0000
Below average color (good)Pill border/text color when the group's leakage is significantly better than the comparison's#007700

How leakage is computed

For each step after the first: leakage % = (1 − score / previousScore) × 100.

How the color is decided

If Main Partition's Base, Comparison and Comparison base are all set for both the current and previous step, the visual runs a two-proportion z-test comparing the group's conditional drop-off against the comparison's own drop-off between the same two steps, using this card's Significance level and Independence. Otherwise the pill uses the Neutral color.

Above = bad, Below = good

"Above average" means the group is leaking more than the comparison at that junction (bad) — colored with Above average color. "Below average" means the group is leaking less (good) — colored with Below average color. Don't confuse this with the Columns card, whose Above/Below refer to the raw Score, not leakage.

Significance level in practice
Levelz-threshold
90%1.645
95% (default)1.96
99%2.576

A higher confidence level (99%) requires a larger gap between the group and the segment before a leakage pill or column cell is colored — fewer false positives, but also fewer cells flagged.

Independence

Two samples sharing observations breaks the independence assumption a two-proportion z-test relies on. Independence tells the test how Main and Comparison actually relate so it can correct for that before running:

OptionWhen to use itWhat happens
Main separate from comparison (default)Main and Comparison are genuinely distinct samplesNo adjustment
Main is part of comparisonComparison is computed over a wider sample that includes Main (e.g. Comparison = "all brands", Main = one brand within it)Main's weighted contribution is subtracted out of Comparison before testing
Comparison is part of MainThe reverse nesting — Main is the wider sample and Comparison is a subset of it (e.g. Score = a market-wide average, Comparison = a single competitor within that market)Comparison's weighted contribution is subtracted out of Main before testing
Picking the wrong direction silently kills significance

If Comparison's base is actually larger than Main's, "Main is part of comparison" is the correct choice — but if you pick it when the nesting actually runs the other way (Main's base is the larger one), the subtraction produces a negative or out-of-range adjusted sample and the test falls back to neutral for every step, no matter how large the real gap is. If pills stay stubbornly gray with an option other than "Main separate from comparison" selected, double-check which of Main/Comparison is really the larger, containing sample.

Each extra column has its own Significance level and Independence, independent from this card's — useful when a column's Score/Comparison roles are wired the opposite way round from the main funnel (see the example above).

Ascending funnel

When Global Settings → Ascending funnel is on, this card is renamed Growth in the Format pane and its Show leakage pills switch becomes Show growth pills (likewise, Competition Leakage becomes Competition Growth) — same cards, same properties, worded for a gain rather than a drop-off. The pill formula and colors are unchanged: the figure is still (1 − score / previousScore) × 100 against the step above, and Above/Below still mean underperforming / outperforming the comparison. The Neutral / Above / Below label legend texts are free-form and aren't rewritten automatically — edit them yourself if you want growth wording (e.g. "Significantly higher leakage" → "Significantly lower growth").

Legend

An optional legend can be drawn below the funnel, explaining what each pill color means in plain text — useful when the visual is shared without this documentation at hand.

PropertyDescriptionDefault
Show legendMaster switch for the legend groupOff
AlignmentLeft / Center / RightCenter
Margin top (px)Space between the funnel and the legend4
Font / Size / Bold / Italic / UnderlineLegend text typographySegoe UI, 10px, not bold
Neutral labelText next to the neutral color swatch"Not significantly different"
Above labelText next to the above/bad color swatch"Significantly higher leakage"
Below labelText next to the below/good color swatch"Significantly lower leakage"
SuffixOptional extra text appended after the labels (e.g. "vs segment")

The legend is only drawn when Show leakage pills is also on and at least one step in the data actually has a computed leakage value; an entry is skipped if its label is left empty. The visual reserves vertical space for the legend below the funnel so it never overlaps the bars.