What This Chart Shows
This chart helps you see which channels or tactics are worth scaling, optimizing, or rethinking. Each bubble is a lever (or a group like a channel or campaign). Its position shows efficiency and impact: one axis is return (e.g. ROI or attribution), the other is impact (e.g. marketing contribution or lift). Bubble size reflects total impact so you can tell big drivers from small ones.
The chart is split into four quadrants (by median) so you can quickly spot "scale these," "grow these," "optimize these," and "reassess these." In 3D mode, a third axis shows how much of the result is explained by the model versus other factors.
Key Questions This Chart Helps Answer
- Which levers or channels should we scale, and which should we optimize or reassess?
- How do ROI (or attribution) and total impact vary across the portfolio?
- Which levers have the largest impact?
Axes, Metrics, and Units
Element | Description |
|---|---|
| X-axis | Costs mode: Marketing Contribution in outcome or currency units. No-costs mode: Marginal Lift per target. |
| Y-axis | Costs mode: Attributed ROI No-costs mode: Marketing Contribution or Marginal Lift |
| Bubble size | Proportional to absolute impact (absolute lift/contribution), normalized so the largest bubble is visually prominent. Size is for comparison only, not a separate unit scale. |
| Color | Quadrant: Green = Scale Priority (high Y, high X). Blue = Growth Opportunity (high Y, low X). Orange = Optimization Target (low Y, high X). Red = Reassessment Required (low Y, low X). Quadrants use the median of displayed points. |
| Reference lines (2D) | Dashed horizontal and vertical lines at median Y and median X (quadrant splits). Costs mode only: solid break-even line at Y = 1 with "Break-even" label. |
| 3D mode | Z-axis: "Model Reliance %" — share of total model-attributed effect allocated to that lever (from AI residual allocation). X and Y are the same as in 2D. |
Control Options Reference
Control | Meaning |
|---|---|
| KPI classification / hierarchy / drilldown | Filter or drill by KPI. |
| Plot type | 2D (default) or 3D. 3D adds Model Reliance % on z-axis. |
| Grouping mode | Lever (default), Series, Channel, Campaign, Funnel, Source Type, or Topic. Aggregates levers into groups and plots one bubble per group. |
| Date range | Time window for attribution. |
| Event categories | Filter by specific event category (include/exclude). |
| Funnel stages | Filter by specific funnel stage (include/exclude). |
| Search terms | Text search on lever/series labels. |
How to Interpret the Results
- Q1 (Scale Priority, green): High ROI and high lift. Best candidates to scale spend, assuming no saturation.
- Q2 (Growth Opportunity, blue): High ROI, lower lift. Smaller but efficient; consider incremental budget or testing scale.
- Q3 (Optimization Target, orange): Lower ROI, high lift. Big impact but less efficient; optimize creative, targeting, or frequency before scaling.
- Q4 (Reassess, red): Low ROI and low lift. Review targeting, creative, or whether to reduce or pause.
- Magnitude: Larger bubbles = larger total impact. Compare both position (efficiency vs impact) and size (how much they move the needle).
- Break-even (costs mode): Bubbles above the Y = 1 line are above break-even; below are sub–break-even. Use with quadrant to decide scale vs optimize vs reassess.
- 3D: Higher Z = more of the attributed effect for that lever comes from the model's allocation. Use to see where the model is most vs least influential.
- Truncation: Only the top N points by absolute Y (impact) are shown; use filters or grouping to focus. Metadata can indicate “Showing X of Y.”
- Uncertainty: The chart shows point estimates (ROI, lift, contribution). It does not show confidence intervals; use ROI bars or other views for uncertainty.
Practical Applications for Marketers
Application | How to use this chart |
|---|---|
| Budget reallocation | Move spend from Q3/Q4 toward Q1/Q2; use size to respect total impact. |
| Scale decisions | Identify Q1 and high-size Q2 levers for scale tests; watch saturation elsewhere. |
| Optimization focus | Use Q3 for efficiency work (targeting, creative, frequency) before adding spend. |
| Pause/reduce | Use Q4 and below-break-even levers as candidates to cut or reassess. |
| Portfolio review | Group by channel or campaign to see quadrant mix and balance. |
| Reporting | Use quadrant labels and break-even to explain “where we stand” to stakeholders. |
Common Mistakes and Misinterpretations
Mistake | Why it is a problem | How to avoid |
|---|---|---|
| Treating quadrant as “good vs bad” only | Q3 (optimize) can be large and strategic; Q2 can be small. Quadrant is efficiency + impact, not just quality. | Use quadrant for action (scale / grow / optimize / reassess), not as a single score. |
| Ignoring bubble size | A small green bubble may be efficient but low impact; a large orange one may matter more for total results. | Always combine quadrant with size (and break-even) when prioritizing. |
| Comparing ROI across different grouping modes | Lever-level vs channel-level ROI are different aggregations; group ROI = sum(lift)/sum(cost). | Compare within the same grouping and same filters. |
| Assuming all levers are shown | Only top N by impact are displayed; low-impact levers can be omitted. | Use filters or grouping to see the slice you care about; check truncation message if present. |
| Reading 3D “Model Reliance %” as quality | It’s the share of attribution from the AI residual; high % can mean less interpretable, not necessarily wrong. | Use for interpretability and validation focus, not as a single quality score. |
Caveats and Considerations
- Data dependency: Requires attribution data. If missing, the chart shows “Data not available.” With no valid points after filters, it shows “No valid data points.”
- Costs mode vs no-costs: When cost data is absent or zero, the chart shows Attribution vs Marginal Lift (no ROI, no break-even). Axis labels and title change accordingly.
- ROI when cost is near zero: Levers with cost below a small threshold are excluded from the scatter (ROI undefined). They do not appear as points.
- Aggregation: In grouped mode, group ROI = sum(lift) / sum(cost) per group, not the average of per-lever ROIs. This keeps the ratio consistent with the waterfall and avoids aggregation artifacts.
- Quadrant bias from truncation: Showing top points by impact can over-represent Scale and Growth (high impact) and under-represent Optimize and Reassess. Use filters or grouping to reduce cardinality and get a more balanced view.
- Assumptions: Attribution and ROI come from the causal graph. Results depend on graph structure, lags, and data quality; use alongside other diagnostics and business context.