What This Chart Shows
This chart shows how much extra lift you get as you scale spend or activity. The curve is the model’s predicted lift at different spend/multiplier levels, so you can see saturation: where lift flattens as you spend more. A large dot on each line marks current spend or 1.0x activity. Use it to decide where to scale, where you’re already saturated, and how much headroom each lever has.
Key Questions This Chart Helps Answer
- Where do I still get strong incremental lift if I spend more?
- Which levers have the most headroom vs already flat curves?
- What would happen to lift if I increased or decreased spend for a channel?
- What is my current position on the curve (large dot) vs potential at higher spend?
Axes, Metrics, and Units
Element | Description |
|---|---|
| X-axis | Cost mode: Total spend in dollars. Grid is current cost times multiplier (0x to 50x). No-cost mode: Activity multiplier (0x to 50x). 1.0 = current activity level. Same multiplier grid, no dollar scale. |
| Y-axis | Incremental KPI lift. Modeled output from full-graph propagation at each multiplier; not raw observed KPI. |
| Lines | One line per lever (up to 12). Color by lever. Spline-smoothed; each point = lift at that x (multiplier or spend). |
| Large dot | Current level: at 1.0x multiplier (no-costs) or at current spend in dollars (costs). |
| Brush range (optional) | Shaded band on the chart for a selected x range; label "Selected spend range." Used to link to other views (e.g. allocation). |
Control Options Reference
Control | Meaning |
|---|---|
| KPI classification / hierarchy | When a KPI is selected, response curves (and attribution alignment) use that target’s block. |
| Brush range | Optional to highlight a spend or multiplier range on the chart. |
| Event categories | Filter by specific event categories (include/exclude). |
| Funnel stages | Filter by specific funnel stages (include/exclude). |
How to Interpret the Results
- Steep slope (more lift per dollar or per 0.1x multiplier) = room to scale; incremental spend/activity is still effective.
- Flat slope near the large dot = saturated at current level; more spend adds little lift.
- Curvature: Often curves flatten at high x (diminishing returns). Where they bend is where saturation starts.
- Compare lines: Higher curves = more lift at same spend; steeper early slope = more responsive. Use for prioritization and reallocation.
- Large dot: Everything to the right is "if we spent more"; everything to the left is "if we spent less." Use it to judge scale-up or cut-back.
- Magnitude: Higher y = more lift at that x. Compare both slope and level across levers.
- Uncertainty: The chart shows point estimates (no confidence bands). Use other diagnostics (e.g. ROI CIs) for uncertainty.
Practical Applications for Marketers
Application | How to use this chart |
|---|---|
| Budget allocation | Find levers whose large dot is still on the steep part; consider increasing spend there first. |
| Saturation check | If the curve is flat past the large dot, avoid adding more spend; consider shifting to other levers. |
| Campaign evaluation | Compare curves across campaigns or periods (same chart filters) to see who has headroom. |
| Planning and forecasting | Use curve shape to set realistic lift expectations when increasing or decreasing spend. |
| Channel mix | View multiple levers to see which channels have more headroom vs saturation. |
Common Mistakes and Misinterpretations
Mistake | Why it is a problem | How to avoid |
|---|---|---|
| Treating the curve as guaranteed outcome | Curves are model predictions; real lift can differ. | Use as a guide for direction and relative headroom, not a guarantee. |
| Ignoring the current-level dot | Without it, you don’t know where you are on the curve. | Always use the large dot to judge “current vs potential” and saturation. |
| Comparing curves with different units | No-costs (multiplier) vs costs ($) change the x-axis meaning; mixing KPIs changes y. | Compare within the same mode and same KPI/outcome group. |
| Assuming all levers are shown | Only a subset (e.g., top 12 by signal) is plotted; others are omitted. | Use filters to focus; if a lever isn’t there, it may be low-signal or filtered out. |
| Extrapolating far beyond the range | Curves are defined on a finite multiplier/cost grid; behavior beyond the last point is not shown. | Stay within the plotted range when interpreting “what if” spend. |
Caveats and Considerations
- Data dependency: If curves are missing or empty, the chart shows a diagnostic message and steps to populate.
- Display limit: Up to 12 levers are shown (those with the strongest curve signal after filters). Use event/funnel filters to focus on the levers you care about.
- X-axis scope: Only the non-negative multiplier segment (0x to 50x) is shown. Negative multipliers (cannibalization) are used in the model but not drawn here.
- Assumptions: Curves come from the causal graph. Results depend on graph structure, saturation parameters, and data; use with other diagnostics and business context.