What This Chart Shows
This chart shows how long it takes for marketing to show up in results across your causal graph. Each edge (link from one node to another) has a lag in days—the delay between cause and effect. The chart plots a distribution of those lags in days and how many edges have that lag (and a weight-weighted version) allowing you to see whether effects are mostly same-day (lag 0), short (e.g. 1–7 days), or long (e.g. weeks). A mean effective lag is shown at the bottom. There is no time series—it is a snapshot of "how long does it take for marketing to produce results" across all edges.
Key Questions This Chart Helps Answer
- How long does it take for marketing to produce results?
- Are effects mostly same-day, short-term, or long-term?
- What is the typical (mean effective) delay between cause and effect?
- How many edges have very short vs very long lags?
Axes, Metrics, and Units
Element | Description |
|---|---|
| X-axis | Causal lag (days). Time delay from source node to destination node for each edge. |
| Y-axis (left) | Number of edges. Count of edges in each lag bin. |
| Y-axis (right) | Weight-weighted count. Same units as edge weights. |
| Bars or line | Histogram mode: Bars = edge count per lag bin.
Scatter/line mode: Line + markers = same counts. |
| Second trace | Weight-weighted: line + markers on right y-axis (dashed in scatter mode). Only shown when edge weights are available. |
| Footnote | "Mean effective lag: X.X days" — weighted average of lag (weights = edge weights when present, else equal weight). |
Control Options Reference
Control | Meaning |
|---|---|
| Plot type / Dist. | Histogram (hist): Bar chart of edge counts by lag.
Scatter: Line + markers for the same counts. |
Note: This chart does not use event categories, funnel stages, grouping mode, outcome group, or date range.
How to Interpret the Results
- Peak of the distribution: The lag (or bin) with the most edges is the most common delay in your graph. Use it to see if effects are mostly same-day, a few days, or longer.
- Spread: Wide spread = mixed short and long lags; narrow spread = consistent delay. Long tail to the right = some edges with long lags.
- Mean effective lag: One-number summary; when weights are used, it emphasizes lags of stronger edges. Use it for attribution windows or reporting.
- Weight-weighted curve (if shown): Where this line is high, most of the "impact mass" sits at that lag. Compare with the count distribution to see if important edges are short- or long-lag.
- Magnitude: Higher bars (or higher line) = more edges at that lag. No time axis—all edges are aggregated into one distribution.
Practical Applications for Marketers
Application | How to use this chart |
|---|---|
| Attribution windows | Use mean lag and distribution shape to set how many days after a touch to credit marketing (e.g. 7-day vs 30-day windows). |
| Channel and tactic planning | Compare with channel-level views: if most lags are short, optimize for quick feedback; if long, plan for delayed impact. |
| Reporting and storytelling | Use "how long does it take for marketing to produce results" and mean effective lag in decks and one-pagers. |
| Model and data check | Odd shapes (e.g. all zeros, or a single spike) may suggest data or model issues; use with other diagnostics. |
| Expectation setting | Use distribution and mean to set expectations with stakeholders on when to expect results after spend or campaigns. |
Common Mistakes and Misinterpretations
Mistake | Why it is a problem | How to avoid |
|---|---|---|
| Treating lag as "time to conversion" from first touch | Lag here is per edge (source to destination), not path-level or journey-level time to conversion. | Interpret as "delay between cause and effect on one link"; use other analyses for full journey length. |
| Treating the chart as a time series | The x-axis is lag in days, not calendar time. Each bar/bin is "how many edges have this lag." | Read as a distribution of delays, not "what happened over time." |
| Ignoring the mean effective lag | The footnote is the main summary; ignoring it can lead to wrong attribution windows. | Use mean effective lag when choosing or explaining attribution windows. |
| Comparing across different graphs or pipelines | Lags come from the causal graph and pipeline; different graphs or TE/lag discovery give different distributions. | Compare only within the same model/graph; do not mix with other systems' lag definitions. |
Caveats and Considerations
- Data dependency: Requires lag data from the causal graph. If lags are missing or empty, the chart shows “No lag data” and no histogram.
- Per-edge lags: Lags are between the two nodes of each edge (time delay from source to destination), not from campaign start to conversion or from first touch to conversion.
- Portfolio-level only: No filters by event category, funnel stage, or grouping—the chart reflects the full graph. For segment-specific timing, use other temporal or attribution views.
- Binning: Up to 50 bins from 0 to max lag; histogram is deterministic. Very large max lag can make bins wide.
- Assumptions: Lags come from the causal graph and pipeline. Results depend on graph structure and data; use with other diagnostics and business context.
- Uncertainty: The chart shows modeled lags (point estimates); it does not show confidence intervals.