+++ title = “MMM 510: Event-Time Metrics for Marketing Decisions” date = “2026-03-26” type = “post” draft = false categories = [“posts”, “stats”] tags = [“marketing”, “causal-inference”, “mmm”, “panel-data”, “event-study”, “carryover”, “metrics”] description = “Quantitative metrics for translating event-study profiles into actionable marketing insights, including ramp-up rates, multipliers, and half-lives.” math = true +++

Translating Dynamics into Decision Metrics

While event-study plots provide a transparent visual of causal dynamics, translating these patterns into business decisions requires quantitative metrics. Section 5.10 introduces the core metrics for evaluating ramp-up, decay, and cumulative impact.

1. Diagnosing Anticipation

Anticipation occurs when customers alter their behaviour in response to an expected future intervention (e.g., delaying purchases before a promotion).

  • Pattern: Typically appears as non-zero coefficients in the periods immediately preceding treatment ($k = -2, -3$).
  • Interpretation: Positive leads may indicate stockpiling or early adoption, while negative leads suggest demand shifting.
  • Action: If anticipation is present, evaluate cumulative effects over a window that includes the pre-treatment anticipation period (e.g., from $k = -L$ to $k = K$).

2. Measuring Ramp-Up Rate

Ramp-up captures the gradual growth of treatment effects over time, common in loyalty programs or network-effect products.

  • Average Per-Period Growth: $(\hat{\theta}_K - \hat{\theta}_0) / K$. This measures how much the effect increases per period on average through horizon $K$.
  • Time-to-Maturity ($k^*$): The event time at which the effect effectively stabilises (e.g., $|\hat{\theta}_{k+1} - \hat{\theta}_k| < \epsilon$).
  • Business Rule: Do not evaluate final ROI until the program has reached maturity. Short-run evaluations of ramp-up profiles will systematically understate long-run benefits.

3. Estimating Effect Multipliers (LRM)

The Effect Multiplier $M_K$ scales the immediate effect to its total impact over horizon $K$.

  • Formula: $M_K = \sum_{k=0}^{K} \hat{\theta}_k / \hat{\theta}_0$.
  • Long-Run Multiplier (LRM): When $K$ is large enough for the effect to have stabilised, $M_K$ coincides with the LRM.
  • Caveat: Multipliers are unstable when the immediate effect $\hat{\theta}_0$ is near zero or noisy. In such cases, focus on cumulative sums.

4. Decay and Persistence (Half-Life)

For transitory interventions like TV campaigns or price shifts, understanding decay is critical for budgeting.

  • Half-Life ($k_{1/2}$): The event time at which the effect falls to 50% of its peak value (${\theta}_{peak} / 2$).
  • Persistence Ratio: $\hat{\theta}_K / \hat{\theta}_{peak}$. Values near 1 indicate persistence; values near 0 indicate transitory effects.
  • Wait Time: Sustained investment is required if the half-life is short relative to the business cycle.

5. Cumulative Effects and ROI

The ultimate summary of impact is the cumulative effect, which serves as the numerator for ROI and CLV calculations.

  • Cumulative Effect: $\sum_{k=0}^{K} \hat{\theta}_k$.
  • Discounting (NPV): For multi-year evaluations, discount distant effects: $\sum_{k=0}^{K} \hat{\theta}_k / (1+r)^k$.
  • ROI Formula: $(\text{Cumulative Effect} \times \text{Scale Factor} - \text{Cost}) / \text{Cost}$, where the scale factor converts outcome units (e.g., sales) into monetary contribution.

Summary

Event-study metrics provide the bridge between statistical profiles and financial accountability. By reporting ramp-up rates, multipliers, and half-lives alongside point estimates and confidence intervals, practitioners can provide a full causal narrative that guides both tactical optimization and strategic budget allocation.