FunnelCausalNet: Funnel-aware Joint Conversion-Revenue Uplift for Multi-tier Coupon Allocation

2026-08-12

AI Summary

This study addresses the challenge of jointly optimizing conversion rate and GMV in multi-tier coupon allocation, where GMV exhibits zero-inflation, heavy-tailed distribution, and a conversion–value funnel structure. To this end, we propose FunnelCausalNet, the first causal uplift model that explicitly incorporates the conversion–value funnel by jointly modeling binary conversion and conditional order value, with GMV estimated as the product μ_GMV = μ_conv × μ_val. Our approach integrates a shared representation network, margin-split conformal CATE aggregation, an RCT-anchored Lagrangian budget allocator, and Bonferroni-corrected joint audit bands to enable subsidy-aware ROI optimization. Evaluated on Criteo-MT7, FunnelCausalNet matches the best baseline in GMV AUUC while reducing treatment effect estimation error by 18–48%; on hotel RCT data, it achieves the highest average ΔROI across all seven anchor points.

Abstract

Coupon campaigns seek to lift both conversion and revenue, but gross merchandise value (GMV) follows a deterministic funnel from conversion to conditional order value and is zero-inflated and heavy-tailed. We propose FunnelCausalNet, an uplift estimator coupling a binary conversion head with a nonnegative conditional-value head through $μ_{\mathrm{gmv}}=μ_{\mathrm{conv}}μ_{\mathrm{val}}$. Under explicit RCT, support, rate-gap, and cross-head covariance-control assumptions, an idealized leading-order MSE comparison identifies a regime in which funnel composition can reduce pointwise variance; this is a heuristic, not a guarantee for the shared-representation neural model. The estimator is paired with marginal split-conformal CATE summaries, combined through a Bonferroni union as audit bands, and a Lagrangian budgeted allocator using RCT-anchored estimates for subsidy-aware ROI accounting. On semi-synthetic multi-tier Criteo-MT7, FunnelCausalNet's mean AUUC_GMV is within one seed standard deviation of the leading feature-interaction baseline among eleven baselines, while a controlled ablation reduces GMV effect error versus direct GMV regression by 18--48% across tested zero-inflation regimes. On de-identified industrial Hotel-Coupon RCT logs with about 4.9 million hold-out exposure records per seed, expected-outcome evaluation sweeps full LP frontiers; FunnelCausalNet has the best seed-averaged mean DeltaROI at all seven correlated anchors from 10% to 60%, which we treat as descriptive frontier consistency rather than independent significance. On sparse binary-spend public benchmarks, revenue-focused rankers can dominate uplift-curve proxies, defining an explicit regime boundary.

Problem

• uplift modeling
• coupon allocation
• conversion-revenue funnel
• zero-inflated GMV
• causal inference

Innovation

• FunnelCausalNet
• uplift modeling
• conversion-revenue funnel
• conditional value estimation
• budgeted coupon allocation