Marketing brand alliances are consumer-facing collaborations in which brands are jointly presented through shared offerings or campaigns. Yet, despite their widespread use in practice, no existing framework supports their market-wide, ex ante discovery with execution guidance. We address this gap by formalizing brand alliance opportunity discovery as a new computational design problem characterized by two desiderata that prior work has not jointly addressed: signal fusion and actionability. We propose BANE (Brand Alliance Network Exploitation), a sociotechnical framework that leverages information networks derived from digital user traces to satisfy these desiderata through four phases: (1) predicting alliance opportunities by integrating offline brand attributes, online consumer co-mention signals, and network embeddings; (2) validating predictions through consumer and brand manager surveys; (3) qualifying predicted positive, non-materialized alliance opportunities for actionability across quadrants; and (4) generating marketing-mix execution guidance. Temporal holdout testing demonstrates that BANE predicts future alliances from pre-announcement signals, ablation analysis confirms the incremental value of each signal layer, and stakeholder surveys show that BANE-identified opportunities are perceived as significantly more promising than controls. Our work contributes three salient design insights to the IS literature: multi-layer signal fusion is necessary for identifying truly promising brand alliance opportunities; consumer co-mention networks systematically align with theory-grounded alliance covariates, providing a distinct informational input for alliance prediction; and coupling prediction with qualification and prescription enhances managerial actionability. These insights enable a scalable, data-driven alternative to survey-based alliance search while generating predictive knowledge that can motivate future causal research on the mechanisms underlying brand alliance formation.
Malhotra et al. (Mon,) studied this question.