LCM-Promo: An LLM-Based Uplift Model for Promotion Allocation

Nesta Midavaine; Floris Fok; Paulo Mendes; Joao Silva; Önder Akaçık; Dimas Lima; Zülküf Genç

cs.LGOct 27, 2026

LCM-Promo: An LLM-Based Uplift Model for Promotion Allocation

Authors: Nesta Midavaine, Floris Fok, Paulo Mendes, Joao Silva, Önder Akaçık, Dimas Lima, Zülküf Genç

Organizations: Prosus, Amsterdam, The Netherlands · iFood, São Paulo, Brazil

Abstract

We present LCM-Promo, an LLM-based uplift model for promotion allocation. LCM-Promo is a task-specific Large Commerce Model (LCM): a finetuned LLM trained on transactional commerce data to predict user intent and promotion response. It learns monotonic promotion-response curves from production logs by interpreting single-token outputs as conversion-threshold probabilities, while conditioning on session context such as day and delivery shift. In a production A/B test on churn reactivation, LCM-Promo increased conversion by 1% and reduced promotion cost by 6% relative to the Segment-EMA production baseline.

Figures & tables

Metric	Relative lift
Conversion	$+1\%$
Promotion cost	$-6\%$

Table 1: Online A/B test. Lifts relative to Segment-EMA baseline.

Model	AUUC	Qini	Uplift@30
Uplift Random Forest	$0.0578\pm 0.0064$	$0.0466\pm 0.0053$	$\bm{0.1708\pm 0.0064}$
LCM-Promo (post-processing)	$\bm{0.0608\pm 0.0032}$	$\bm{0.0490\pm 0.0028}$	$0.1626\pm 0.0035$
LCM-Promo (waterfall)	$0.0278\pm 0.0036$	$0.0226\pm 0.0029$	$0.1208\pm 0.0026$

Table 2: Offline uplift metrics on churn users.

Uplift Modeling

See all 13 papers

ECP-Bench: Benchmarking and Learning Entertainment Content Promotion with Foundation Models
Aug 26, 2026Hyomin Kim, Bowen Chen, Jin Huang +3LLM Evaluation Domain Generalization
Sony Group Corporation · KAIST · University of Michigan
LLMAdBench: A Human Preference Benchmark for Advertising in LLM Responses
Sep 26, 2026Rui Ai, Yuqing Liu, Sitao Qiu +14Language Model Generation Evaluation Human Preference Modeling
Massachusetts Institute of Technology · University of Michigan, Ann Arbor · Shanghai Jiao Tong University +7
Just Ask for a Table: A Thirty-Token User Prompt Defeats Sponsored Recommendations in Twelve LLMs
May 12, 2026Andreas Maier, Jeta Sopa, Gozde Gul Sahin +2LLM Auditing LLM Prompting
Pattern Recognition Lab, Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany

Aug 26, 2026J/K move · Enter open · S save

ECP-Bench: Benchmarking and Learning Entertainment Content Promotion with Foundation Models

Hyomin Kim, Bowen Chen, Jin Huang, Zhao Wang+2

Sony Group Corporation · KAIST · University of Michigan

LLM Evaluation Domain Generalization LLM Post-Training

Open paper AskPDF arXiv

Content promotion spans a broad set of skills, from understanding content to forecasting its market reception. However, LLMs' ability to support such promotion decisions remains underexplored. Existing studies are often limited to a single task (e.g., popularity prediction) or a small set of tasks within a single domain (e.g., movies). As a result, there is a lack of understanding of LLMs' abilities in the full promotion process and how these abilities generalize across different tasks and domains. In this work, we introduce ECP-Bench, a benchmark containing 1.9M movie, game, and music items and 423,451 questions across 33 tasks in five content-promotion skill families. Our evaluation shows that frontier models achieve only 51.9% overall accuracy and lose much of their advantage on post-cutoff content, with drops of up to 19.1 percentage points. In contrast, open-weight models fine-tuned on ECP-Bench achieve up to 60.3%, remain substantially more stable across the knowledge cutoff, generalize to unseen content and tasks, and exhibit meaningful cross-domain generalization.

Abstract

Uplift Modeling

Explore similar work

ECP-Bench: Benchmarking and Learning Entertainment Content Promotion with Foundation Models

LLMAdBench: A Human Preference Benchmark for Advertising in LLM Responses

Just Ask for a Table: A Thirty-Token User Prompt Defeats Sponsored Recommendations in Twelve LLMs