Sponsored product plays a key role at advertising business which attracts an increasing number of advertisers seeking better platforms to promote their products. Connecting advertiser's products (demand) to the customer interests (supply) becomes a crucial factor in forecasting advertising campaign success. In this work, we propose a general framework for learning the inverse product retrieval process, a more complex procedure compared to the conventional forward approach. The first set of challenges arise due to a significant level of noise in the retrieval data caused by users who deviate from the intended platform usage. To address this, we propose a contrastive learning approach that minimizes the effect of such noise in data, ensuring robustness against false positives and false negatives. In addition, we propose a calibration procedure for the inverse retrieval that handles dynamic size of a set of queries that may retrieve any particular product of interest. Our framework is universally applicable across various contexts of an advertising platform, including but not limited to search pages, item pages, and browsing pages. We showcase the efficacy of the proposed approach using Walmart's advertising data in ten product domains.

Ads demand forecasting for Walmart Connect’s ad products plays a critical role in enabling effective resource planning, allocation, and overall management of ads performance. In this paper, we introduce a comprehensive demand forecasting system that tackles the hierarchical time series forecasting problem in business settings. Though traditional hierarchical reconciliation methods ensure forecasting coherence, they often trade off accuracy for coherence especially at lower levels and fail to capture the seasonality patterns unique to each time-series in the hierarchy. Thus, we propose a novel framework "Multi-Stage Hierarchical Forecasting Reconciliation and Adjustment (Multi-Stage HiFoReAd)" to address the challenges of preserving seasonality, ensuring coherence, and improving forecasting accuracy. Our system first utilizes diverse modeling techniques, ensembled through Bayesian Optimization (BO), achieving individual base forecasts. The generated base forecasts are then passed into the Multi-Stage HiFoReAd framework. The initial stage refines the hierarchy using Top-Down forecasts and "harmonic alignment." The second stage aligns the higher levels’ forecasts using MinTrace algorithm, following which the last two levels undergo "harmonic alignment" and "stratified scaling", to eventually achieve accurate and coherent forecasts across the whole hierarchy. Our experiments on Walmart’s internal Ads-demand dataset and 3 other public datasets, each with 4 hierarchical levels, demonstrate that the average Absolute Percentage Error (APE) from the cross-validation sets improve from 3% to 40% accross levels against BO-ensemble of models (LGBM, MSTL+ETS, Prophet) as well as from 1.2% to 92.9% against State-Of-The-Art models. In addition, the forecasts between all hierarchical levels are proved to be coherent. The proposed framework has been deployed and leveraged by Walmart’s ads, sales and operations teams to track future demands, make informed decisions and plan resources strategically.

This paper solves a specialized regression problem to obtain sampling probabilities for records in databases. The goal is to sample a small set of records over which evaluating aggregate queries can be done both efficiently and accurately. We provide a principled and provable solution for this problem; it is parameterless and requires no data insights. Unlike standard regression problems, the loss is inversely proportional to the regressed-to values. Moreover, a cost zero solution always exists and can only be excluded by hard budget constraints. A unique form of regularization is also needed. We provide an efficient and simple regularized Empirical Risk Minimization (ERM) algorithm along with a theoretical generalization result. Our extensive experimental results significantly improve over both uniform sampling and standard stratified sampling which are de-facto the industry standards.