Leading the science rigor with business strategy of Ad measurement methodologies, implementations, and customer-facing tools for Walmart Ads

Fixed Marketing Measurement: I provided causal inference interpretations of models from different scientific communities (ie econometric causality, potential outcomes, deep-learning models, longitudinal analyses) in the context of brand marketing measurement. Also, I consulted and provided recommendations to the Amazon-wide efforts to align and certify marketing measurement models and Randomized Control Trials from multiple science teams and business units. Retail Pricing: I designed and built a prototype decision making system to contextualize the Amazon Retail Pricing policy of matching or not competitor prices in the presence of 3P regional vendors. I led the development of the project: 1) getting leadership sign-off, 2) estimating customer-level price elasticities for all active Amazon users in US, 3) optimizing price-matching decisions within a statistical decision making for a national pricing policy, 4) executing and analyzing the product-level A/B testing of the system.Brand Lift Measurement: I designed and built from the ground an incrementality attribution prototype to scale brand advertising effect measurement on customer sentiment for all Amazon homepage visitors using onsite surveys. This product is automatic, addresses the campaign target selection with Double Machine Learning techniques, and corrects the responder self-selection bias with propensity score techniques leveraging big data ML models. I regularly provided consulting to the Amazon Ads incrementality measurement and optimization initiatives.

I led the design of the incrementality testing platform for both Yahoo! ad network and its DSP platforms, partnering with product managers, architects and engineering, leading to 2 pending patents. I led the science development and data analytics of all incrementality tests run in the company, delivering results directly to large advertising customers and partnering with sales leaders. Lead scientist for the development of CTV and TV offline measurement modeling for bias correction and offline calibration for a real-time reporting product. The incremental revenue impact of this work was tens of millions of dollars per year. I mentored one software engineer and two data analysts.Research Conference tutorials: - https://www.cikm2021.org/programme/tutorials/online-advertising-incrementality-testing-practical-lessons-and-emerging-challenges- https://joel-barajas.github.io/kdd2021-incrementality-testing/CTV and Linear TV reach measure for real-time reporting release (featured by the CEO):- https://www.linkedin.com/posts/gurug_blog-new-unified-tv-report-verizon-media-activity-6821504624785260544-xRI9/- https://www.thedrum.com/news/2021/07/14/verizon-media-bridges-linear-and-digital-with-unified-measurement-tool- https://www.adtech.yahooinc.com/post/understand-incremental-reach-with-the-yahoo-unified-tv-reportDigital and Linear TV Cross-Screen Planner release (featured by the CBO):- https://www.linkedin.com/posts/ivanmarkman_bridging-the-gap-between-linear-and-digital-activity-6643543666545831936-Moli

Data Science leading role in the development of Media Mix Models for optimal budget allocation. We made budget recommendations and channel effectiveness assessments with time-series based models, synthetic control techniques, and the support of A/B testing. I worked cross-functionally with Marketing Managers, Ad technology team, engineering, and product in the allocation of the US/CAN budget. Through rigorous experiments, we saved millions of dollars of inefficient marketing channels.I assessed A/B tests analysis for product experiments of new features, in loyalty programs, Login and On-boarding funnels.I mentored one junior data scientist and four data analysts. I author/co-authored a statistics course and a Experimentation course for internal DS/DA community.

I designed strategies to re-distribute audio ad loads among listeners end-to-end, from designing Machine Learning based strategies to setting up A/B test experiments and analyzing results of the strategies. I championed ML based strategies with listener implicit feedback to audio ads. I covered all aspects of the project from production coding to engaging in the vision of the project product management and assessing strategic decision making. These developments improved the established rule based strategy in long-term A/B test experiments in retention metrics and revenue. References to my contributions appeared in the company earnings report. One of the most successful strategies I championed improved retention metrics worth millions of dollars, measured by A/B testing.

Hard drive analytics. Time series analysis. Detection of outliers using principal component analysisAuthored a patent:Title: STORAGE ANOMALY DETECTION US Patent Office 20180032385

Primary PhD student of the attribution research project between AOL and UCSC. Online display advertising campaign attribution estimation and research based on: the Potential Outcomes Causal Model and randomized experiments. Dynamic attribution modeling based on time series and survival analysis. Campaign performance analysis and attribution for large scale data (millions of users and thousands of campaigns).

Research in the following areas: online advertising attribution, statistical topic modeling, information retrieval, and recommender systems. These projects have been sponsored by multiple companies including AOL, AT&T, Google, Microsoft, Crowdscience (acquired by YuMe), and Serendio. Modeling energy consumption by smart buildings in collaboration with UC Berkeley energy group.

Mobile advertising targeting enhancement. Cold-start conversion prediction using Bayesian Hierarchical models. Analysis of A/B testing evaluation of algorithms.

Research in local search, and user review aggregation. Statistical topic modeling of user reviews based on business categories to improve business listing retrieval in local search.

Reconstruction and registration of Intravascular Ultrasound (IVUS) images with the University Hospital German Trias I Pujol, Barcelona Spain and Boston Scientific. Analysis and estimation of the cardiac motion from Tagged and CINE MRI with the Santa Creu i San Pau Hospital, Barcelona Spain.