CUSTOMER STORIES
LTM Improves Employee Onboarding Probability with Snowflake
Snowflake helps the LTM talent solutions team to predict employee onboarding with 80% accuracy
80%Model Accuracy
700+Users adopted across different business units
Industry
TechnologyLocation
IndiaOverview
The hiring landscape has become increasingly complex, marked by fierce competition for top tech talent, shifting employee expectations for seamless digital interactions, and high attrition rates within the first 90 days of joining. For large enterprises, these challenges are compounded by siloed HR systems, manual processes, and delayed insights, often resulting in reactive decision-making post-offer dropout rates of up to 30%.
LTM, a global technology consulting and digital solutions company, recognized that traditional hiring models were no longer sufficient. To deliver hiring strategies at scale and improve candidate retention, LTM required a unified and robust data platform that could bridge business goals with advanced analytics and AI. LTM selected Snowflake for its unified data cloud platform that can support scalable hiring strategies through robust AI and data capabilities.
Snowflake’s AI/ML model predicts the probability of candidate onboarding by analyzing historical data from successful and unsuccessful joiners. The model delivers predictions 25-30 days prior to the date of joining, enabling early identification of potential dropouts and proactive strategies to enhance retention. It runs at an 80% accuracy level.
Story highlights
- Strategic AI/ML Deployment: The predictive ML algorithm is trained on 9 –12 months of historical data with 12 predictive features. Each prediction includes top 3 influencing factors enabling talent acquisition teams to take precise actions.
- RAG-Based Prioritization: Candidates are segmented into red, amber and green based on the onboarding probability. Enables targeted follow-ups and strategic interventions for low-probability candidates.
- Scalable Rollout: Adopted by 700+ users across talent acquisition, business, workforce management, and Operations teams. Used across multiple service lines and delivery units.
From legacy constraints to data agility
Prior to Snowflake, LTM’s HR analytics relied on legacy, on-premises systems, plagued by data silos, high maintenance costs, and poor scalability, and slow processing speeds, often delaying candidate predictions by weeks and inflating expenses by 2-3x during peak hiring seasons.
By migrating to Snowflake’s single, unified platform, LTM unlocked real-time analytics, elastic scalability, and significantly reduced total cost of ownership (TCO). The migration took place in phases: lift-and-shift of critical workloads using Snowflake’s Snowconvert tool for minimal downtime; followed by the modernization of ETL pipelines with zero-ETL connectors to sources such as SAP, SuccessFactors, and ATS platforms; and third, deploying ML models via Snowpark for end-to-end operations.
This approach delivered 10x faster query performance, 70% TCO reductions, and automatic scaling aligned with hiring volumes.
The hiring landscape is evolving rapidly, and technology is no longer just an enabler—it is a strategic partner in shaping better candidate experiences and driving measurable outcomes. By leveraging Snowflake as our data platform, we’ve achieved seamless alignment between technology and business, enabling advanced models like joiner prediction with exceptional accuracy. This milestone is a testament to how innovation and technology converge to redefine and transform recruitment, while allowing us to customize candidate engagement strategies and prevent last-minute disruptions, and most importantly, avoid customer disappointments.”
Rajeev Menon
Executive Vice President, Human Resources
Building a resilient data foundation
To support AI-driven hiring at enterprise scale, LTM implemented a robust and modular data architecture for the successful AI/ML solution, and this was achieved through:
- Modular data modeling: Instead of monolithic code—large, inflexible scripts that ran sequentially and choked legacy systems during high-volume HR data processing—Snowflake enabled logically separated intermediate steps through its architecture of stages, views, and Snowpark user-defined functions (UDFs). This modular approach utilized virtual warehouses for automatic parallel processing across nodes, distributing workloads like feature extraction from candidate profiles or aggregation of joiner/dropout histories.
- Automated data pipelines: Dynamic Tables and Automated Tasks streamlined the transformation of raw data into consumption-ready datasets, reducing build times and manual orchestration.
- Governance and traceability: Mandatory object and query tagging ensured complete visibility into costs and data lineage, which was critical for scaling enterprise AI models. This is essential to deliver full transparency on costs and data flows, preventing budget overruns and ensuring compliance with regulations like India's Digital Personal Data Protection Act or the General Data Protection Regulation. Tagging enables teams to assign labels, such as "HR-Recruitment" or "Project-TalentAI" to tables, queries, and pipelines, enabling precise cost allocation across HR initiatives. It also creates an audit trail showcasing how data moves from raw applicant logs to ML predictions, vital for regulations like India's DPDP Act or GDPR. This traceability also helped LTM prove model fairness, secure sensitive candidate data, and scale AI safely across more than 80,000 users.
75-80%
Enhanced prediction accuracy from 75% to >80% by increasing training iterations and expanding training data.
AI/ML innovation with Snowflake Cortex
Snowflake’s fully-managed AI/ML capabilities, such as Cortex, played a pivotal role in transforming LTM’s workforce analytics. LTM leveraged these integrated capabilities to transform talent functions. Cortex provides serverless access to large language models such as Snowflake Arctic, Llama 3, and Mistral along with ML functions like forecasting, anomaly detection, and sentiment analysis–eliminating GPU management, model hosting, or data movement. By integrating Cortex via BlueVerse, LTM enabled natural language queries on employee data, automated resume screening, personalized retention insights, and proactive governance and token budgeting for AI workloads. This approach reduced model development time significantly and allowed AI applications to be built 5x faster.
- Predictive modeling: A key outcome of this transformation was the Joining Probability Predictor, an AI/ML model, designed to shift hiring from reactive to proactive.
- Capability: Predicts whether candidates are likely to join or drop out based on historical data.
- Performance: Identifies potential dropouts 25–30 days prior to the date of joining with 75–77% accuracy.25–30 days pre-DOJ provides critical lead time for targeted actions—personalized offers, counter-offer negotiations, relocation support, or culture-fit sessions. Furthermore, each post-offer dropout typically costs 1–3x the candidate’s salary in re-recruitment, lost productivity, and administrative work.
- Efficiency: Automated stored procedures handled feature engineering and algorithm training within the Snowflake environment using SQL and Python notebooks.
Operationalizing insights across the enterprise
The transition to Snowflake has empowered LTM to deliver proactive strategies at scale:
Model registry and versioning: Models are registered within Snowflake to ensure better tracking and deployment of model versions.
Automated insights delivery: Automated tasks run models daily, with results seamlessly integrated into Power BI dashboards.
- Enterprise-wide adoption: Over 300 users across Talent Acquisition, Workforce Management, and Operations leverage these insights for early intervention to enhance candidate retention.
