Predicting a Global Fintech Company’s deal outcomes with 80% accuracy

Challenges

C3 AI’s customer is at the epicenter of the fast-transforming fintech industry. However, its business needs are typical: its leaders must identify and react to hot spots in the sales pipeline. They must plan quotas and incentives, identify where to accelerate and decelerate hiring, and plan teams and territories. But the company’s rapid sales team growth has strained its revenue operations and ability to forecast and manage effectively. As the sales team has expanded, its organizational structure has grown complex, making it even harder to plan.

Because customers do not make upfront commitments, sellers and revenue operations need help to predict steady-state recurring revenue for new accounts. As a result, sellers tend to focus on more visible measures of success, such as new logos and signups, at the expense of more important measures like driving adoption and helping new accounts ramp usage quickly. As a result, the company suffers from sub-optimal revenue attainment.

Additionally, with sales highly susceptible to external macroeconomic factors, foreign exchange rates, and multiple other external factors, ranging from geopolitics to inflation, it is difficult to forecast sales and manage the business for the future as this company operates in over 40 countries worldwide.

C3 AI configured C3 AI CRM to help the customer address these challenges.

Approach

Over eight weeks, C3 AI configured the C3 AI CRM application to build four predictive machine learning models to forecast quarterly sales and each deal’s likelihood of being won. C3 AI configured the C3 AI CRM user interface (five screens) to enable an intuitive, flexible, efficient forecasting process suited to the customer’s needs.

The C3 AI team began by ingesting, cleansing, and unifying several years of historical data from the company’s CRM system, which amounted to ~85,000 opportunities.

The C3 AI team then augmented the CRM data with 113 exogenous data sources, including economic indicators, foreign exchange rates, stock market indices, and industry indicators, to create a unified federated data image. This comprehensive and clean data image allowed us to generate accurate AI insights. In addition, it formed the foundation for addressing additional adjacent AI use cases, such as consumption forecasting and churn risk detection.

Using the historical CRM data and the exogenous data, C3 AI trained core machine learnings models to generate the following AI predictions:
• Quarterly sales forecasts
• Which opportunities will be won each quarter
• When each opportunity will be closed (won or lost)

Each opportunity-level prediction comes with an AI evidence package that indicates how to improve the probability of each opportunity, making AI insights complete, comprehensive, and actionable.

Finally, the C3 AI team configured five user interface screens to deliver application insights. The screens enable users to forecast, manage their sales pipeline and accounts, and take action against AI insights, ultimately operationalizing a more efficient, effective sales motion guided by AI.

Project Objectives

• Forecast sales (bookings) and the likelihood of sales opportunities to be won across all organizational hierarchies with advanced machine learning models.
• Enrich core CRM data with additional ~100 external data sources provided by C3 AI.
• Configure C3 AI CRM’s user interface to expose accurate sales forecasts and enable sales pipeline management globally.

About the Company (2022)

• 40+ countries of operation
• More than 1 million customers
• $5+ billion annual revenue
• $300+ billion annual transaction volume

Project Highlights

• 8 weeks from kickoff to pre-production application completion
• 8 quarters of data integrated
• 6x forecast error reduction (from 35% to 6%)
• ~80% accurate opportunity outcome predictions on day 1 of quarter
• 94% accurate sales predictions on day 1 of quarter
• 113 exogenous data sources integrated into predictive ML models
• 2,000+ machine-learning model features created and used
• ~85,000 opportunities analyzed and scored
• Up to $27M in potential annual economic benefit is estimated due to increased win rates and sales pipeline

Solution Architecture