Forecasting Financial Metrics using Machine Learning and Data Science

You will learn:
• Application of machine learning and data science in Financial Planning & Analysis (FPA)
• Machine learning methods for forecasting
• Role of prediction pipelines in automated rolling forecasts
• Benefits of driver based forecasting using machine learning in FP&A
• Use of AI modules such as Scenario Modeling, Root Cause Engine and Anomaly Detection in financial forecasting

1. ©2023 Institute of Business Forecasting | All Rights Reserved | www.ibf.org
Forecasting Financial Metrics using Machine Learning and
Data Science
Vijay Dakshinamoorthy
Senior Manager (Data Science), Finance
Bell

2. ©2023 Institute of Business Forecasting | All Rights Reserved | www.ibf.org
Agenda
1.About BCE
2.Forecasting and FP&A
3.Data Science Methods
4.AI Modules
5.Evaluation
6.Insights for Executives

3. ©2023 Institute of Business Forecasting | All Rights Reserved | www.ibf.org
Bell Canada Enterprises (BCE)
Canada’s Largest telecommunications Company
3
About BCEAbout BCE About the Financial Analytics COEAbout the Financial Analytics COE
Delivers a wide range of innovative products and services to consumers, businesses and government customers.
•Mobile data and voice plans for 4G LTE, 5G
and 5G+ wireless networks
•FibeInternet / TV, Wireless Home Internet.
Residential/business voice, cloud-based, mobile
edge computing, Internet of Things (IoT)
•Bell Media operates media brands such as
CTV, RDS, Crave and iHeartradio. Leading
investor in content creation, sports /
entertainment and original TV and film
•One of Canada’s major retailers with more than
8000 retail points of distribution across Canada.
•Finance 2025 vision to better leverage data,
tools, systems and processes to enable a more
effective and efficient finance function
•59 team members across Data Intelligence
(Data Engineering, Data Science, Service
Management), Business Insights and
Visualization, Data Strategy & Architecture,
Program Office
•Standardize and streamline processes, drive
efficiencies and leverage Artificial Intelligence
(AI) and machine learning (ML) as key enablers
About BCE

4. ©2023 Institute of Business Forecasting | All Rights Reserved | www.ibf.org
Financial Forecasting at Bell
•Business Units prepare their plans internally during August and September
•The plans are submitted to corporate in October and reviewed by Senior Executives
•In early December the plans are approved and detailed budgets are developed by
Business Units and documented in the systems
Business Unit
Plans
Business Unit
Plans
•An update to the in-year budget, also known as YEE (Year End Expectancy)
•3 forecasts are submitted to corporate in the months of April, July and October
•Additional forecasts may be produced based on corporate’s needs
ForecastsForecasts
•SAP P-50 –SAP instance
•P58 –BI (Business Intelligence)
•BPC (Business Planning and Consolidation)
SystemsSystems
Forecasting and FP&A

5. ©2023 Institute of Business Forecasting | All Rights Reserved | www.ibf.org
Next Generation Financial Planning & Analysis (FP&A)
•12-24 months forward; routine processes automated
Rolling Forecasts
•Incorporate external drivers; collaborate with operations
Driver-based Modeling
•Annual event to continuous process; Scenario modeling
Planning
•Centralized data warehouse with integrated forecasting and
planning processes
Integrate processes,
systems and data
•Qualitative and quantitative measures beyond forecast accuracy
Measuring Impact
Forecasting and FP&A

6. ©2023 Institute of Business Forecasting | All Rights Reserved | www.ibf.org
Data Science Methods for Forecasting and Planning
Data Science Methods

7. ©2023 Institute of Business Forecasting | All Rights Reserved | www.ibf.org
Forecasting Financial Metrics with Data Science at Bell
Data Science Methods

8. ©2023 Institute of Business Forecasting | All Rights Reserved | www.ibf.org
Driver-based Models and Explainability
Data Science Methods
Driver-based modeling fits well with rate, volume forecasts and a P x Q calculation of final forecasts. Having
components of forecasts also yields well for explainability.

9. ©2023 Institute of Business Forecasting | All Rights Reserved | www.ibf.org
AI Modules to support FP&A
AI modules are re-usable components that can be applied across multiple business unit use cases. These
modules use fairly sophisticated algorithms to detect patterns in data.
•Scenario Planning
–Sensitivity analysis
–Forward Scenario modeling
–Reverse Scenario modeling or Optimization
•Root Cause Engine
•Anomaly Detection
•Commentary engine
•Virtual reporting agent
AI Modules

10. ©2023 Institute of Business Forecasting | All Rights Reserved | www.ibf.org
AI Modules at Bell
AI Modules

11. ©2023 Institute of Business Forecasting | All Rights Reserved | www.ibf.org
Evaluating Forecasts
Evaluation

12. ©2023 Institute of Business Forecasting | All Rights Reserved | www.ibf.org
So What for Finance Executives
Insights for Executives

13. ©2023 Institute of Business Forecasting | All Rights Reserved | www.ibf.org
Where to Start?
Insights for Executives

14. ©2023 Institute of Business Forecasting | All Rights Reserved | www.ibf.org
Tips for data-driven FP&A
1.Start small and iterate
2.Ensure sufficient historical time series and account for COVID periods
3.Balance accuracy with other benefits (timeliness, automation, granularity)
4.Incorporate external drivers (Driver based modeling and Scenario modeling)
5.Aim for explainabilityand transparency
6.Adopt a product management and agile data science mindset
7.Measure impact and refine
Insights for Executives

15. ©2023 Institute of Business Forecasting | All Rights Reserved | www.ibf.org
THANK YOU!
[email protected]