Unlocking the Value in FP&A Data Science & Analytics

Unlocking the Value in FP&A Data Science & Analytics

By Emmanuel Jibodu, FP&A Manager​ at Stericycle Inc.

This is the second part of a three-part series that focuses on the business value that data science & analytics can provide to enterprises.

First a quick recap on my earlier column. Enterprises, whether publicly traded or privately owned, in their drive to grow and unlock value must make critical decisions about the following:

  1. balance sheet optimization 
  2. pursue reasonably priced acquisitions 
  3. re-evaluate their portfolio businesses and engage in spinoffs, if needed
  4. grow their product/service portfolios. 

Rather than passivity, enterprises have deployed data science and analytics as the linchpin of their growth initiatives. In this article, we will cover three areas of analytics that Financial Planning and Analysis professionals (FP&A) perform to drive business value. You can find the first part of this series here.

Diagnostic and descriptive analytics, within an FP&A context, are classified as backward-looking finance activities. Prescriptive and Predictive analytics are classified as forward-looking finance activities.
It is paramount than an organization has a robust performance measurement process to derive business value from its analytics activities. An enterprise’s strategy, key performance indicators (KPIs), and internal operational metrics and external indicators underpin all analytics initiatives.

Variance Analysis

This represents a hybrid activity that is undertaken by FP&A. It can be categorized under both Diagnostic and Descriptive analytics. It allows executives to ascertain the performance of a region, business unit, and enterprise by juxtaposing current performance with forecasted, or prior-year levels.

Within enterprises that embrace advanced performance measurement processes, the assessment of enterprise strategy formulated by the C-suite is paramount. And one method of assessment is: monitoring the variance between actual versus forecast/plan key performance indicators (KPIs). KPIs for a strategy centred on capturing market share via innovation will be different from one that's focused on capturing market share by attracting clients that are price driven.

Variance analysis, for example, would monitor metrics such as: sales data (e.g., average size per deal of sales opportunities, sales force productivity), customer data (e.g., churn rate), employee data (e.g., headcount, attrition rates) and tie them to financial outcomes (e.g., Third quarter revenues). For example, in a software company, for a given quarter, if the forecasted levels of sales metrics—such as conversion rates for sales opportunities, the average price of deals and number of expected deals differs widely from actuals, the result is an unfavourable variance for Sales From New Customers. However, sales from existing customers, maintenance fees, consulting and training could preclude an unfavorable variance in total revenues.

Further, publicly traded companies that disclose key performance metrics to the investment community can forge stronger relationships with investors. For example, Badger Daylighting (TSX: BAD), a Canadian hydrovac producer and operator that provides low-risk excavation of underground pipelines and infrastructure reports publicly discloses its revenue per truck per month—an important key performance indicator.

Badger Daylighting has been able to establish itself as the market leader in North America. Other examples include publicly traded Canadian telecom companies that disclose operational metrics such as customer churn rates and average revenue per user (ARPU).

Rolling Forecasts

Rolling forecasts are de rigueur for FP&A practitioners, with many deeming annual budgets as outdated once completed; and it represents the latest outlook for an enterprise. Rolling forecasts lie within the Predictive Analytics zone of the analytics framework described above.

Enterprises leverage forecasting for a litany of reasons, which include: providing guidance to the investment community on how effective its strategy is being executed; peaking around the corner 18 months - 2 years ahead (by quarter/month) at performance pitfalls or opportunities; manage the timing and roll-out of operational activities such as marketing promotions and major product launches; evaluate enterprise performance in comparison to a benchmark or peer group; identify early warnings of increased borrowing needs.

Rolling forecasts are updated monthly or quarterly. Best-in-class enterprises build rolling forecasts based on the key drivers that are integral to the performance of their enterprise—this is called a driver-based rolling forecast. These drivers are internal and external metrics. Internal metrics include, but are not limited to the following examples: customer metrics (revenue forecast), product demand (revenue forecast), employee productivity metrics (revenue and expense forecast), employee compensation data (expense forecast), manufacturing capacity (expense forecast), marketing efforts (revenue and expense forecasts), etc.

Further, outputs from revenue and expense forecasts are also used for cash flow forecasts. Examples of external metrics include economic variables such as price level indices, unemployment data, auto sales, and housing statistics.

Driver-based rolling forecasts leverage predictive analytics tools, mathematical algorithms, at the same time deploy internal/external metrics to build income statement, balance sheet and cash flow forecasts. Software tools now exist that help enterprises determine the external metrics that relate to their internal results.

Further, machine learning solutions have made it possible to analyze multi-structured data types in real-time. BI data coupled with increased computing power has enabled us to build robust models that can provide predictive power for a variety of scenarios. Of course, the convergence of these factors has represented a boon for FP&A teams in the development of driver-based rolling forecasts.

For enterprises that have not implemented a robust driver-based rolling forecast—rapid prototyping can be used to get things going. First, enterprises need to find out if they have access to the right data, that will yield accurate outputs, even if machine learning algorithms are robust. Secondly, we focus on a quick win—build a driver-based algorithm and model to forecast revenue for a product line or service line with high visibility across the enterprise to gain support early on. Third, we backtest the results of the predictive model to corroborate its efficacy.

Further, machine learning tools, specifically linear regression models, have made it possible for enterprises to revisit and refine the cause-and-effect relationships between operational metrics and financial results in an automated manner. Therefore, we can constantly reiterate predictive models to make them more accurate and responsive to evolving business needs.

Customer Value Management

Customer profitability analysis enables enterprises to ascertain which of their customer segments to prioritize for acquisition and retention. Customer profitability entails the profits an enterprise garners from its customers, not only after accounting for direct costs/variable costs, but also after accounting for costs to serve—such as selling, distribution, advertising, marketing, logistics, warehousing, and after-sales service. The result of customer profitability is the categorization of customer segments into quartiles or deciles—each with differing profitability; and the evaluation of how each customer segment contributes to an enterprise's bottom line.

Enterprises can monitor different customers by monitoring profit and loss statements for each customer segment in real-time due to advancements in business intelligence and transactional systems. The end goal of customer profitability analysis is to calculate the customer lifetime value (CLV) for a customer/customer segment.

As a result, enterprises can prioritize which customer segments to target for customer acquisition and retention initiatives—this lies within the Prescriptive Analytics zone of the core business analytics framework. Based on the current profitability and future profitability of customer/customer segment discounted to the present using the enterprise's cost of capital, we can calculate the value of a customer/customer segment to an enterprise.

Formerly, product profitability or service line profitability was the focus of FP&A teams, however, in recent years, a palpable sea change has occurred, and the focus is now: customer profitability. Instead of the focus being on the profit margins of products (e.g., why has the profitability of product A reduced by 50 basis points?); FP&A teams are now charged with optimizing the return on investment (ROI) from marketing activities, product development and extension initiatives. Of course, this is achieved by targeting the customer segment with the highest CLV or transmuting relatively lower CLV customers into higher CLV customers. Consequently, an enterprise must find a happy medium between the profitability of a customer/customer segment and the costs incurred to serve them.

Further, machine learning tools can aid enterprises in managing customer retention initiatives. ML models can relate to enterprises the churn propensity of a customer, for example, particularly a relatively high CLV customer, by leveraging multi-structured and unstructured data on that customer. ML models can also relate to enterprises cross-sell or up-sell opportunities—that serve to transmute customers/customer segments with low CLVs to higher CLVS.