Companies need working capital for inventory, product enhancements, labor and shipping. They also need an efficient way to manage customer orders, credit approvals, invoicing, payments and collections. The order-to-cash (O2C) cycle is a critical process that impacts numerous financial and customer activities. In addition to using traditional process improvement strategies such as Six Sigma or Lean to improve their O2C practices, companies can further improve operations and financial outcomes by applying artificial intelligence (AI) and real-time advanced analytics.
AI capabilities like machine learning can identify patterns and anomalies that indicate risks and exceptions. Natural language processing can also support the process by extracting unstructured data from multiple sources, such as invoices and email, and translating it into a more structured form that systems can analyze and act on. This data provides insights that enable better decisions, drive automation, and reduce exceptions.
The most significant O2C challenges that companies face are collections, disputes and cash flow. We believe organizations should consider three intelligent finance strategies that will improve their O2C cycle, speeding payments and ensuring consistency.
Collections optimization: Know whom to target and when
Optimizing collections by more accurately predicting whether and when customers will pay helps accelerate the O2C cycle and shorten the time that a company must fund inventory or receivables. Machine learning can identify the key drivers of delayed or unpaid invoices by learning payment patterns based on historical payment data and other factors (such as seasonal impacts, specific products, or salespeople associated with payment issues).
Using pattern recognition and predictive characteristics, accounts receivable departments can more effectively target accounts that pose the highest risk, prioritizing follow-up accordingly. For example, on past-due invoices, a customer behavior pattern that indicates payments typically are received after 60 days or more should be a higher priority than one that typically pays within 40 days. With more accurate predictors, collections teams can improve cash flow, reduce outstanding invoices and lower bad debt expenses.
Case study: A large health services provider wanted to accelerate receivables closures through faster payments and denials (by payers). Using machine learning predictive models, Cognizant analyzed historical invoices and claims to identify drivers of payments and denials, and to flag invoices likely to be paid and denied. Instead of continuing to use a first-in-first-out method, flagged invoices were prioritized for a second collections follow-up. The benefits were substantial:
Collections increased 13% and were received 33% faster.
Denial registrations increased 6% and occurred 50% faster.
Appeals of denied payments improved 10%.
Closure rates increased 6%.
Time to close decreased 17%.
Overall, by optimizing collections, the estimated annual benefit is $1.9 million in additional payments collected.
Dispute analytics: The heart of discontent
Understanding why customers dispute invoices can reduce negative effects such as slow receipts, complicated cash applications, necessary manual interventions and impacted customer satisfaction. Organizations can use machine learning techniques to analyze data from all relevant sources, identifying patterns and predictive factors of disputes not revealed through traditional reporting and descriptive statistics.
We often see causal relationships between propensity to dispute and parameters such as total invoice amount, geography, price structure, salesperson or specific products. Once drivers are identified, companies can take corrective action by either resolving the root causes of disputes or escalating higher risk cases to the collections department for faster prioritization.
Companies can also use process mining to understand invoicing issues. Process mining takes a comprehensive, fact-based approach by extracting event log data from a receivables management platform, ingesting it, and generating a process flow along with process variations, cycle time and number of cases. This approach helps identify inefficient dispute management processes as well as their root causes, providing a more granular view than traditional process mapping. The insights gained allow companies to focus on process improvements that can help avoid or minimize disputes and tied-up cash in disputed invoices.
Case study: A leading medical devices company lacked a mechanism to provide visualizations and insights into pricing disputes in two foreign markets. The company’s data resided in multiple systems and was often incomplete and difficult to use. We applied advanced analytics and AI to capture and analyze this data, pulling from all relevant sources. We created more than 20 insights that are displayed on a simple dashboard. Our analysis helped the company identify the cause and effect of pricing disputes, which it now uses to proactively avoid future disputes.
Cash flow forecasting: Intelligent data accelerates working capital management
In the business world, cash is king; finance organizations must continually develop cash flow forecasts to determine and manage working capital requirements. Traditional forecasting models use factors such as year-over-year trends, sequential rolling forecasts and seasonality, and these may be adequate when operating in a steady state. However, for more dynamic businesses — such as those experiencing high growth or launching new products — traditional forecasting techniques probably cannot achieve the desired degree of accuracy. Moreover, macroeconomic pressures may sometimes force virtually all companies to rethink how they forecast cash flow; for an example, we need look no farther than the COVID-19 pandemic.
A machine learning approach to forecasting enables companies to leverage detailed receivables and payables data to identify changing cash flow patterns. This “intelligent data” offers insights that support more accurate cash flow forecasts and provide the basis for more predictable and effective working capital management.
Case study: A large U.S.-based retailer needed visibility into cash requirements for its stores, which kept idle cash on site because the company lacked a cash retention policy. We analyzed three years of historical data and used machine learning to build a predictive model, accessible via mobile devices and the web, to forecast 30-day cash requirements for each store. Prediction accuracy surpassed 97%, which reduced working capital requirements and saved the company more than $4 million annually.
O2C platforms and systems contain extensive data assets, but traditional reporting and descriptive analytics only scratch the surface of insights that can be extracted to drive improvements. Advances in AI technologies help make this data more accessible to drive behavior predictions, pattern recognition and anomaly detection. By leveraging these strategies, CFOs and finance leaders can ensure that their O2C organization is more efficient, and reduce working capital requirements and costs.