“A goal without a plan is just a wish”- that’s how the saying goes. The Global Operations Analytics team at AB InBev started with a dream of transforming business operations through analytics, but it made sure that the wish had a plan and hence converted it into the goal.
The team started with five members heavily focused on using analytics in operations to go ahead and deliver $18 million in EBITDA savings. The head-start with a highly focused team enabled leadership to pump in funding to scale the solutions to all the zones where AB InBev does business. In 2020, the team spanned five more zones and scaled the team 3x.
Even during COVID-19 in 2020, the team was able to realise a value of $30 million in EBITDA. The same trend followed in 2021, where the team added $58 million in EBITDA and productised the flagship projects, thus enabling a zero-touch ML pipeline execution.
Data Science
The data science team consists of AI and ML experts ranging from 4-to 16 years of industry experience having a product mindset. The team supports all facets of the business in operations and dives into building an MVP in no time. This enables the buy-in from the business to develop a solution that is best suited for the business. All the processes in AB InBev have an analytics footprint from the Global Operations Analytics team, with the USP of the team being “Limitless boundaries in operations and deliver high value”.
Productization
Another arm of Global Operations Analytics ensures the project is taken to the stage of fruition. There is a lot of drive and rigour from the team that the consumption of every solution reaches every user, and the adoption rate stays on the upswing. In order to do that, the team engages in the development of an interface for the solution, which is supported by a data pipeline and driven by the algorithms.
| Project | Tower | Description |
| Payment Leakage | PTP | The payment leakage detector solution helps PTP teams to identify potential duplicate invoice postings. The solution is powered by AI and leverages the last three years of data to find similar patterns between existing and new invoice postings through the use of NLP and pattern matching. |
| Smart Collections | OTC | Improve OTC cash collections process by predicting delay in invoice payments and dispute reasons to enable customized collection actions. |
| ZBB Optimisation (Budget Optimisation) | FP&A | Analysis of budget trends in order to understand spend evolution market effectiveness and identify of opportunities to improve spend efficiency. |
| Credit Risk | OTC | Opportunity to enhance current credit allocation by leveraging transaction history & analyse drivers for noncompliance. |
| Cashflow Forecasting | ATR | Develop a robust analytical model that provides finance team(s) with an accurate cashflow forecast which helps with financial planning and maintain the solvency of the business. |
| Tax Optimisation | ATR | Preventing overpayments, underpayments and penalties charged to the company due to erroneous tax applications. |
Tax analytics
In the VAT domain, the structure of VAT keeps changing from country to country in the way it’s levied and the way it’s imposed. Every year, the ABI’s Accounts-to-report (ATR) teams process more than $20 billion worth of taxes and more through the goods and services that ABI provides. The complexity, legacy systems, and high volume of transactions make this process prone to error.
Sometimes, because of manual intervention by all the people working on the tax computation, there can be over claims and under claims. If there is a situation where an underclaim or overclaim happens, we can have a dispute, and it will delay the whole process itself. Once the government raises a dispute, they will put their points forward. So the money gets stuck during this period which can act as a working capital.
The team has created a structure to prevent overpayments, underpayments and penalties charged to the company due to erroneous tax applications. In addition, it has created an algorithm that typically detects where anomalies pop up or if incorrect VAT postings occur.
Architecture
There is an AP (Account payable) and AR (Account receivable) invoice. It goes into SAP, where the financial data resides. The automated invoice posting gets extracted and finally goes to the tax analytics engine, where anomaly detection occurs. These are then sent to the concerned team in a certain file. They would make those corrections. In SAP, the corrected VAT posting would go with the entire exercise being automated and made intelligent by using analytics and data science.
Cash flow forecasting
The objective here is to develop a robust analytical model that provides finance team(s) with an accurate cash flow forecast that helps with financial planning and maintains the business’s solvency.
If the Accounts Receivable is stuck, there is a lack of cash with the company, and it is deprived of working capital. It could have used that money that was coming for some project. AB InBev wanted to know the point of time at which the cash is going to be with us and how much.
Account payable plays a crucial role as well. Though it is controllable, businesses would like to see, based on the current financial ecosystem, when they are able to make the payment. This is where the forecasting engine has been built. It gives a flavour of what amount will be coming to the company and what amount will be payable in what amount in time.
Procure to Pay
The procure-to-pay process is one of the most business-critical technology stacks. It extends beyond your enterprise and involves suppliers. Therefore, it is crucial to adopt a seamless, efficient, and human-centric approach to designing and developing procurement solutions. When AB InBev is trying to procure goods, it encounters various challenges. Initially, it used to track one metric payment on time, but it was not giving the company the entire picture. So the team now looks at overpayment and early payment as well.
The payment leakage detector solution helps PTP teams identify duplicate invoice postings. The solution is powered by AI and leverages the last three years of data to find similar patterns between existing and new invoice postings through NLP and pattern matching.
Smart Collections System
AB InBev’s Order-to-Cash (OTC) teams are responsible for processing over $50 billion worth of invoices every year. However, even after having fully optimised solutions in place, the company has encountered many issues. The hurdles arise out of system dependencies and the business silos of various departments. These problems can create invoicing disputes that result in financial and strategic burdens on ABI business. Invoicing disputes can arise due to a variety of reasons such as inept order processing due to inefficiencies in the order management system, inconsistent payment terms, inaccurate and incomplete invoicing etc.
Some repercussions due to invoicing disputes include
- Nonadherence to scheduling agreements leads to undercharging issues.
- Delayed collections and recovery leading to impact on DSO (Days of Sales Outstanding)
- Negative impact on customer relationships, incorrect discounts
To solve such problems, the analytics team at ABI designed the Smart Collections System. This AI-powered platform assists the OTC teams in avoiding revenue leakages through solving the invoicing disputes in the OTC cycle and ensuring the timely collection of correct payments.
Three modules:
- Early Warning System
This uses heuristic rules and fuzzy logic to solve the invoicing issues during order generation.
- Short Paid Analytics
With the help of Optical Character Recognition and Natural Language Processing (NLP), assists collections teams in short payment recovery from the customers (Proof of Deliveries).
- Delay Prediction Model
This model uses cutting edge machine learning algorithms to predict the customers who would delay their payments. This significantly helps the collections team to recover the dues and impact the company’s working capital.
Credit Risk
The objective here is to enhance current credit allocation by leveraging transaction history and analysing drivers for noncompliance.
Here the data used are the invoice data (total invoices, delayed invoices, invoice amount), customer data (credit utilisation, payment term), Credit Bureau data and AR ageing data (AR ageing, balance, overdue).
It goes into two machine learning models.
- Model 1 – to determine credit risk category.
- Model 2 – to determine payment terms and credit limit.
Outcome
- Customers grouped into Credit Risk categories
- Dynamic Payment term determined for each customer
- Credit Limit determined for each customer
With analytics and AI taking such a central role in key business decisions,
AB In Bev as a leader is paving the way (and staying ahead) for integrating analytics and new-age technologies in business verticals. The use of cutting-edge tech in various aspects of the company’s finance systems has helped it optimise its performance and stay ahead in the game from its competitors.
