Building an Analytics Driven Budget

The Quintessential Technology Source for Corporate Financial Professionals

Building an Analytics Driven Budget

Patrick Harms, AVP Data Analytics at Credit Union of America

Patrick Harms, AVP Data Analytics at Credit Union of America

Given the ever-increasing complexity of the business, just the word “budget” is enough to make any CFO’s head spin. While forecasting for an accurate budget is never going to get less complex, there are more options than ever for making sense of all the potential factors involved. Those working at financial institutions, or in any industry for that matter, know that this year has been more challenging and that those challenges will also be present going forward. This has only added complexity to budgeting, but by using a combination of data analytics and budgeting techniques, these challenges can be overcome.

While most budgets are built under the watchful eye of the CFO, all managers should be involved as they have the best knowledge about what drives their income and expenditures. This is not always the case, though. Managers may have a high-level idea of what effect the environment will have on their budget but may not know what which factors produce an accurate forecast or what coefficients to use. At this point, a data analyst should become involved to determine what outside factors correlate to income and expenditures. These drivers will have different effects on the output and should be run through a regression model and then retested for accuracy.

Depending on what is being forecasted, there may be many different drivers and therefore, many different coefficients produced by such a model. To make accurate and reliable forecasts, a broad scope of factors may need to be input leading to extremely complex models. At this point, two things must be taken into consideration—overfitting and reasonability of inputs. Overfitting may seem simple to avoid, however many do continue to introduce inputs to models until the model is “right” all the time. As it turns out, this means that the model will predict its training data correctly and nothing else. Only if the future strictly follows the past would such a model be accurate or even useful. Inputs to any model must be reasonable and relevant. In heavily regulated industries, this is not only a best practice but also mandated by regulators. Additionally, staff performing the modelling will be thankful not to have to forecast or at least enter additional factors.

Even though statistical learning is becoming more common, not every company may have access to data analysts with a statistical learning experience. Luckily, there are now several different companies that make statistical learning software that does not require coding experience or more than a basic understanding of modelling. While such drag and drop applications may not allow for as much fine-tuning, they can still provide an accurate forecast for budgeting purposes. They will always be more precise than straight-line estimates based on a gut-feeling percentage.

Technology has only made it easier to predict income and expenditures for the coming year. However, technology is only one piece of the puzzle. Uncertainty has presented itself in nearly every aspect of the business. To help deal with this, it may be possible to budget certain items on a rolling basis so as to get only the most up to date and relevant inputs. Depending on the technology available, there may even be an opportunity for automating aspects of the budgeting process using information from the company’s operational system. In all cases, CFOs should use all resources at their disposal to create a reliable budget.

Weekly Brief

Top 10 CPM Solution Companies - 2020
Top 10 CPM Consulting/Service Companies - 2020