“Fast is fine, but accuracy is everything.” — Wyatt Earp
Creating an Accurate Forecasting Capability
SITUATION
In 2012, I was asked to take over the product management area and the additional responsibility of managing fee income. As I came in, I saw that forecasting had been embarrassingly inaccurate with monthly actuals coming in 30% below projections. Clearly, this was unacceptable to me and to the leadership of the firm.
To tackle this problem, I organized my team and led an in-depth review of all aspects of the forecasting cycle. That review revealed three principle areas needing improvement–accountability, data, and process. Let me now explain how I addressed each.
ACTION
Accountability
While fee income was my responsibility, the forecasting process was divided between product management and finance. The problem was that, although product would give input to the process, finance would increase the forecast based on influence from their leadership. Just as bad, the input the product team provided was not based on a full understanding of the fee income drivers. I requested and received approval for the product team to develop and own the forecast model going forward.
Data
I began by leading my team through a line-by-line review of the general ledger and all journals feeding it. The journals were mapped back to the core processing systems of the bank where we identified every transaction type that impacted fee income. A model was built where we accounted for all transactions, multiplied them by their unit prices and calculated the expected revenue to work our way back to the GL totals. While we found some isolated cases where transactions weren’t properly mapped, for the most part, transactions were flowing through to the GL the way they were supposed to.
After performing our mapping exercise, we identified 31 primary drivers of fee income that fed to 8 revenue lines on the GL (not surprisingly, we found that the current forecast model had missed several of those drivers). Using historical transaction data, we developed seasonal models of transaction volume and dependencies. This seasonality was particularly important for certain fee drivers whose behavior fluctuates greatly depending on day-of-week, and month-of-year. Accordingly, factors for day-ality, and month-ality were added to our forecast model.
Process
With solid data in hand, I created a “walk-forward” process that looked at the 8 GL revenue line items and the transaction drivers. The walk-forward started with the previous month’s actuals and made rate and volume adjustments based on the seasonal projections of driver activity over the forecast period. This formed the basis for forecasting the coming months.
The walk-forward was checked for accuracy then presented to the executive leadership team for approval. After approval it was input by finance.
RESULT
The changes I implemented addressed all deficiencies and resulted in an accurate forecast with clear accountabilities, solid integrity of both the data and the models, and a smooth, repeatable process for monthly and yearly forecasting. Forecast error was reduced to under 1%.
The methodology was so effective, we went further and also applied it to unit forecasting turning what had been an embarrassment into a point of pride. Because of my leadership, revenue and unit forecasts are now accurate, credible and effective tools to manage the retail bank.
Contact information:
Alpine Jennings
Senior Financial Services Executive
