Our algorithms have beaten official forecasts and accrual adjustments consistently

Our forecasting algorithm beat official accrual adjustments by 45 percent

OnCorps built an AI-powered system that analyzed five years of historic expense, invoice, and accrual data. Using this data, along with AUM data, our data science team implemented a machine learning model that learns to improve forecasting accuracy by fund and expense type.

More accurate TER accrual forecasts

Outcomes

Backtesting. All our new forecasting systems are thoroughly backtested before implementation. For one asset manager, we tested the algorithm by challenging it to predict expenses in mid year for a prior year where actual expenses were known. We provided the SARIMAX algorithm with only the accrual adjustment recommendation for mid year and all the months history prior to mid year. We then ran 13,000 samples using the actual data in a Monte Carlo model to improve the certainty of the forecast.
‍
Accuracy Rose 45%. On average across all cost categories, we found the SARIMAX algorithm to be about 45 percent more accurate than the asset manager’s accrual adjustments at mid year. Each cost category save Support showed a dramatic accuracy improvement from the firm’s official mid year accrual adjustment. The largest expense, Custody, showed a 32 percent improvement. Distribution expenses indicated a 46 percent improvement.

Case study

Increasing accuracy of fund expense forecasts with Machine Learning models

The Request

The budgeting team at a large international asset management firm was dissatisfied with the accuracy of their fund expense forecasting.

Excel models retained the same simplistic calculations even as more data became available to refine them.

Updates from expense owners in relevant departments (e.g., Legal) were not collected and incorporated into the models.

Sizeable accrual adjustments were required at year-end due to under-forecasting and under-accruals.

The Solution

OnCorps built an AI-powered system that analyzed five years of historic expense, invoice, and accrual data. Using this data, along with AUM data, our data science team implemented a machine learning model that learns to improve forecasting accuracy by fund and expense type.

the results

Our trained models were run against both the legacy forecast and actuals for the prior year. Six months out from year-end, the OnCorps forecast beat the legacy forecast for 25 out of 30 funds. Three months out, OnCorps was more accurate for 29 of 30 funds. Improved AI forecasting will allow the firm to change controllable expenses sooner and improve profitability.

83%

of funds for which OnCorps' AI beat the service provider's forecast

88%

of cost categories for which OnCorps' AI beat the legacy forecast

45%

improved accuracy driven by our first run of algorithms

13K

models trained to improve forecasting accuracy

"OnCorps' algorithms analyzed over 13,000 models in order to select an optimal model for each fund and expense type combination."

Learn more

If you'd like to learn more about OnCorps' Forecasting solutions, please complete our contact form or email us at inquiries@oncorps.ai

See a demo

Thank you! Your submission has been received!

Oops! Something went wrong while submitting the form.

Thank you! Your submission has been received!

Oops! Something went wrong while submitting the form.

Privacy Notice Careers

OnCorps, Inc.
116 Huntington Avenue
Boston, Massachusetts 02116

© 2024 OnCorps, Inc. All rights reserved.