Forecasting Goodwill Costs Before They Arise

The Challenge

Goodwill costs occupy a uniquely difficult position in the financial management of any large manufacturer. Unlike warranty claims, which follow defined rules and contractual obligations, goodwill payments are discretionary decisions made to maintain customer satisfaction when a vehicle is outside its standard warranty coverage. The manufacturer absorbs the cost as a gesture of goodwill. The amounts can be significant, the triggers are varied, and the volume is hard to predict.

For a global manufacturer of light and heavy trucks and buses, goodwill costs had become a meaningful budget challenge. The sheer number of variables involved — vehicle model, production year, damage code, warranty status, service contract coverage, extended warranty terms, and the presence of active recall campaigns — made accurate forecasting extremely difficult with conventional methods. Budgeting was largely reactive. When goodwill costs exceeded expectations, the organization had limited ability to anticipate the overrun or respond to it early enough to matter.

The business case for a more intelligent approach was clear: if goodwill costs could be forecast accurately at the level of individual cases and vehicle populations, the organization could manage its budget proactively, identify risk concentrations before they materialized, and make more consistent, defensible decisions about when and how much goodwill to extend.

The Approach

Building the Forecast Model

change2target developed a machine learning based forecast model capable of predicting goodwill costs at the level of individual goodwill cases and vehicle populations, with a two-year forecast horizon. The model was designed to handle the full complexity of variables that drive goodwill cost outcomes, including the integration of standard warranty periods, goodwill coverage windows, service contracts, and extended warranty terms, as well as granular damage code details that reflect the specific nature of each claim.

A proof of concept was developed first, establishing that the model could generate meaningful predictions with the available data. This was followed by a minimum viable product that demonstrated the model's performance at a level of maturity sufficient to inform real business decisions.

A particular challenge in goodwill forecasting is the effect of warranty campaigns — recalls and similar events that generate large volumes of claims in concentrated timeframes and can distort cost trajectories significantly. The model was designed to account for these events explicitly, ensuring that campaign-driven spikes were distinguished from underlying trends rather than absorbed into baseline forecasts.

Creating an Early Warning System

Beyond point-in-time forecasting, a structured early warning system was developed to alert the organization when goodwill costs for specific models, damage categories, or markets were tracking above expected levels. This gave budget owners the ability to act before an overrun became a problem rather than after it had already landed in the accounts.

The system also supported a more flexible approach to individual goodwill decisions. With clearer data on expected cost exposure by vehicle and damage type, decision makers could calibrate their responses more precisely, approving goodwill in cases where the commercial and customer relationship value justified it while applying tighter scrutiny where cost risk was elevated.

Establishing Goodwill Governance

Accurate forecasting is only valuable if the organizational structures around it are capable of acting on the insights it produces. As part of the engagement, change2target designed and established a goodwill governance structure that clarified budget responsibility across the organization, defined decision rights for goodwill approvals, and created a consistent framework for evaluating field events and their cost implications.

A concept was also developed for converting technical data from vehicle sensors and field information into structured inputs for the goodwill decision framework, creating a pathway toward more data-driven and consistent customer decisions as sensor data availability increases over time. Additionally, an index for measuring customer satisfaction in the context of goodwill decisions was designed, connecting the financial management of goodwill directly to its underlying purpose: maintaining customer loyalty.

Building the Data Foundation

Underpinning all of this was a purpose-built data model and structure capable of consolidating the diverse inputs the forecast model required. Bringing together claims data, vehicle data, warranty and contract data, and damage code information from multiple source systems into a coherent and reliable structure was a foundational prerequisite for everything else the project delivered.

The Results

The project delivered a working AI forecast model with a two-year planning horizon, an operational early warning system for budget overruns, and a governance structure that gave the organization clear ownership of goodwill decisions and their financial consequences for the first time.

Goodwill costs were reduced through more accurate identification of cases warranting approval and tighter management of exposure in high-risk segments. Budget accuracy improved significantly as forecasts replaced estimates. The flexibility and consistency of goodwill case handling increased, with decision makers operating from a defined framework rather than individual judgment alone. And by connecting goodwill management more directly to customer satisfaction outcomes, the project established a foundation for treating goodwill not simply as a cost to be minimized but as a customer relationship instrument to be managed intelligently.

Key Takeaway

Goodwill costs are one of the more overlooked financial risk areas in after sales management. They are discretionary by nature, which makes them feel unforecastable. In practice, they follow patterns — patterns that become visible when the right model is applied to the right data.

For manufacturing executives, the broader lesson is that the boundary between cost control and customer management is where some of the most valuable AI applications in after sales currently sit. Forecasting goodwill accurately does not just improve the budget. It gives the organization the confidence to make better decisions for customers, knowing the financial implications of those decisions in advance rather than discovering them at year end.