Financial planning systems are geared to secure two main financial goals of a company: gaining profitability and securing financial stability. Forecasts are a crucial addition to any financial planning system to steer corporate performance, i.e., identify measures to secure the financial goals early on. While financial planning allocates appropriate resources to corporate operations, financial forecasting predicts how key performance indicators (KPIs) will develop as objectively as possible. Importantly, these forecasts are independent of plans, thus enabling management to make informed decisions and adapt actions to reach corporate goals.

 

The case of the multinational crystal jewelry designer Swarovski illustrates the challenges in the evolution of the planning process from traditional budgets to analytics-induced driver-based planning and forecasting for better decision making (see also Renita Wolf’s 2015 Strategic Finance article Broken Budgets?). These driver-based planning and forecasting systems, when distilled to their smallest unit, consist of a set of value drivers. Value drivers represent the key influencing factors of a business model. In a driver model, value drivers are mathematically combined as interdependent variables, meaning they systematically relate the performance-driving activities of companies to the financial and strategic performance measures they influence.

 

Driver models help analyze organizational developments as they combine cause-and-effect relationships and time effects. These cause-and-effect relationships can be visualized in a driver tree chart to further enhance the understanding and communication of a business model. By translating the business model into value drivers, driver models can be used to simulate different scenarios for planning and forecasting. This characteristic is particularly vital for companies operating within a VUCA (volatile, uncertain, complex, and ambiguous) environment. Driver models shed light on the effects of changes in individual value drivers on top KPIs (e.g., return on capital employed) and enhance the understanding of the business model with its specific steering levers.

 

Value drivers can be financial and nonfinancial. This duality is also reflected in Principle 9 of the IMA® (Institute of Management Accountants) report Key Principles of Effective Financial Planning and Analysis. For example, the DuPont (return on investment or ROI) analysis tree is a sole financial driver model that analytically decomposes the ROI into multiple financial accounting metrics (e.g., sales, inventory, or operating expenses). However, financial driver models don’t consider how revenue is generated. Revenue generation is highly dependent on the business model of the individual company and requires the inclusion of nonfinancial value drivers (e.g., store types, experience of the sales force, price elasticities, or social media sentiment). As business models and value drivers differ considerably across industries, strategies, and setups, there’s no all-encompassing driver model off the shelf. Even within a company—from the group level to business units, functions, or markets—many individual driver models can be used autonomously or combined toward the top KPIs. For example, an operating expense driver tree can be broken down into several individual trees (e.g., per business unit). The resulting cascade of driver trees reduces individual driver tree complexity, making them easier to understand, communicate, and discuss.

 

Given the company-specific nature of driver models and their division into sub-models, value driver selection is a complex task and turns out to be a major challenge for companies in the implementation of driver-based planning and forecasting.

 

The Era of Driver Models

 

Swarovski, a global manufacturer and retailer of high-end cut crystal stones, offers a representative case study on the evolution of financial planning and forecasting within established companies. The focus of this article is the crystal business of Swarovski, operating in more than 150 countries and including more than 2,400 retail stores at prime locations. With a history spanning more than 125 years, the company has navigated various stages that are typical in the development of financial planning systems and practices, influenced by both internal and external factors—a phenomenon well-framed in the contingency theory of management accounting.

 

Historically, Swarovski used an annual, locally driven, bottom-up planning and forecasting approach rooted in a decentralized organizational structure for decades. However, this method proved to be time-consuming due to the granularity of the planning depth and silo thinking within the organization. In the early 2000s, the need for more efficient financial control became evident in the face of rising competition from Asia and the shift toward a direct distribution model, which required increased (store-level) planning granularity and capital expenditure planning.

 

The 2008 financial crisis further strained Swarovski’s financial planning, prompting the adoption of shorter quarterly forecasting cycles to better navigate economic uncertainties, as also suggested by research. Yet the postcrisis period brought renewed competition, pressuring the company to streamline its processes and reduce planning granularity combined with a rolling forecast approach. Despite these efforts, the persistence of siloed thinking and management’s demand for detailed information led to an even more time-consuming planning process, hindering agile decision making.

 

During the COVID-19 pandemic, Swarovski had to shift to a weekly, top-down predictive forecasting model out of necessity to deal with the extreme uncertainty. This solution was significantly faster and more accurate than the previous manual, bottom-up forecasting approaches. This enabled local financial managers to focus more on the operative steering of the business amidst the crisis, as there were too many uncertain variables to produce meaningful bottom-up forecasts anyway. Ultimately, this environment facilitated the decision to adopt a more centralized planning approach and created a certain level of acceptance for predictive analytics in financial planning and forecasting.

 

Swarovski’s Driver-Based Approach

 

In the aftermath of the COVID-19 pandemic’s peak turbulence, Swarovski pinpointed driver-based planning as a promising approach to address the challenges of reinstating the high level of planning granularity traditionally demanded by management with more frequent (monthly) forecasting cycles due to the increased uncertainty, while keeping the planning and forecasting processes efficient because of market pressures. As the top-down predictive planning and forecasting approach demonstrated its efficacy during the pandemic, enabling proactive responses to environmental changes, Swarovski decided to incorporate predictive analytics and implement a predictive driver-based planning and forecasting solution across the entire group. To generate quick wins and gain a deeper understanding of the potential of driver models, Swarovski decided to focus resources and commence its driver-based planning journey with one of its top KPIs—the net sales of its business-to-consumer crystal business.

 

Swarovski’s net-sales driver model. As a retail-driven organization, Swarovski decided to implement the net sales driver model as one core element of its future financial planning system. The model employs a hybrid planning and forecasting approach by combining a top-down automated forecast of the baseline net sales trend using a predictive model and a bottom-up adjustment of the net sales value drivers to account for effects not reflected in historical data patterns (e.g., the planned refurbishment of a flagship store). The resulting net sales figures can then be broken down to individual stores using an allocation key logic.

 

Swarovski’s predictive model. The data-driven baseline net sales trend is predicted using a time series model that learns from 10 years of historical net sales data. Multiple time series models (e.g., Prophet, autoregressive integrated moving average, and exponential smoothing) compete against each other, and the best-performing single model or ensemble of multiple models is selected for the final prediction. The prediction doesn’t account for changes in the other net sales drivers and thus is merely the starting point of the planning and forecasting process. However, Swarovski is also experimenting with causal models that take into account the effect of macroeconomic factors such as gross domestic product (GDP) growth, consumer spending, or inflation. The baseline prediction provides management with a rapid initial forecast and leaves more time for discussions on the developments of the business model and its drivers. Predictive analytics enables automated and evidence-based forecasts at the push of a button, and its merits and pitfalls for financial planning and forecasting have been widely discussed.

 

Swarovski’s value drivers. Besides the baseline net sales trend, three aggregated value drivers best explain Swarovski’s net sales: the portfolio mix, marketing, and distribution (see Figure 1). For the portfolio mix, price elasticities, average selling prices, and units sold at the market level were selected as main value drivers to estimate how Swarovski’s product portfolio and pricing strategy influence net sales. The marketing effect on net sales is driven by the marketing expenses and the expected marketing ROI. The distribution driver deduces the net sales effect from the number of active retail stores and, thus, is mainly driven by the opening, closing, and refurbishment of stores. For example, the refurbishment of a store usually decreases sales during construction. But after reopening, sales increase due to elevated customer interest. Swarovski uses the historic patterns of comparable stores to model these sales effects.

 

12Weiserfigure1Final

 

Augmented intelligence. The predictive driver-based model combines a top-down approach using a predictive model with bottom-up adjustments using value drivers. This form of an augmented intelligence system employs value drivers to augment the net sales trend prediction, following a traceable driver model logic. By separating the planning and forecasting process into a predictive model and manual adjustments, the strengths of both predictive analytics and the business acumen of the management accountant can be combined. The driver tree helps to structure the judgmental adjustments of the net sales trend prediction while also enhancing transparency. This is achieved by making business assumptions explicit through the driver model and focusing on the lowest level of analysis (the value driver), rather than directly adjusting net sales.

 

Technology. Swarovski has implemented its driver-based planning and forecasting solution by integrating various data and analytics platforms. A business warehouse solution aggregates the granular primary data from an enterprise resource planning system. This data is then transferred to a data and analytics cloud platform that runs the baseline net sales trend predictive models. In the final step, the results are loaded into a cloud planning platform that supports driver-based data models and interactive user interfaces to adjust the predictions and simulate different scenarios using the value drivers.

 

Since the inception of the net sales driver model, Swarovski has rolled out driver-based planning and forecasting to several markets and to further top KPIs. The application of driver-based planning and forecasting has numerous benefits in terms of efficiency and accuracy for the company. For instance, manual labor and error rates could be significantly reduced through an automated data-integration process. Using the net sales driver model has more than doubled the accuracy of its net sales forecasts compared to the previous traditional bottom-up approach. Additionally, transparency about the buildup of plan figures was enhanced while shortening the planning and forecasting cycles to enable more agile financial steering. Overall, it took one year to implement the first fully driver-based planning cycle.

 

Implementation of Driver Models

 

Several interviews within Swarovski on the implementation project corroborated that organizations should consider three essential levers for a successful transformation toward driver-based planning and forecasting: people, process, and technology. Across these three levers, management accountants take the role of orchestrators driving driver-based planning initiatives. The orchestrators need to acknowledge the change management characteristics of financial planning system transformations, where soft factors mostly outweigh hard factors (see Figure 2).

 

12Weiserfigure2Final 

 

People. The key to success in designing a driver model is acceptance and buy-in from all involved stakeholders, regardless of the driver model’s accuracy or deliberate construction. If stakeholders don’t accept the selected drivers, even the most sophisticated model is worthless. Nearly everyone with business acumen can select value drivers, as most are already known and common sense within organizations. Nevertheless, selected drivers won’t be accepted without proper change management efforts due to a lack of trust in the driver-based planning and forecasting methodology. Mistrust in the driver model results in business units establishing shadow forecasts and plans that subvert the implementation of the driver model. Likewise, a CFO study reported that, besides costs, insufficient implementation and change management skills are among the most severe barriers to adopting new technologies in financial planning and forecasting.

 

Process. To achieve a consensus on steering relevant value drivers, it’s crucial to follow a comprehensive process to systematically structure the driver selection. This process should be driven and continuously further developed by an orchestrator. The orchestrator is familiar with the driver-based methodology and has the appropriate business acumen paired with financial modeling skills and soft skills required to manage stakeholders. These characteristics are tailor-made to the skill set of the contemporary management accountant. The driver-based model should be validated using statistical models and evaluated against previous planning and forecasting practices systematically. This not only makes the success of the new planning and forecasting approach measurable but also facilitates learning and trust by comparing the two approaches side by side. Let the data speak for itself.

 

Technology. Appropriate software and infrastructure are prerequisites for driver-based planning. Although technology isn’t the primary driver behind successfully implementing driver-based planning, it’s certainly a driver of failure if not appropriately managed. The pivotal characteristic of technology is its flexibility to integrate predictive modeling, human adjustments, interactive simulations, visualization interfaces, and the ability to adapt the driver logics according to prevailing business dynamics.

 

Issues in Driver Selection

 

The selection of relevant and accepted value drivers is one of the main challenges of implementing driver-based planning because managers have different interests and perspectives as their departments face different business dynamics. The driver selection process is far from straightforward, encompassing a range of soft and hard factors. Given the complex nature of driver selection, it often leads to extensive discussions and alignment meetings. The primary issues that Swarovski faced during its transition to driver-based planning exemplify the potential barriers that may surface during the driver-selection process (see Figure 3).

 

12Weiserfigure3Final 

 

Level of detail. For management accountants as orchestrators of the implementation process, discussing the first driver tree proposal requires balancing the right level of detail in the preselection of drivers. There’s a fine line between jointly developing the driver model and the risk of having to buy stakeholders into your way of thinking and losing acceptance. In addition, orchestrators must ensure that the selected value drivers are mutually exclusive and collectively exhaustive, both in the preselection and in the advanced stages of driver selection.

 

Persuasion work. Getting all stakeholders on board and convincing them of the value of driver-based planning is a challenging task. In particular, regional financial managers may resist top-down predictive models due to algorithm aversion and limited model explainability, which can foster resistance as they feel they’re losing control over the numbers. Thus, the question remains: How can humans and machines collaborate as complements? The benefits of the machine must be substantiated. Machines may improve forecast accuracy and reduce forecasting time, which could be allocated toward more value-adding business partnering activities.

 

Data availability. Even when an agreement on value driver selection can be reached swiftly, missing underlying data across the organization may pose a substantial challenge. Adequate data availability is required to ensure traceability of results, particularly when a company operates in different markets or offers a diverse range of products. Companies must ensure that all data for the respective drivers is equally available at the required granularity. Otherwise, approximations using suboptimal drivers as alternatives that undermine acceptance and trust in the driver model could be required.

 

External drivers. Incorporating external variables, such as GDP growth, into a driver model takes time and resources. Isolating the effect of an external driver on a KPI without overshooting the complexity of a driver model presents a challenging task. For example, for Swarovski’s product categories, the yearly granularity of publicly available household panel data was insufficient. But a growing number of professional data platforms provide not only macroeconomic and financial data (e.g., Bloomberg or Refinitiv) but also market (e.g., Statista or Euromonitor), customer (e.g., Nielsen or Lotame), weather (e.g., Meteoblue), or health data (e.g., Oracle Health Sciences) for predictive analytics applications.

 

Biases. The selection of value drivers calls for a delicate balancing act between stakeholder mindsets and requirements, as biases inevitably emerge during the decision-making process. It’s crucial to acknowledge that managerial decision making isn’t always rational. For instance, stakeholders often introduce an excessive number of drivers in their pursuit of comprehensive and seemingly accurate results. This was evident in the case of Swarovski, where numerous factors such as store relocations, temporary closures, the annualization rate of opening effects, and other drivers were initially considered, in addition to the distribution sub-drivers illustrated in Figure 1. Such extensive lists of drivers can quickly become unmanageable and dilute the effectiveness of the decision-making process. To address this issue, it’s beneficial to prioritize key drivers and limit the time allocated to theoretical discussions. Instead, a greater emphasis should be placed on the implementation of prototypes. This approach not only streamlines the decision-making process, but also enhances the quality of discussions by providing tangible outcomes for review and consideration.

 

How to Master the Driver Selection Process

 

Driver selection isn’t a straightforward process but an iterative procedure involving all stakeholders (e.g., top management, finance, and sales departments at various levels). Developing the first value driver proposal and involving all stakeholders can take approximately 20% of the time, while the remaining 80% is devoted to discussions on driver selection. In other words, 20% is groundwork and 80% is alignment work. Based on the Swarovski case, we developed the following structured process as a blueprint to master value driver selection (see Figure 4).

 

12Weiserfigure4Final 

 

1. Gain top management support. Top management support is the cornerstone in the initiation of the driver selection process. The support from top management acts as the catalyst, setting in motion the driver selection journey. Without this essential backing, the subsequent process steps are at risk, due to a lack of user acceptance. Furthermore, top management’s role extends beyond mere support; it may also function as a steering committee, having the decisive vote in case of unresolved disagreements along the driver selection process. The influence of top management is multifaceted, underpinning the entire process and ensuring its smooth progression.

 

2. Define the orchestrator. Additionally, the driver selection process needs an orchestrator to take leadership. Finance departments, especially management accountants, serve as highly suitable orchestrators. On the one hand, they have the necessary financial, business, and stakeholder management skills to draft initial driver model proposals since they’re involved in planning and business partnering. On the other hand, they have the financial modeling skills, technology acumen, and data access to support an analytically driven planning transformation. Taking the net sales driver tree at Swarovski as an example (see Figure 1), the sales and marketing departments play an essential role in net sales planning alongside the regional management accountants responsible for the operative planning of stores. The orchestrator thereby coordinates the departments, which influence the net sales through their business activities, to reach a mutual agreement on relevant drivers.

 

3. Develop a value driver proposal. Driver selection doesn’t necessarily need to start from scratch. As illustrated by the Swarovski case, the transformation toward driver-based planning was preceded by a strategic realignment and an organizational restructuring, which included the creation of a long-term financial plan. This provided an opportunity to leverage the existing corporate knowledge and identify the key drivers through workshops with internal process experts. For the draft of the first value driver proposal, it’s recommended to focus on key drivers only and add complexity to the proposal only over time, as simplicity leads to acceptance. For example, for a net sales driver tree, work your way up from the fundamental logic that sales are driven by price and quantity. Avoid incorporating external drivers in the early stages to reduce complexity and maintain an understanding of all stakeholders involved.

 

4. Capture and engage stakeholders. The finance department can’t manage driver selection independently but must involve affected stakeholders. Relevant stakeholders that must be brought on board need to be clarified early. Bring all stakeholders into one room to create a common understanding of the purpose and setup of driver-based planning and each other’s requirements. Having stakeholders on board early on will most likely increase their acceptance of the driver-based planning system as they turn from affected stakeholders to involved participants. This step is important, as stakeholders might possess tacit knowledge or market understanding crucial for crafting a driver model proposal and evaluating the drivers. While these expert opinions form a robust foundation, they alone are insufficient for driver selection. This is particularly true when considering external drivers beyond the company’s control. For instance, Swarovski employed statistical models to assess the impact of GDP on net sales, thereby validating qualitative driver selection hypotheses and increasing trust in the model.

 

5. Present, discuss, and validate. The driver selection process requires several meetings with all stakeholders. This phase primarily encompasses discussions and consensus building on the identification, definition, and future application of value drivers. These discussions are driven by the pivotal question of the most significant business decisions pertinent to each stakeholder. It’s essential to understand that this step is highly iterative, requiring substantial time and resources for the preparation and execution of multiple workshops and follow-ups. This process step often involves the creation of prototypes, such as Excel models, between workshops to assess the viability and potential improvements of workshop ideas, considering practical constraints like data availability. Using proposals on spreadsheets or slides as a basis for discussion fosters agility over perfection in favor of multiple iterations. It’s also important to remember that there’s no single correct solution to driver selection. Therefore, preparing for iterations and allocating sufficient time is key to finding a solution that’s understood and accepted by all stakeholders.

 

Driver-based planning and forecasting offers an agile and analytical approach for navigating the complexities of the business environment. By selecting relevant and accepted value drivers, companies can make informed decisions and timely adapt their actions to reach their corporate goals in a VUCA environment. The insights from this article can help to guide practitioners in implementing driver-based planning and improving their financial planning systems.

 

The journey toward driver-based planning and forecasting can be a challenging one, as demonstrated by the case of Swarovski. The company found that a purely top-down, quantitatively driven approach to value driver selection, devoid of stakeholder involvement, hindered the acceptance of the new planning methodology. A series of iterative discussions aimed at building consensus, understanding, and acceptance proved pivotal to their successful implementation of driver-based planning and forecasting.

 

The role of the finance organization, and especially management accountants with their comprehensive skill set, was crucial to orchestrate the driver selection process. Simplicity emerged as a key factor in the acceptance of a driver model. Swarovski constantly balanced complexity and explainability of its net sales driver model to find an optimal trade-off that met the requirements of the organization.

 

The case of Swarovski underscores the importance of a structured, stakeholder-inclusive approach to value driver selection, underpinned by top management support and a careful balance between complexity and simplicity. Value driver selection isn’t a purely analytical task—especially in the iterative alignment process, soft skills are key to getting all stakeholders on board.

About the Authors