Agile Forecasting For Finance

Author: Tsholo Masakale

 

INTRODUCTION

Over the last few years, we considered disruptions as purely emerging from digitisation. Few, if any, finance functions foresaw a pandemic as a more disruptive element than digital.

Due to the pandemic, all finance functions had budgets rendered unrealistic almost overnight. However, CEOs still had questions about the organisation’s performances in light of the pandemic. They required a revised strategy coupled with financials. New financial forecasts needed to be created at speed to give answers in an environment riddled with uncertainty.
Furthermore, various iterations of the forecast were needed over time as more certainty was obtained and previous assumptions were found to be no longer valid.

In the simplest terms, forecasting is the process of drawing likely pictures of the financial future based on today’s assumptions. Through this process, we try to predict how that picture would change in case our assumptions change.
The intended outcome of the process is to enable business to make optimal decisions – keeping in mind that these assumptions impact what the optimal decision may be.

We thus recently found ourselves in an environment where we realised the necessity to forecast in an agile manner.

CHALLENGE DESCRIPTION

How did organisations fare considering the need to adapt given constant changes rapidly? Simply put, most failed spectacularly.

Most organisations discovered the following:

  • It took a long time, many people and multiple data sets to perform detailed forecasts based on previous budgets.
  • The focus was on providing accurate financial measures as opposed to directionally correct operational outcomes given known information and lots more unknowns.
  • Executives could not receive timely updates on the various scenarios they needed to be modelled. In turn, decisions were based on “gut” rather than the results of modelling based on key drivers.

why is this a problem?

  • Forecast information is not delivered promptly to enable analysis, thus rendering finance an ineffective support function and not a key input provider in strategic decision making.
  • The desire to provide accuracy versus directionally correct answers delays information provision, and thus, parallel forecast processes are initiated by the business.
  • By the time forecasts are provided, using a bottom-up approach, assumptions have changed, and the information is of significantly less value for decision-makers.

what are others doing to solve it?

    Finance functions in more successful companies have identified key drivers of performance and are creating summarised forecasts on a rolling basis to provide quick answers on scenarios to be modelled.
    Upon executive acceptance of the forecast scenario, they can cascade information down to the rest of the business based on an agreed allocation basis for the revision of budgets into detailed revised budget scenarios.
    Further, they are continuously looking for ways to streamline the forecast process to provide likely outcomes based on multiple what-if scenarios.

    what is the solution model?

      One of the leading methods we have encountered in forecasting was to have a base set of assumptions and critical drivers used throughout the forecasting process to change once. When coupled with historical data, the assumptions and drivers would flow through the scenario planning process and generate results.

      The key was to use historical data and analytics on the relationship with other drivers to improve the assumptions continuously. This approach allows a lot more agility as external influences could be easily incorporated and modelled without relying on tedious manual processing and agreements inherent in the planning and budgeting process.

      what are the steps to consider?

        In developing a driver-based forecasting process, what are the key elements to bring a viable forecast whilst looking for ways to improve on the process over time?

        We have identified nine steps to guide organisations through the change:

        1. Start simple

        Don’t overcomplicate the first iteration. The model aims to support decision-making, and the model should be understandable to test the logic with other stakeholders in the organisation and the forecasting process.

        2. Define the target output

        The target output for many forecasting models is often overlooked, resulting in forecasting for its own sake. Defining the intended output e.g. “Increase profit margin” or “optimise product mix”, helps to separate the signal from the noise and highlights areas for separate models if needed. Engage the business, elicit their requirements to ensure the target output is aligned with them.

        3. Identify the main assumptions

        This is often where the magic happens. As a start, it is useful to assume that the main inputs to the model will stay constant over time or grow at a constant rate. Many clients either over or under-think this component. Our experience shows that assumptions will evolve and mature as more data and experience are incorporated into the process.

        4. Model the mechanics

        Depending on the output target, the right model type is selected, and the mathematical relationships developed (e.g., the 3 Statement financial model or the CAPEX prioritisation model). Ideally, this should be done and prototyped in MS excel and only moved to scalable systems when the model is understood.

        5. Build the initial driver trees

        Driver trees help us visualise and articulate proposed causal relationships, whether mathematical (e.g., accounts in the income statement building up net profit, or intuitive (e.g., trained employees have fewer production mistakes). Driver trees are a valuable component to modelling but could often become onerous to maintain, so it is recommended to limit driver trees for the significant value contributing buckets.

        6. Link assumptions to the driver trees

        This is where formulas are introduced to drive components of the driver trees with assumption-based data. In some cases, this is also where assumptions are made where company data ‘is not available. Here it is helpful to check the extent of influence that the business has on the chosen drivers. Further, an opportunity exists to test linkages to other metrics, e.g. how revenue increases are likely to have on working capital and ensure the impact filters through to the working capital section of the balance sheet.

        7. Incorporate data

        In this step, we incorporate organisational and external data into the model for all the base accounts and assumptions where data is not available. Once we have done this, the overall model should work and respond effectively to changes in assumptions or updated data when it is published within the organisation.

        8. Apply analytics to test the relationships to drivers over time

        As more data is collected over time, it is prudent to leverage analytics to test assumptions and the drivers that impact the outcome.

        9. Re-engage the business, iterate, and maintain the model

        We have found it helpful to iterate the model over time to increase accuracy to the point where the output could support meaningful discussions. After that, the model should be maintained but not necessarily grow in complexity as it increases maintenance and decreases understanding. Business must then be re-engaged to ensure the model reflects the reality of their operations and buy-in is obtained from them.

        What are the impediments or pitfalls to consider

        • Excessive detail – This does not equate to accuracy. Forecasting on every expense line does not result in greater accuracy, especially if two to three key drivers drive expense increases.
        • Lack of agility – Forecast scenarios should be changed relatively quickly; therefore, the team working on Forecasts must be agile and capable of quick adaptations/iterations of the models.
        • Accuracy – Accept that the organisation cannot be accurate in a scenario where the organisation does not have all the answers, and there is no mechanism or time to find the perfect answer.
        • Accounting mindset – A forecast does not necessarily have to yield a balanced balance sheet. What is the impact of changes in consumer demand on share composition in the company?

        Jigsaw Advisory specialises in guiding CFOs and finance teams through their digital transformation journey. We value and nurture our niche position as a trusted transformation advisor to many top-level CFOs and are so confident about our recommendations that we choose to implement the solutions ourselves. That way, we fully understand the business context for the challenge and use that to serve as the focus throughout the implementation and benefits realisation process. Get in touch with our senior team here.