| blog 2 of 2| 3 minutes read |
In our last blog ‘’Machine Learning for Finance’’, we explained the difference between AI and ML. We also provided an example about how ML could be used in practice for an investment analysis. In spite of all the advantages of ML, companies struggle to extract the real value from this technology and do not know how to apply this within their organizations. In this blog we want to explain how ML can be used in an organization for a common problem. We will also dive deeper into how ML can be used within the Finance function.
Sales forecasting
In finance and business, a common task is forecasting the sales. This is usually done by a process using Excel spreadsheets that require input from different departments in an organization. This process is very time-consuming and the forecasts are usually based on a gut feeling. This input is not based on business drivers and therefore leads to biasedness. Another problem is that the accuracy can only be measured at the end of the forecasting period by comparing the actuals with the forecast. By doing so, the company has limited effectiveness in steering their business and could have achieved better results only if they had earlier, more accurate, insights.
Solution
To solve this problem, ML models can be used to make forecasts that are based upon business drivers. By making use of these drivers the organization could observe patterns and effects, such as trends and seasonality. This gives important and new insights into the numbers by using standard visualizations.
In the ‘old-fashioned’ way of sales forecasting, a company would look at their pipeline and opportunity data and could analyze the corresponding sales figures. When a company uses ML, it has the advantage that the data is presented in a way that could be translated into an action plan. For example, this forecast method will analyze past opportunities and translate this into probabilities for future sales. Companies can use this information to focus forecast sales even when opportunities are not known or expected. By combining internal and external data from multiple sources, additional trends and relations can be discovered to increase forecast possibilities.
Lastly, the accuracy of the model can be analysed immediately by comparing a backward looking forecast with the realized, historical data for past months or years. By doing so the accuracy of the model is tested for different trends and seasonality. When the accuracy of the model is lower than desirable, we have the option to let the algorithm search for strong business drivers or dynamics. The company has then the information and gains new insights to shift focus to these new drivers to achieve better results.
The sales forecasting model could be integrated in the financial and business management, by doing so you can make better overall decisions. Without a good idea of what your sales are going to be, managing your inventory and cashflow becomes almost impossible.
Interested?
CPMview provides tooling which can help to implement ML in financial reporting and analysis of daily operations. In the next blog we will dive deeper into how this ML model can be implemented into your company. We will explain this process and describe the important steps.