Financial forecasting is one of the most impactful tools that the finance department has. When, in 2016, Lufthansa had a profit warning its stock value plummeted 20%. Another, more recent example, is IT company Atos whose shares plunged by 15% due to misstatements and inaccurate forecasts. This resulted in the company’s free cashflow going down from 600 million euros to merely positive. Likewise, many companies have been known to radically cut workforces when profit warnings arise. In more positive scenarios increased profits can lead to big investments. This goes to show that financial forecasting has a big impact on the future actions of a company.
However, it is extremely common that there are significant differences between the forecast and reality. Meaning that the corresponding actions to positive and negative results might not always be as justified as we think. Would it therefore not make sense to improve the forecast to improve its reflection of reality? Using AI in your forecast is the best solution for this. Read why that is below.
Financial forecasting is of course based on data, but when using AI you are delving into the world of big data. That means that you have more data and more detail to your data. You are then able to take into account each customer's specific individual behavior and therefore have a more precise forecast.
For example: For a telecom organization it is impactful to know what devices their customers use, as it has been discovered that the device has strong correlations with the amount of data or minutes the customer uses. Knowing when these devices will be exchanged for a newer model is vital. This is because when customers upgrade their devices, their usage most likely changes as well. As you can imagine this has direct consequences for revenue. Being able to accurately model this gives you precise insight into a revenue stream.
The second benefit AI has in financial forecasting is also related to behavioral forecasting. The data that is now most often used is compiled of totals or averages. It lacks a granular view of the data. The effect of this is that small customer groups can get skewed in results and that, when used in calculations, snowballs into a bigger gap between reality and forecast.
When using AI, you gain a better understanding of what your customer’s behavior is and how this affects the predictions. Take for instance the difference between ‘early adopters’ and your ‘sleeper’ customers. The early adopters jump at every opportunity your organization gives them but the sleepers don’t react to anything you do. Knowing not only the difference between the two but also the size, worth and behavior of both customer types drastically improves your forecast.
Because AI gets into the most detailed level of data you can expect better insights into seasonal effects. Of course, you know the general seasonal effects in your company but does the model you use to apply these take into account what is truly dependent and independent of the season?
A good example is the impact of a weekend relative to a working day within the telecom industry. On Sundays, customers make fewer calls (up to 40% less) than on a weekday. Of course, holidays also affect the number of calls someone makes. In retail, we see the opposite effect. The insight into this data on customer level increases the accuracy of your forecast.
Learning from mistakes is one of the most valuable things you can do as a human. The thing is that, when it comes to forecasting, AI is much better at learning and adjusting than we are. This is not only due to the speed with which a computer can understand and reevaluate data but also because it is unbiased. Model management (which is minimizing the gap between reality and forecasting) is a key element to a good AI-powered financial forecast.
When we analyze the forecast and reality we might compare and adjust a tenfold of combinations. With AI the number of combinations that can be made between result data and forecast data can easily reach five thousand. It is therefore concluded that AI creates more and better improvements than we could do manually.
Want to have a chat with us to discuss how you can use AI in your financial forecasts? Get in touch here.
This blog is based on the article ‘Improving financial forecast through customer behavior data’ by Wynfrith Meijwes and Marnix Bügel. You can download the article here.
We’ve made starting easy for you. Use our checklist to get ready for an AI implementation within Finance and you won’t encounter any unforeseen problems.