Exclusive portal designed for our Customers and Partners

Log in

Request for Access

countries flags

Automatic forecasting – does it help to improve planning?

Apr. 24 2020 by Prof. Dr. Karsten Oehler, Solution Architect - CCH Tagetik DACH

Performance Management Budgeting & Planning

It’s not surprising that the demand is high for automated and improved data quality through artificial intelligence and machine learning. Budgeting and planning are usually considered time consuming with limited, often unrealistic results which can quickly go out of date.

So what can automation achieve? Automation beyond simple standard tasks is not as easy as it sounds. And opinion is even split about the function of planning and forecasting: there are two broad schools of thought about where planning and forecasting should be headed.

The first says planning in volatile times is more or less useless. It would be better to focus on agility and flexibility. Consequences of actions must be understood if increased speed in decision making is to be achieved. Simulation on the basis of cause and effect could help to test the possible effects of activities. Unbiased forecasts are not useless but planning efforts should be reduced drastically. This view is best presented by the Beyond Budgeting Roundtable (BBRT).

The other group claims that with the right information it is possible to predict the future in a meaningful way. Planning can be significantly enhanced by better forecasting and other improvements. Of course, negative aspects of information hiding cannot be avoided completely. However, with a better system it is possible to overcome these problems.

Both positions are valid. The right approach depends on many factors such as the volatility of the industry sector in question, the economic climate, customer behavior, competition, and of course the cultural climate: How openly are expectations communicated? How strongly do individual objectives influence the forecast?

A compromise to reconcile both directions might be a leaner planning process using machine learning for forecast automation, sophisticated analysis and better understanding through simulation. This can become a cornerstone of a new planning process.

automatic forecasting

Forecasting and simulation should not be viewed separately. Both are based on the principle of cause and effect, and can share the same business model. For example, if you have better understanding of all the inputs relating to your discount policy, you not only improve your sales prediction capabilities but are also able to simulate decisions with their possible consequences.

A lot of the data required to establish cause and effect is available and already included in planning and forecasting models. Production calculations can be extracted from the ERP environment. In CPM, sophisticated planning tools like CCH Tagetik come with predefined calculation rules for P&L, cash flow and other KPIs. But what about drivers you expect to have an impact but which you can’t quantify? Or drivers you don’t even know about yet?

A lot of drivers exist outside the traditional CFO’s scope. Machine learning can help to identify dependencies like the effects of campaigns, discounts etc. on sales figures. A good understanding of these effects can help improve forecasting and simulation because they make results more realistic. But this means that planning has to cover more than just the financial sphere.

So it is surprising that sophisticated machine learning approaches are rarely used in the area of the office of finance, particularly for planning and budgeting. Innovation in finance is often restricted to robotic process automation (RPA) and transactional improvements like account reconciliation or intercompany matching.

RPA can definitely help to reduce effort. Yet if planning is based on a single platform like CCH Tagetik with an integrated workflow there is no need for RPA automation. Investments could be better targeted at more important challenges like improved business understanding through machine learning.

Planning and forecasting it is not only about automation but also about new insights and – let’s not underestimate this aspect – better alignment of planning units. Do all participants share the same assumptions? Do they have access to machine learning methods and the necessary data to get insights?

The best way forward is to adopt an interactive and flexible approach to embedded machine learning. There will be trial and error but even simple support features like identifying adequate seasonality figures or calculating days of sales outstanding per customer group can significantly improve forecasting quality and, consequently, planning.

For the manager it is useful to work interactively with machine learning results. The results must always be challenged: are you sure you have used all related data? And human expertise is still necessary because machine learning provides correlations, nothing more.

Without a theory behind them, pure statistics doesn’t lead to better understanding of your business. Automation is important but not the only goal. It is learning about cause and effect, supported by automation, that really helps improve planning and forecasting quality.

Discover more about the CCH Tagetik for Budgeting, Planning & Forecasting here.

Share this post!