Technology Blogs by SAP
Learn how to extend and personalize SAP applications. Follow the SAP technology blog for insights into SAP BTP, ABAP, SAP Analytics Cloud, SAP HANA, and more.
Showing results for 
Search instead for 
Did you mean: 

Time Series Forecasting on SAP Analytics Cloud

Time series forecasting is a new capability in SAP Analytics Cloud for planning. The forecasting function uses SAP’s proprietary time series technology to analyze trends, cycles, and fluctuations in time series data. Utilizing historic data, it predicts future values using the SAP HCP predictive services. Embedded in a business user-friendly interface, users can simply select their desired forecasting periods and click run without having to choose between different time series algorithms. Thus, this functionality assists business users in making more data-driven decisions especially when combined with their business planning processes.


What is the difference between a predictive and planning forecast?

Let’s get a better understanding on the difference between predictive and planning forecasting. Predictive forecasting and planning forecasts go hand in hand. A planning forecast uses historical data, business acumen, planning models, and predictive forecasting to determine future outcomes. In addition, planning functions such as spreading, distributions, and allocations are also key in helping to build a planning forecast - all of them available today in the product. Predictive forecasts run a time series algorithm on historic data that considers trends, cycles and/or fluctuations to predict future values. The future values predicted through a predictive forecast can then be leveraged to help business users take a more data-drive approach in the business planning process.  Below is a visual diagram of how predictive forecasting and planning forecasts relate to one another.





Scenario: Expense forecasting for data-driven budget planning

In this scenario, we are in the role of a business user who is planning their company budget for the rest of the year. We want to know what the company’s forecasted travel expenses will be for the next few months using time series forecasting, this will help us better plan our budget and ensure we are optimizing our company’s financial resources.


Setting Up:

  • Create a “New Story” and select “Add Grid”

  • Click “Insert” and select the “Concur_Predictive” model to import the travel expense data we are going to perform the forecast on



Next, we add the time dimension under columns in order to break down and analyze our travel expense by quarters and months.

  • Select “Builder”

  • Click “Add Dimensions” and select “Time” as the additional dimension to add

Before we start our time series forecasting, we should create a private version of this data so that we don’t accidentally make changes to the actual data set. We do this by:

  • Clicking on Version Management under the Plan tab section to create a copy of the Actuals data

With this current data set, we have travel expense data available until April of 2016 and we want to forecast the travel expense for after April. We can do this with our predictive technologies through clicking on Forecast to start our time series forecasting.

  • Select the cell that shows our total Travel Expense, in this scenario cell C6

  • Click on the time series forecasting icon

After clicking the time series icon, a dialogue will open which automatically identifies the last date we have data for and will start forecasting from that date until a future date that we choose. For this scenario, we want to forecast until the end of the year which will look something like this:

After clicking ‘Preview’, the time series algorithms returns the best model by analyzing cycles, trends, and fluctuations. Once the analysis is completed, we will be able to see what the model forecasts as our travel expenses to be until the end of the year and will also provide us with the confidence interval along with it.


Clicking "Ok" will copy the forecasted values into your previously specified version.


Additional Resources:

Further details on the time series algorithm can be found here: