SAP Analytics Cloud: Introduction to Predictive Scenarios, and their applications
This blog post was authored by Seth Magehee, who works as a Technical Analyst for the NIMBL Finance Practice, SAP Analytics Cloud channel.
Hey everyone! My name is Seth Magehee and I am a Technical Analyst with the EPM Finance practice at NIMBL, TechEdge group. I consult on matters related to SAP Analytics Cloud, and I am based out of Denver, Colorado. In this blog, I will explain the usage and philosophy behind SAP Analytics Cloud’s Predictive Scenarios.
Predictive Analytics and Data Analysis sounds like an insanely difficult topic to cover without a Master’s degree and a handy dictionary to follow along with the technical jargon. This is the truth in most cases; however, SAP has completely done away with this notion though SAP Analytics Cloud (SAC) Predictive Scenarios. SAC Predictive Scenarios provides business analysts and other users with the tools to create more robust insights into data, with no need to sort through mountains of data and craft complex algorithms manually. SAC Predictive Scenarios allows users to create three different flavors of data analytic scenarios very quickly. In this blog, we will cover the following functions of SAC Predictive Scenarios:
- Time Series Forecasts
Fig. 1: Predictive Scenarios available in the home screen
Time Series Forecast
A Time Series forecast identifies key trends that directly influence future performance. All time series predictions are time dependent and require time data. Additionally, Time Series forecasts do not account for anomalies and adverse circumstances outside of the prediction. In SAC Predictive Scenarios, the software generates these key insights and generates trends in data automatically. The Time Series Forecast scenario in SAC can use both imported datasets and data models as long as there is a date dimension present. SAC calculates all the individual variables and their related statistics automatically when the user trains the forecast, and each variable can have its own predicted forecast. Because Predictive Scenarios is a feature within the SAC software suite, users can apply their predictions to either a new dataset or a private version. Private version export functionality is dependent on whether the forecast uses a data model to generate a prediction, and otherwise the user can export prediction data into a new dataset/CSV file.
Fig. 2: Gross Revenue Forecast from a demo forecast.
Use cases for Time Series Forecasts
Time Series forecasts, especially within SAC, have a myriad of uses for the end user. Some use cases can include, but are not limited to, the following:
- Revenue projections for product performance
- Cost projections to gauge how COGS can change over time
- Measuring cash flow projections
- Gaining insight into storage, warehousing, and other costs for capital and goods
- Using projection data to compare with budget data to find out how to adjust your budget for the next quarter
Overall, these use cases are just a small sampling of what analysts and other end-users can accomplish with time series forecasts in SAP Analytics Cloud. For users who already use SAC, it includes the added benefit of seamless integration with SAC Planning with version management. When SAC does most of the heavy lifting when it comes to forecasting, end-users can enhance their workflows by automating the mundane task of generating a time series forecast.
Classification is the process of separating data into classes. In this scenario, the target is discrete, and the goal is to separate data into groups in a meaningful way. Say for example you want to analyze the user base of a company’s mobile application. In an unorganized database, each line of data contains the user information, phone platform, location, and usage data. Classification is the process of separating the user base between iOS and Android users, and it shows how particular data influencers like location or usage data change based on OS type. In SAC Predictive Scenarios, it parses through the data using a binary discrete variable and simultaneously considers other variables in the dataset. These are known as “Influencer Contributions,” which show how influencers affect the output of the classification. The analyst can also view individual statistics for each influencer in SAC, and the software calculates all these statistics immediately when you train a classification scenario. SAC additionally generates contribution charts which visually show how each class contributes to the total dataset.
Fig. 3a: Influencer Contributions of a Classification, ranked by influence on the classification.
Fig. 3b: % Detected target performance curve from a sample classification scenario in SAC.
Use cases for Classifications
Overall, classification is a powerful tool that allows end-users to gain insight into subsets of data. Some examples of questions that could be answered by classification include the following:
- Who has a higher probability of responding to web ads?
- Which store locations have higher sales potential for my product?
- What is the likelihood that the email addresses I send my marketing newsletter to are active or not?
All these use cases are pertinent to business decision-making, and SAC can generate those insights automatically without having to sort through thousands of lines of data manually.
Regression Analysis is a way of sorting out which variables have the most impact in a dataset. The target is continuous in this method and is primarily used to show the relationship between two or more variables in a dataset. For example, an analyst would use regression analysis to determine whether there is a correlation between users’ location and average daily screen time. Traditionally, this involves taking all the data, plotting it on a scatter plot for the two variables, and finding an average line if the data fits a particular trend. This would then be done for other variables that are not included in the initial regression in the traditional method. In SAC, a user would simply pick one variable and the software automatically generates regressions, complete with relevant statistical data, for all variables in the dataset and how they relate to the target variable.
Fig. 4: Predicted vs. Actuals from a sample regression scenario in SAC.
Use cases for Regressions
Regressions are insanely powerful in business decision-making. They can take seemingly random subsets of data and generate insights that would normally not be visible. Some of the questions that can be answered by regression analysis can include the following:
- Will the reach of my billboard marketing campaign be impacted by daily highway traffic?
- How will sales of my product change based on a competitor’s new product release?
- How can web traffic for my website change based on my social media activity?
- How is quality control impacted by implementation of a new process on the production floor?
SAC can take all of these considerations into account and provides deep insight into the individual impact multiple variables have on a dataset instantly. Like the other concepts mentioned earlier, SAC makes regression analysis substantially less of a headache than doing it manually and increases the capabilities of the end-user.
In conclusion, these methods demonstrate how SAP has made predictive analytics and data analysis accessible and simple for all users, and how they provide deep and powerful insights into datasets instantly and seamlessly integrate with other features within SAP Analytics Cloud.
Thank you so much for reading! If you enjoyed reading this post, please drop a like and share it! If you have any questions, please feel free to comment on this post. If you want to explore the SAP community, go to the SAP Analytics Cloud, SAC Predictive Scenarios, Smart Predict, and Predictive Planning tags respectively to view and engage with related threads on this topic.