Product Information
#ASUG Webcast Recap Smart Predict Update Recap
You can watch the replay here BI: Smart Predict Update and slides are here
If you missed a previous webcast, go to ASUG BI Webcast Recording/Slides Reference List for 2019
Join us for these upcoming webcasts:
April 9 BI: Data Protection & Privacy with BW/4HANA
April 10 BI: What’s New with Application Design in SAP Analytics Cloud
April 16 BI: What’s New in Transitioning from SAP BW to SAP BW/4HANA
April 25 BI: Preview of Business Intelligence and Analytics Sessions at ASUG Annual Conference
Source: SAP
Richard Mooney, SAP, provided this update, with overall Smart Predict capabilities in SAP Analytics Cloud
How does it fit inside Augmented BI, and SAP Concept of Intelligent Enterprise and roadmap
Introduction: Augmented BI and the Intelligent Enterprise
Source: SAP
Key concept behind augmented analytics, how rethink analytics using intelligent augmentation
Take concepts today of dashboards, agile BI, reporting and connect intelligent capabilities to them
Data driven alerting, predictive and prescriptive, suggested points of interest, conversational AI
Bringing together human authored machine guided analytics to make it simpler for users to interact with analytics data
Source: SAP
Analytics Framework – what do you do with it?
Themes of assistive -machine learning to make it faster
Descriptive to describe it
Diagnostic, where you go into a deep understanding of something interested in
Predictive – build predictions, generate new data, make decisions
Prescriptive – use new data generated to make decisions
Combine with 2 product categories – planning and BI
Build capabilities to support business users – search to insight – conversational AI – find insights using natural language, and add it to your story
Interacting directly with data without understanding UI, makes it accessible
Smart Insights, Smart Discovery
Smart insights – instant analysis, use machine learning and statistics to tell you about your data, helps analyze data faster
Smart Discovery is a deeper analysis; articulate a question, such as order value, and what drives that; run machine learning algorithms against that data set to understand it better
Bring together all in one place; storybuilder to build story, key driver analysis, natural language generation, find relationships, outlier detection to find unexpected values, also has forms simulation
Smart Discovery does all the work for you and you can change, update to it and build story to tell story
BI features powered by machine learning – built within context of a chart or story, such as time series data, build forecast
Look at dataset, look at smart grouping
Apply machine learning augmentation in analytics process; extend AI, bring prescriptive and next best actions – bring predictive in prescriptive analytics
Intelligent learning to proactive actions/alerts
Augmented data preparation to draw more insight – planned for Predictive this year
Increase level of machine learning in planning
Source: SAP
Key areas looking at today
Will start as internal project, and publish material externally
Recognize that for people, predictive can be challenging to figure out best place to use it
Supported at data mining level, classification and regression level – open ended
Classification could be several use cases; want to do use cases but not know where start
Use case library – 5 use cases at moment
Provide high quality material that is useful for business users
Include demo data to try out sample
Presentation includes use case, ROI
KPI’s to improve, measure better, and a hands on tutorial to use Smart Predict
Use cases are above
Payment forecasting – already built out analytics content network capability (based on S/4HANA Finance OData Connection)
Are the use cases useful, and other use cases
Usually, use case comes down to having enough data to build predictions
Source: SAP
Launched Smart Predict June of last year, went GA in September 2018
German Football club is using Smart Discovery and Predict to improve performance, push to dashboards, used by executives to see where invest via wages and player acquisition
Shows benefit of predictive and BI together
See interest in segmented forecasting, automated forecasting of expenses by business unit and product area, what drives customer retention
Source: SAP
Not covered
Source: SAP
Context driven AI every where, as a menu option or button in area where doing the function
How connect Smart Predict process to story builder and planning
Not want to rebuild SAP Predictive Analytics on premise into SAP Analytics Cloud
Explainable AI – how made prediction and patterns found
Continue to invest in LIVE; need underlying platform to support the algorithms; easy to do with HANA as it has underlying algorithms; look to Q3/Q4
Already have live for Search to Insight
Near term support for Live will be about HANA
Source: SAP
Q2, publish predictive scenarios to SAP S/4HANA using the Pai framework; 12 use cases already delivered
Customers can use SAC and publish to S/4HANA and it will execute in S/4HANA
Payment forecasting is available in SAP Analytics Content Network
Choose explanatory variables you want to output to dataset for consumption in stories
Q3- HANA data connection, starting with tables and SQL views
Can already acquire data from OData
Refresh stories and datasets in Q3/Q4 – work with predictive data like agile BI
Longer term – acquire data in SQL databases, augmented data prep – to draw predictive insight
Source: SAP
Way Smart Discovery works today; flattens data to get insight to lowest level
Source: SAP
Where get to; get it at the customer level, takes transaction information by doing joins, merges and pushes it into auto derive features, like they did in Data Manager in Predictive on premise but make it easier
Source: SAP
Roadmap for live
Q3/Q4 timeline
Train and apply on remote HANA is planned and include augmented data preparation – “full self services for business users”
Question and Answer
Is there a way to upload or submit your own machine language algorithm, which would handle all the features SAP’s ML does today?
If you want to use algorithm outside of SAC, use HANA or Data Hub
When will it be possible to use time series forecasting on top of a model in a user friendly way? Moreover would be great to use a created model inside a story over the standard button for prediction
3 places where time series exist in SAC: 1) within a story – exists today 2) in planning 3) segmented in smart predict – look at how provide debrief story form in an easier way than today; not on near time road map
Would BW on HANA enough to use live scenarios?
if willing to move view to table, yes,
Does Beta program for Live Smart Predict is only relevant for customers? Can SAP Partners join as well?
if fulfill criteria described.
Is there a Special liecence needed to develop with Application designer in SAC?
comes with the BI license
Questions answered after the webcast:
| Possible Use-Case for use case library: Pricing prediction for most optimized sales and margin return |
| We will definitely consider it, we need to add optimization libraries first which we eventually plan to do. |
| is there any kind of predictive asset management on top of use cases? e.g : adjust the predictive Algorithm %2B get it’s exact math ? |
| We provide configuration options to configure the business question and also debrief to explain the model. |
| will smart predict be useful for P&L ? |
| Yes, forecasting both the top and bottom line can show you risk factors and reasons why |
| What is the approach to tailor the predictive models for scenarios where the model accuracy is not satisfactory ? Is there an option to switch between the models in SAC and models designed on HANA platform |
| Accuracy is not typically to do with algorithms, it is to do with data. More data, greater accuracy. In future we will provide data preparation to improve accuracy. |
| It is always mentioned that every user can use predictive. Isn’t it important to know how machine learning algorithms or other statistical algorithms act in the background? I think it’s very dangerous to let all users work with this feature. Results can be misunderstood. What is your opinion or who should work in your opinion with this tool? |
| Users need to be able to interpret the results and we provide easy to understand explanations to help them, we also make it clear where a predictive model is not good and explain how make it better. People are capable of driving cars without being mechanics. |
| Is there going to be a SAP course dedicated to Predictive? Currently there are two visible on training.sap.com: SAC01 – Introduction To SAP Analytics Cloud | SAP Training (SAC01_EN_Col17) and SACP20 – SAP Analytics Cloud for Planning | SAP Training |
| We are adding Smart Predict to the new SAP Analytics Cloud OpenSAP course, soon to be released. |
| I’m interested in reviewing the demo in SAC, how can i get access? |
| We will make it available soon with the use cases. |
| is there a use case recommendation when to use predictive in SAP Data Hub and when in SAC ? |
| Data Hub is for data engineers, SAC for analytics |
| Smart Predict is not available in our labs. Is there any limitations ? |
| You need to have a Cloud Foundry enabled tenant, we are working to make sure all customers have this. |
| Could I have Predictive algorithm in my tenant demo? |
| You need to have a Cloud Foundry enabled tenant, we are working to make sure all customers have this. |
| I there predictive scenarios SAP data HUB isn’t covering ? |
| Same use cases different use types. SAC is designed for business analysts. |