What is SAP doing in the realm of Big Data, and how apply predictive to Big Data? This is why I watched this SAP webcast last week.
Figure 1: Source: SAP
Figure 1 shows the personas for predictive analytics, as SAP sees it. Before, there was the data scientist who used algorithms, neural net, and coded in R
The Data Analyst would obtain the right reports
The rest are business users
Figure 2: Source: SAP
The speaker explained that SAP’s current product, Predictive Analysis, is a graphical layer above R
R is a well know language for predictions with over 3000 algorithms
Infinite Insight (formerly known as KXEN) allows business analyst to build a model (no code, no knowledge of algorithms)
SAP wants to build visualizations on predicted data, offering Lumira
Figure 3: Source: SAP
Figure 3 shows how customers are using SAP predictive solutions.
The speaker said over 600 customers using this today
Biggest customer is Wal-Mart
Vodafone Netherlands has 700 models
Figure 4: Source: SAP
Figure 4 explains with HANA you can process data real time unlike traditional
The idea is to run Infinite Insight and Predictive on HANA real-time
Figure 5: Source: SAP
What is targeted towards which person – the business, analyst, data scientist
The data is fed into the in-memory database, including unstructured, geographic, structured data
As an example, a major aircraft manufacturer needs to know when will a part fail in real-time to help prevent an incident from happening
What about Big Data?
Figure 6: Source: SAP
Figure 6 shows the example of a phone bill, using a CDR – call data record
This “big data” records every call received so it is a “long table”
A possible prediction is whether you leave Verizon this month; the speaker said the answer is in the data
They look at usage patterns , people calling, duration of calls, something will say if you leave
Will you buy new iPhone 6 when it comes out? Are you interested in new data plan? Answers will come in form of prediction.
They derive attributes from Big Data (see slide) and different data sources – CRM, who you are, where you live – ERP – know what phone you have – what types of products you bought
CDRs change constantly
Using predictive models, they target customers, graphs show the amount of lift
Example: Call Center Agent
The speaker said it is not just a lot of data, but it is unstructured, and it is all the data
It goes to another level when you want to make predictions with the data
Figure 7: Source: SAP
Figure 7 shows the prediction of offers – in this example, the Call Center agent is targeted to offer Showtime to the caller.
Figure 8: Source: SAP
Figure 8 is another example of a bank with 2,500 pieces of information about each customer.
This big data came from different sources
If bank sends e-mail, how likely are you to open it?
The data shows that a doctor is more likely to open an e-mail than a Ms (true)
Figure 9: Source: SAP
Figure 9 shows the interface of Infinite Insight (former KXEN)
Roadmap: Connect to Hadoop
(subject to change of course)
In the roadmap they are looking at connecting to Hadoop for realtime predictions
Predictive Analysis and Infinite Insight will be one tool
SAP plans to embed predictive into applications like CRM
Question & Answer
Q: Is it possible to use Predictive Analysis on top of non-SAP data?
Q: What are the areas where predictive is used?
A: CRM, fraud, predictive maintenance
Q: Use predictive with HR ?
A: PoC’s ; looking at integrating with SuccessFactors and predicting employee churn
Q: Using SAP Predictive Analysis now; how does that relate?
A: Tools will begin to merge
Predictive Analysis is more to the data scientist
The plan to add Infinite Insight to Predictive Analysis
Will be both, soon be one
Q: Is it possible to automate business processes to respond to predictive model?
Q: How fit into Integrated Planning and BPC?
A: SAP is meeting with teams and to see how to get these tools embedded into applications
There are Predictive Analysis sessions at BI 2014
For ASUG Annual Conference, we are planning some special sessions around Infinite Insight/Predictive – with customer speakers (to be announced)