BI Innovations – Go Beyond BI with Predictive Analytics Webcast Summary
This was an SAP webcast this past week. The presentation is here.
Figure 1: Source: SAP
Most not using predictive solutions today (of those attending the webinar)
Not a single solution for predictive
Figure 2: Source: SAP/Gartner
Evolution from raw data, clean data, to BI for prepared reports, analytics, OLAP analysis, then move to predictive modeling, until you can do recommendations based on company
This slide doesn’t mean you are going to evolve your existing technology; companies who work with data move on evolutionary line
Figure 3: Source: SAP
This doesn’t mean the same people doing BI today will do predictive
The speaker said there is a “chasm”
People who do BI and people who do predictive are different
You have to expect some difference in workflow to introduce predictive functionality
Figure 4: Source: SAP
It is different at different companies
Figure 5: Source: SAP
Advanced analytics encompasses predictive analytics
Idea is you have a lot of data about the past; you know who your best customers are, and how products are performing in the market, when machines have issues, who are the best employees
Advanced analytics applies mathematical algorithms against data to see if hidden patterns exist in data
Figure 6: Source: SAP
Where will you use the model?
Understand how past behaved, and apply model to new data and will give you recommended actions
Operational analytics usually means there is a machine at the end; the goal is to be quick
Exploratory analytics – look at past data, look at model, rules to better understand business. Usually a human looks at results and decides
Predictive can be used in both situations.
Figure 7: Source: SAP
Question asks whether you hire data scientists for predictive projects
Data scientists are hard to find, so most are using BI people
Figure 8: Source: SAP
Why is there a chasm between predictive and BI?
Humans have a limited ability to comprehend while a machine does not
The data for predictive is typically bigger than BI with a lot of columns and rows
Machines can handle complex questions than a human can
Figure 9: Source: SAP
With BI, humans slice and dice, aggregate, filter, simplify, and the goal is to make visualization simple so anyone can understand
Advanced analytics, start with data, enrich data, to give more information to machine so machine finds interesting data
Figure 10: Source: SAP
BI and advanced analytics are “complementary”, says the speaker
With BI you can identify an opportunity or a risk
What you can do with predictive, give more data to machine and let machine tell you what to do to take advantage of opportunity or risk; machine can tell you next best action
Figure 11: Source: SAP
Most attendees do not have data scientists in their company
Figure 12: Source: SAP
SAP BusinessObjects Predictive Analytics is the flagship product
The solution is used by a data scientist or BI users familiar with predictive workflow
It can work as a stand-alone solution or with the predictive libraries in HANA
You can embed the model in application, such as hybrid marketing
The model can be exported as Java code, C++, and then copy code in your own application. The end user uses it without knowing the complexity
Figure 13: Source: SAP
Examples of embedded predictive analytics we use everyday
Question & Answer
Q: Can you please suggest how can i do my sales data analysis on map (shape files / postgis / google map) ? is BI the product for it ? or any other ?
A: the whole BI suite (Web Intelligence as well as Lumira and Design Studio) could support such a scenario. With partner extensions available with the Analytics Extension Directory (https://analytics-extensions.enter.sap/) add even more richer visualizations in the mapping space in these products. The predictive information would be persisted in your DB of choice (best integration with HANA & Hadoop, but any other DB would work as well) or computed on the fly (e.g. in streaming environments) and SAP’s BI tools
SAP Predictive Analytics uses social analysis where things are and when they are there and an impact
Q: Can I access universes with Predictive Analytics?
A: Yes, there is a query panel
Q: How to use predictive with BW?
A: If you build specific queries to provide details you can use Predictive