Empowered Self-Service BI with SAP HANA and Lumira with Honeywell – ASUG Webcast
We had a great ASUG webcast today with Dennis Scoville of Honeywell Aerospace. My notes are below. Thank you to Dennis for sharing his story with ASUG and agreeing to this to go on SCN. Also meet Dennis here Meet the ASUG TechEd Speaker Dennis Scoville – Empowered Self-Service BI with SAP HANA and SAP Lumira
Figure 1: Source: Honeywell
Figure 1 shows what Honeywell did in the past for self-service BI.
Their environment was vanilla-based columnar based SAP BusinessObjects Web Intelligence reports
They gave their users personalized reports and allowed them to publish to a group folder or favorites.
They were not allowed to report to enterprise report folders.
The data was an Oracle database without real-time
They extracted data from SAP using Data Services and other sources.
The process took 13 hours from 6:00 pm to 7:00 am following morning.
Figure 2: Source: Honeywell
Figure 2 shows the “before system” look. SAP source systems include ECC, GTS and non-SAP
They had transformations that were put into staging tables that went into Rapid Marts.
That data went into a data mart to SQL Server Analysis Services.
Then they used Web Intelligence against the Rapid Marts.
What was wrong in Figure 2? They had several sources of failure, job-wise, several transformations.
It took a host of skill sets to maintain this, he said.
The end user had to go to different places and there was no drill-down from SQL Server to the Web Intelligence.
Figure 3: Source: Honeywell
Figure 3 shows what Honeywell has done since Figures 1 & 2
They acquired HANA and the BI4 tools and are now on BI4.1
They started using the other tools
They are determining use cases for newer solutions to them
Lumira Server is not out yet.
They are also looking at Predictive Analysis.
Figure 4: Source: Honeywell
Figure 4 shows what their architecture looks like today.
They no longer use flat file interfaces into their data warehouse
All SAP applications use SLT to replicate tables “as-is” – such as MARA tables – MARA tables in ECC look the same in HANA. There is no ETL process.
They use Data Services to pull in non-SAP data to harmonize data.
SAP and non-SAP data is used to create analytic views. Some reports go against HANA views and the others go against the UNX universe.
It is a better architecture as they have fewer ETL tools and fewer transformations and a common skill set.
Figure 5: Source: Honeywell
Figure 5 shows what Self-Service BI
They want to provide enterprise wise governed data. He emphasized governed so they get the same answers.
They want no IT involvement after the initial data delivery.
The business then handles it themselves and they don’t have to create a request for a report.
A governed data set provides a single set of truth and allows decisions being made on information.
Figure 6: Source: Honeywell
Figure 6 shows IT responsibilities – they will do ETL, any data modeling in HANA, information spaces for Explorer, and universes.
Figure 6 shows Honeywell’s tool decision construct. Key performance indicators are included in Dashboards and Design Studio
The middle layer of the pyramid is the “pealing back” including Analysis, Lumira and Explorer.
The details are for the individual contributor includes focused details of where there might be issues.
Figure 7: Source: Honeywell
Figure 7 shows Business Users tasks at Honeywell, which are lower in complexity.
Business users create Explorer View sets to create dashboards
They use Analysis OLAP and Lumira.
Web Intelligence is used against universes.
Figure 8: Source: Honeywell
Figure 8 shows their first business case, working with Finance on a predictability model.
Some of this is custom.
All this data was acquired and created in 30 minutes!
Figure 9: Source: Honeywell
Figure 9 shows their anticipated revenue in a few years and by quarter.
It says what could but not by business segments or classifications
Figure 10: Source: SAP
They dropped in business segment and now they have a stacked bar.
Segment 1 in blue has the most revenue.
Segments 2 and 3 are more linear, while segment 1 is cyclical.
Figure 11: Source: Honeywell
Figure 11 shows dropping in region
Figure 12: Source: Honeywell
Figure 12 shows contributions by business segment in a pie chart
Figure 13: Source: Honeywell
One region, the Americas has the greatest cost every year
APAC has the best margins
Americas is next with the best margin
Figure 14: Source: Honeywell
Each business segment shows the “heat”
He said you can’t format legend on bubble chart
Figure 15: Source: Honeywell
Figure 15 shows the keys to success. You need to define roles and responsibilities. You create models with business in mind; they decide on the key metrics.
They found JAD to be an iterative process and the end user has “skin in the game”.
Figure 16: Source: SAP
He said there was very little training for “data junkies” with a 30 minute session with Lumira – the data junkies “ran with it”.
Custom hierarchies cannot be used in Lumira
Universes should have pre-defined prompts. You cannot filter in Lumira until after you acquire it.
You can use built in hierarchies such as time and region
With BusinessObjects Explorer, he said 12-15 facets are recommended
They have used up to 25 facets in Lumira
Figure 17: Source: Honeywell
He has 16 GB memory without too many problems
Freehand SQL has some difficulties – you have to login for each query and serial refresh fashion
Our thanks again to Dennis for his time and sharing his company’s story.