This blog entry is a continuation of my summary of the Reporting & Business Intelligence seminar for Day 2. The first day of the seminar is covered here: SAP Insider: Reporting and Business Intelligence conference – Day 1
I’ve had a chance to talk with some of the other attendees here and the feedback I’m getting is consistent: frustration. It appears that many people came to the seminar to learn how to do reporting and business intelligence with the new BOBJ tools and all they are hearing is a sales pitch for buying new software. Everyone wants to know how to use their current software license to do reporting and BI, but instead they just keep hearing, “Upgrade to our new tools to get the best functionality”. In other words, they feel like they paid a steep admission fee just to get a sales pitch. The people I talked to aren’t real happy about it either.
I think that this feeling stems from the fact that the presenters are only discussing the premium offerings that are available. All the software discussed is in addition to your existing license. In fact, it seemed that every time someone asked about the licensing for one of the products being discussed, the standard answer was that there will be upgrade costs (and the guy sitting next would shake his head and mumble, “It’s always more money.”) The “light” versions (i.e. free) coming out were only mentioned briefly at the very beginning of the first session.
I can’t really blame the presenters for this confusion because the light versions aren’t even available yet, nor is there any solid information about what they will offer us. How can you talk about something that doesn’t exist? In fact, the only place I see the light versions mentioned in the SAP marketing materials are on the long-term roadmap. But everything else on their site only pushes the premium products.
It’s understandable that the seminar presenters can’t talk about the light version, and it’s also understandable why the attendees think that SAP is just trying to sell them new software that they really don’t need. I did my best to explain to everyone I talked to about what Crystal Reports Light will be and that it is a free option that will replace BEx Reporting and Voyager. But that doesn’t help all the other attendees who think that the BOBJ acquistion is giving them a raw deal.
Today’s sessions, just like yesterday, were very informative for me and I learned a lot. Here is a summary of what was covered. Just like yesterday, the presentations were packed with information and the take home materials had even more information to read through back at the hotel. If you find this information useful, you should really look into signing up for the full seminar ( http://www.sap-businessobjects.com/homepage.cfm?usergroup=us )
Session 1: Tools, technologies and best practices to leverage the web for interactive analysis
The morning started off with Bobby Coates doing the first two presentations. The first covered the web analysis offerings. One of the biggest benefits of the BOBJ acquisition is that BOBJ tools are designed to access disparate data sources. The SAP tools can only access SAP data and most companies have various databases that they need to pull data from. The BOBJ tools make this easy and the final result is transparent to the user as to whether the data came from SAP or non-SAP data sources. Training is also more efficient because end-users only need to learn one interface.
The web analysis tools are:
- WebI (Web Intelligence) – Serves the casual user and the professional analyst. Can access any relational database.
- Voyager – Specifically for the professional analyst. Can only access OLAP data sources.
- Pioneer: Not built yet, but its goal is to be a hybrid of WebI and Voyager. Effectively, it will be the next generation of WebI.
When deciding between which tool to use, there are two considerations: the user and the size of the data. If casual users want to do ad-hoc analysis, then you need to use WebI. The size of the data is important as well because WebI caches all the necessary data on the WebI server. This gives fast performance, but you are maxed out around 10,000+ records. If you plan on doing analysis on millions of records, use Voyager because it lets the OLAP server do the heavy lifting and can process these large queries very efficiently.
When using WebI you have to access the BEx queries via a BOBJ universe. This is a semantic layer on top of the complex query structure. The goal is to design the universe using terminology that the end-user is already familiar with. Too many tech people create universes that use the same names in the BEx query and end-users are very confused and can’t understand what it means.
Session 2: A visual guide to designing dashboards
The first order of business was to make sure everyone was on the same page and clarify the differences between dashboards, cockpits and scorecards.
- Dashboards – see my metrics ‘right now’ so you can take immediate corrective action.
- Cockpit – a predefined dashboard with data that may not be as urgent. It’s really the same thing as a dashboard, but it is meant to ‘spice up’ boring/non-critical data to give it more style.
- Scorecards – one of the underused tools. They compare one period to another period. Shows high level metrics that change slowly (employee satisfication, client retention). Can show target/actual/variance %.
Xcelsius is ideal for designing dashboards, but the problem people run into is that they design the entire dashboard at once. When users see it, they either want to update it now or later down the road and you have to make various changes. It’s better to create the indivual elements on the dashboard as separate Xcelsius widgets and then add them to a larger dashboard. This makes it easier to swap out widgets in the future as the requirements change. It also makes it easy for users to drag and drop individual widgets onto their desktop without using a dashboard.
A common problem people have is expecting Xcelsius to do too much work. Since the final widget is a flash based object, it is inherently slow. Thus, you don’t want to have Xcelsius process too much data or perform many calculations. You should design the query so that it aggregates all the data and perform as many calculations as possible. This lets Xcelsius focus on presenting the final resultset in a professional manner.
The biggest tip he gave was to read the Xcelsius Best Practices Guide first. That will save you a lot of headaches and not make the common mistakes. https://boc.sdn.sap.com/node/2120
Session 3: New tools, options and strategies for quey design and optimization
Jenny Shah started out this session with an overview of all the anylytics tols. All the new tools still leverage BEx queries: WebI, Crystal Reports and Voyager. However, WebI needs a universe to as the semantic layer around a BEx query. Ideally, you would create a generic query that has good performance and then build universes which return subsets of the query for specific business groups.
Although Crystal Reports and Voyager can connect directly to the InfoCube, BEx queries are the recommended data source.
You shouldn’t use ODS/DSo objects because they kill performance.
Prior to Crystal Reports 2008, you had to connect to SAP NetWeaver BW via the BW Query Tool. With the release of Crystal Reports 2008, you can now use the MDX driver to connect to SAP. This gives you more optimized data access and also improves upon the type of data that can be reported on. All new development should be using Crystal Reports 2008 with the MDX driver.
Near the end of the presentation, she gave us some excellent planning tools. There were a couple grids showing what you need to be aware of with the Universe Designer and mapping BI elements to OLAP Universe elements. This was also best practices for building universes. This will be very useful on the job.
Sesion 4: Data modeling architecture and landscape design.
I think that David Dixon’s final session of the day was the highlight of the entire seminar. He discussed best practices for integrating BOBJ into your SAP server landscape as well as implementing the BOBJ tools into the BI system architecture. There was so much great information here that he went well over the 1.5 hour presentation time. Even though everyone was pretty much brain-dead at this point of the day, most people stuck around during the extended Q&A session to get more information. I’m certain that the slide deck for this presentation alone will get a lot of use when people get back to the office and start planning their BOBJ implementation.
Let me try to give a few highlights from his presentation (although nothing I write here will do it justice).
BOE and SAP have very different architectures. SAP puts everything into a package and you put that package on a server. BOE introduces the concept of having individual services that can be scaled up or down. You can run all services on the same server and when you start to see performance issues, you can separate the service that is handling the largest load and place it on its own server. As your demands grow, you can continue to monitor the services and scale up individual services when necessary. With SAP, its all or nothing. He gave us an excellent “thumbs up, thumbs down” chart that looks at various factors (performance, administration, security, etc.) and how each one is effected by being either centralized or decentralized.
It’s important to realize that BOE integrates with SAP and not the othe way around. Ideally, you will put BOE on a separate server and it talks to your SAP server(s). He gave us quite a few examples of possilbe server landscapes for different scenarious.
Just like Jenny, David stressed everyone to use the MDX drivers to talk to NetWeaver BI and not use the BW Query drivers anymore. In addtion, BOE XI 3.0 is far superior to XI R2 and should be upgraded to.
There are three ways to access your data:
* BI (via ABAP): This is the recommended way.
* ERP (direct access to the data tables): This is bad b/c you go around SAP security and can’t take advantage of the acclerator.
* Web Services: This is the newest way of accessing data and it is highly recommend as well.
Some of his final slides focused on the best practices for data modeling and gave us critical questions that need to be asked during the planning stages. One of the most important aspects of data modeling is integrating data, which is handled in the staging layer of BI architecture. Although MDM and data federation are emerging architectural paradigms, they are not mainstream yet and should be avoided for now.