When I wrote my first blog on our upgrade to NetWeaver BW 7.3, I had expected this to be a simple technical upgrade with minimal business impact, or impact on our in-flight Project Santa. Now admittedly I would not normally recommend a customer to go live with the very first support package of a new SAP Ramp-Up product (SAP’s product market introduction mechanism, to a small number of customers). Wait for SP01 at least.
And indeed we were so early that we were by the first people to hit most of the kind of problems you only find in a productive environment. We knew this would be the case, and we thought we were up for the challenge.
The power of regression testing
Now we had a set of test scripts for our environment, which involved running a report suite and some data loads. Unfortunately it turns out whilst these are representative, they are not exhaustive. And it also turns out that BW 7.3 had a lot of weird regression bugs which only came out when we got into the depths of detail.
For example if you export a Web Report to MS Excel, the numbers are saved as text in Excel and you can’t easily manipulate them.
I also get the sense that SAP were still running internal regression tests when they released SP01 because a lot of the things that we ran into were already fixes that were in the proposed SP02 – which has only now been released in late February – 2 months after the product launch. I think they should have delayed launch by a few weeks to get these fixes in, because they then had to fit into the release schedule for SP02.
Who needs sales data anyhow?
It turns out that our regression tests didn’t cover our Sales or Support Services data loads. Which is a bit of a pain, considering that our Commercial Director Michael Eldridge quite likes to look at our pipeline. Actually our BPS integration for commercial planning was also broken and we hadn’t noticed, so SAP CRM Opportunity Planning and therefore our sales funnel were non-operational.
I actually think there was a silver lining in this cloud because we had data problems with this area for a while – and this made us get to the bottom of a few niggling data problems in our CRM reports. But, it required some pretty swift fixes from SAP – which mostly seemed to be down to problems with BW Business Content after the upgrade. A bunch of reactivations of characteristics fixed a lot of problems.
BI Java goes badly wrong
We got really caught out with BI Java – reports ran fine in Development but not at all in Production. It turns out that they split out some components between SP00 and SP01 – in BI Base Services specifically. And you have to apply some additional patches manually. It took us and SAP 2 weeks to figure this out and we had to connect Production back to the Development portal in the meantime.
So what’s actually new?
The main thing for me is Portal 7.3 from a user perspective. It’s so much nicer, faster to start, more usable, faster to use and just generally better. The BI Java components have been tweaked and tuned and it really feels like a new system. Plus from a technical perspective there are a load of improvements from modelling to administration and performance. It’s not going to change the world, but all the little things add up.
Developers are much happier with the new Planning Modeller and there are lots of nice usability tweaks for monitoring tasks. We’re told it comes of age with HANA, but the details of that are yet to be released.
For me the big lesson learnt was how important complete test scripts are. These test scripts had served us fine for things like Support Package Stacks and Enhancement Pack 1, but it’s just not enough for a full upgrade. You need a map of what BW is used for and who the business owners are for each area, and they need to select a set of reports to run from their area. And in the context of a customer project we would always have this, but for our own project we thought we could get away with it.
I think the other lesson learnt is to do what we always do on customer projects and do a Proof of Concept (PoC) for a major upgrade. This way you can get SAP bugs fixed outside the critical path. But of course we wanted to Go Live in a specific timeline and there just wasn’t time…
Check out parts one and two of this blog series.