Here are the slides from my 2-hour #PowerDesigner presentation to the UK SAP Database & Technology User Group (UKSUG) today

This was an opportunity I could not miss – a 2-hour presentation on modelling to an audience of database experts!
Starting with a brief look at using Visio and/or Excel for data modelling and governance, I talked about the extras we can gain by using PowerDesigner to design databases.
Of course, it’s not ‘just’ databases we’re concerned with, the relationships those databases have with our business and technical architecture is also important.
The next key topic is the role of Data Models and others (such as the Requirements model) in Governance and Design.
Next it’s mapping data sources and targets to demonstrate and create data lineage, showing how PowerDesigner supports multiple DBMS Versions (and what you can do to change how it does that), creating a CUBE for Business Objects, finally (almost) focusing on the support provided for SAP IQ.
Finally, I described some real-world uses of PowerDesigner.

Buying another company? Have you checked the data models? #datamodeling #ITintegration #M&A

There’s an apparently never-ending discussion of the role of data modelling. Is it an IT activity? Is it only about designing databases? My friend Chris Bradley has given numerous sessions on this (you can see one of them here), as have many others, so I don’t want to labour the point. I’d just like to focus on a possibly overlooked role for data models in mergers and acquisitions.

I heard a tale a few years back, of a foreign mortgage provider that bought a UK company that had a comprehensive enterprise data model. This was the first UK company that the foreign company had ever purchased, so I expect that they carried out a lot of research into the new market before deciding how much to pay for the acquisition, and how much effort would be involved in migrating the existing UK IT applications to their own platform. During that research, they ought to have taken a look at the enterprise data model and compared it with their own data models, assuming they had any, but did they?

Here’s where the tale gets interesting. The foreign company was used to a market with a simple view of mortgage accounts; each mortgage account only has a single product, making the interest and repayment calculation processes quite simple:

Mortgages - simple view

Unfortunately, a glance at the UK company’s enterprise data model would have revealed that their business processes and IT systems were much more complicated:

More complex mortgage

This is a simplified view of the data model, so it doesn’t look like there’s much difference at first. Look at the full data model and you’ll see that the UK mortgage account can include multiple products with different:

  • start date
  • maturity date
  • interest rate %
  • interest rate type (interest only or repayment)

The business processes for managing these accounts would be much more complex.

According to the tale, the foreign company had a modular component-based IT platform, and their projected cost savings were dependent on migrating the UK IT functionality to the existing components. Without the knowledge gained from looking at the data model, they may have seriously underestimated the amount of work needed to migrate mortgage accounts.

Within three months of the takeover, the new owners shut down the entire enterprise data modelling organisation – all the people, the data modelling tools, and the metadata repositories. I’ll leave you to draw your own conclusions.


Do we really have Employee master data? #MDM #MasterDataFail #MasterDataSlaves

Sometimes when I’m providing a service to a client, I need to connect to their domain so they set me up as an internal user. This can give me some interesting insights into the way they operate, and it’s amazing how little has changed since the distant time when I was a salaried employee. For example, I recently supported a user department within a very large UK company, and the team moved from office A to office B. We all received an email advising us to update our current location in our email signatures, and also in the following systems:

  • Human Resources (two separate systems)
  • IT’s ordering system
  • the Intranet Directory

What on earth is going on at this company? How much time is wasted by employees carrying out manual Master Data Management? How much variation is there in the location data? Why are the Employees expected to act as Master Data Slaves?

Whatever happened to Information Management at this bank? #DataQualityFail

Yesterday I set up SMS alerts for my business bank account for the first time. When I registered for the service, it showed me the last four digits (9901) of the number they proposed to use for messages advising me of activity on my account. That worried me, as I stopped using that number several years ago when I changed providers, and the bank are well aware of my current number. In fact, they already send me text messages to confirm certain activities via internet banking.

With trepidation, I registered for the service, and there was an option to update my contact details, so I took it. Lo and behold, the ‘9901’ number is not in my contact details; the mobile number is correct.

Here’s my mobile phone number in my “Personal Details” (there is nowhere to record a second mobile number):

my banking mobileSo, I decided to carry on and sign up for a couple of ‘text alerts’. Here are the choices I’m given when setting up a ‘text alert’ – note the first phone number:

two numbers for text alerts

Interesting that, I can choose from a list of mobile numbers, though I’m only allowed to record one mobile phone number.

I was careful to select the correct number, and expected the service to work OK. My first email alert arrived at 02:06 this morning, but no matching SMS arrived. This may have been due to the patchy mobile service I’ve had in the last few days, but I decided to ring the bank anyway. I called at 10:20 and explained the issue to the customer service representative, and she promised to get back to me today.

I asked her for the contact details for their data protection team, as I wanted to raise this concern with them, in case other customers’ alerts were going to the wrong people, and also because this is very obviously a data management problem. I was told that the bank does not allow customers to contact data protection directly, that has to be done via customer service or by making a complaint.

Well, the missing text alert arrived at 11:57 this morning, nearly 10 hours after the email. The time is now 22:20, and I’m still waiting for my call back from the bank. I’m not surprised they haven’t sorted it yet, it’ll take a while to sort out.

I used to work for this bank in an information management capacity, leaving ten years ago when the new owners guillotined the Information Management department. I’ve heard a few things from the inside since then about how effective their information management is these days; this definitely would not have happened in ‘the old days’, there was a business department in place to make sure of that. I wonder what happened to it.

P.S. Why is there a drop-down list of phone numbers? Here’s a suggestion from a friend of mine, which I’ve paraphrased – “during project testing a developer found >1 mobile number, so they decide to add a drop-down list. Not scenario-checked. Unit tested but no end to end test.”  Could be true.

But that’s just experience bitching.

How much can you rely on tool comparisons from tool vendors? #EDW15

The quality and reliability of comparative evaluations issued by vendors of modelling tools varies significantly, from the completely unprofessional to the merely incomplete. I include the comparisons I wrote for Sybase (now SAP) PowerDesigner in 2011 in the latter category.

The worst I’ve seen (very recently) was ‘unofficial’, presumably produced by a sales rep for a particular customer. It was completely unprofessional, the sole intention was to rubbish a competitor. For example,

  • claiming that the other tool doesn’t support feature Y, just because it doesn’t have a feature called Y – in this case, it does support that feature, just happens to give it a different name
  • missing information – “my tool supports both relational and dimensional modelling” – doesn’t mention the fact that the other tool also supports both of them
  • apparent hearsay – throwaway comments such as “they say that my tool can leverage colour better than the other tool” with no supporting information
  • our model comparison feature can compare more objects and properties than the other tool – hmm, really?
  • unanswered questions, such as “Does the other tool support inheritance?”, presumably intended to sow doubt

I think it’s safe to say that the author of this comparison is not actually a user of the tools in question.

It’s very difficult for one person to produce an unbiased and detailed comparison of tools, as very few people know the target tools in sufficient detail. To create unbiased and detailed comparisons you need access to experts in all the tools involved, and you have to ask all the right questions.

Take everything with a huge pinch of salt – take time to come to your own conclusions when you’re choosing a modelling tool.

I’m speaking with my friend Chris Bradley on the topic of Evaluating Data Modelling Tools at EDW in Washington next week, come along and find out more –