Saturday, 29 August 2009

A Novel Way of Dealing with Excel Models

Excel is both heaven and hell in the business world. It's a great calculator, modeler, reporter, you name it. With this incredible flexibility comes a trade off, your data is stored in multiple locations throughout the company in an unstructured and often unsecured manner. This data is not visible to the Enterprise and in most cases is not integrated into your reporting stack. Loads of capturing and data copying results in huge inefficiencies and in many cases errors.

My current employer embarked on a project 18 months ago to "black box" a number of these high profile models. Without going into the gory details let's just say that it did not work. I joined them in April of this year when my family and I made the move to London and was tasked with firstly evaluating the models to give an opinion on whether they were fit for purpose and could they be supported into the future. Sadly they could not and it was my unfortunate responsibility to stop the bleeding and bring the project to a conclusion. That was the easy part, the hard part is what do you do now as the original issues around the Excel Models were still relevant.

The obvious choice would be to take a look at the short listed products again that were selected at great expense and find the next suitable candidate. I was not comfortable with this approach and we did not want to try and rebuild them in PerformancePoint as all development for the product has been discontinued by Microsoft. (Don't even get me started on this subject, I have bitten my tongue since they shook the market with this announcement in January 2009).

I have come up with a rather "out of the box" approach which my team is in the process of designing at the moment. We needed to achieve a few objectives from the project being:
  • The models should be flexible
  • They should have access control
  • Changes should follow appropriate change control
  • They should be auditable
  • The data should be stored centrally and available to other business systems
Package these requirements up with the new Solvency II requirements and SOX etc, and you have quite a tall order.

Off course the most flexible model is still Excel. So what if you could keep Excel and also apply security, access control, change control, auditing etc. This led me into a journey of trying to find software that would give us this functionality. And guess what, it does exist, albeit a software concept in its infancy. Gartner have even released a paper on it (March 11, 2008) under the heading "MarketScope for Spreadsheet Control Products, 2008". It was quite comforting to find out that we had made contact with all the major players prior to this report and that our findings were in line with Gartners. Here are the major players in this space:
They all have a presence in London. Cimcon and Prodiance are US based suppliers whereas ClusterSeven and Finsbury are UK based.

Now they all do pretty much the same thing but in slightly different ways. Essentially each product has 3 major functionality groups:
  • Discovery and Risk Assessment - Basically trolling through all your locations looking for spreadsheets, categorising them and creating a baseline.
  • Monitoring - Keeping track of your spreadsheets and keeping a full audit trail of all changes and versions.
  • Development Tools - Some are add-ins, others applications that assist in developing your spreadsheets to minimise risk, errors etc.
So in a nutshell you can see who did what to which spreadsheet over time and set up alerts and workflows etc. You will be able to see broken links between spreadsheets and a whole host of interesting things that are out of the scope of this paper. We are in the process of getting RFP's from the suppliers and although I have some opinions will reserve them until a later date.

This piece of software should take care of most of the issues surrounding the Excel Models. Now for centrally storing the data and making it available to the rest of the business. My team is in the process of requirements gathering for a framework that will allow us to upload many different data sets from these models into a SQL Server database and thus make it available to our reporting layer. This may sound quite simple but to build a framework that is scalable. supportable and flexible is no easy task.

Watch this space!

- Paul Steynberg

No comments: