Head of Business Analytics & Technology
Bluefin Solutions
Every time you say "Single Global Instance", god kills another kitten
01 Dec 2010
Agile Methodology, Business Intelligence (BI), Business, ERP, Project Management & Methodology
I must confess that this has been troubling me for a while. If you have something – anything – to add then please chime in below. I don’t claim to know the whole story here, or to offer all the answers. So let’s get it out there. What I want to know is: is there any business value to do this, or is it just a CIO’s wet dream? If you're the CIO or CFO or a large enterprise, please feel free to get in touch with me and talk about this privately.
What’s an SGI anyhow?
I’ve spent a substantial amount of time over the last 2 years invested in talking to organizations who are looking to move their IT into a single, big system. A SGI, or Single Global Instance, is the end-state of such an organization change. In SAP land, this means having a single ERP environment which contains information for all markets, countries and consolidation. Or at least most of them. It is often viewed as the nirvana of success of a CIO – consolidating all finance, supply chain and other information into a single information.
The question is, do the benefits of doing this outweigh the negatives that such a behemoth brings? Let’s take a look at that:
Reduced Total Cost of Ownership (TCO)
The types of organization that I’m talking about normally pull in revenues in the $30bn+ category and usually operate in 100+ markets through 4 distinct regions: NA, SA, EMEA & APAC. It’s clear that moving from a position where they have 100+ IT systems to… less than that… will deliver significant reductions in cost. In addition to that, common operating models and process homogenization aid in organizational efficiency and give a platform for future business change.
What troubles me is that I don’t think there’s much difference in cost in operating a system per market, and a single system – in terms of cost per employee. You trade 4 regional systems that have to operate independently on different equipment in different locations, for a single (bigger) system which runs in one location and requires expensive communications infrastructure through the globe. I’d challenge anyone who can show me a real return on investment from 4 down to 1.
Single view of the truth
One thing which is clearly a benefit with SGI is that you only have one product master, one customer master etc. So there’s no confusion between whether you are buying a product in China, or in the UK. Which is great until you realise you probably have different products in different regions anyhow. Made in different factories and with different labels and different ingredients.
This is exacerbated by the fact that for most organizations of this size, they have to invest significant resources into maintaining quality information. Most of these guys have dedicated teams of data people, just cleansing the customer and product data over and over. This is costing them $10m a year or more just in employee resources alone.
I’m not convinced what difference it makes therefore, having to reconcile master data through 4 regional systems. There is clear data ownership that way in the regions, and they can sort their own data out. Consistent process can still be maintained.
Financial consolidation
The one major thing you can’t do with a regional system is to get a global P&L sheet. Who actually cares about this though? Certainly no one in the regional offices do, they just care about their regions. So the only people who care are the people who are involved in head office finance, reporting to shareholders and posting consolidated charters of accounts. Financial Consolidation accountants are a special breed and they are used to manage the often manual process of making things match.
And there’s no problem here in the regional model, because we can easily build a consolidated Global Data Warehouse that sucks information from the regions and presents it in a consolidated format. On top of that we can use various consolidation tools if we need to, so that we can present information neatly to the CFO and shareholders.
Release management
This is where it all falls apart with the SGI. The cost of release management skyrockets when you consolidate outside of regions. Because when I want to make a business change that effects a global process, I have to go to a Global Change Approval Board (CAB). I then agree with representatives from each region (argue, I mean) and get my change through. 3 weeks later I have agreement to put a new field in a business critical report, but because it’s considered to be an invasive change, it gets pushed into the next release.
And here it gets worse. The amount of effort to test things means that you can do far fewer releases. Most organizations with SGIs have 1-2 releases a year, and all change is pushed into this window, which usually represents 4-6 weeks of real “development” per release. Some organizations try to fix this by running parallel IT systems for projects of different types and this just increases the cost of release management. Anecdotally, one global organization told me they spend $2m just building out a release – and often things are pushed out of releases – meaning that releases can happen, with very little change included.
Quality of life
I’ve worked on projects on SGIs and they suck. The 6am calls with Australia and the 10pm calls with California. The nightmare of orchestration and the lost sleep and burnt out consultants. It’s no fun for the people who have to work there and they know it. When will this become an important element for IT departments? Presumably when they can't recruit the best people, because they care about their own wellbeing first. That situation isn't here yet, but it's coming.
Conclusion
I'm not claiming to know all the answers here – far from it – but I really want to encourage discourse on this topic. From everything I have seen and heard in the industry, there is a law of diminishing returns after moving to a regional instance model. And I've not seen any true case for a Single Global Instance. If you think otherwise and would like to chime in, I'd love to hear from you. Or do you think god is just killing kittens?
Comments
John Appleby 28 Mar 2011
Hi,
Actually it really depends on the type of SAP app. ERP isn't too bad provided latency is OK - worldwide latency can be below 250ms provided you have a good MPLS cloud. So this needn't be a restriction.
With some web based apps, there can be a lot of round trips unfortunately, and this can make quite unacceptable usability - BI and CRM in particular.
Regards,
John
Michael Li 28 Mar 2011
Agree.
My biggest concern is actually the performance.
If the SAP backend is hosted in USA and the users are in Asia, Will the SAP UI be very slow?
Anand Gupta 28 Feb 2011
True, but I think it depends on the scope of the project, the organization structure and how the program affects the strategic bits. In marketing driven company if operations are undergoing transformation, probably CEO would not bother. However would the chief be hands off, should it hit its customer experience?
PS: CEO's Office.
John Appleby 28 Feb 2011
Hi,
Thanks for the comment - interesting that you think it's a CEO agenda. I'd tend to think that CEOs don't really care about that stuff - they just want the numbers.
I think in the end, it's one of those things that may prove to be cyclical. We'll see but there aren't that many single instances out there yet - it may come back full circle and become regionalised.
John
Anand Gupta 26 Feb 2011
John I read with interest your blog on Single instance. I share your view. I have also been trying to figure out (check my company name) why big companies choose to fancy single instance. I think the main driver for single instance is the CEO's / Central office desire to have control and visibility of local businesses.The so called chase of single version of truth. I would also assume that this, inspite of senior managers being aware; that an increased central controls correlates well with increased bureaucracy and inflexibility. For example many simple but critical process changes would require central approval, delaying what in many instances could be simple things to do. On the other hand in many case the regions stop following published processes without central office realizing it.
Apart from the business issues, I think single instances are quite problematic even from technical point of view. And I mean in every sense, including technology and functionality. In one of the projects where I worked, we have had single instance for both processes and data-warehousing. It is was tricky to design it and it is turning out to be equally challenging to maintain it.
But I think this quest of Central Controls is here to live its life again. After-all the quest of Central Control, planning and execution is age old quest of leaders. Unfortunately history have not been kind to such quests. Until now, atleast.
Personally I think the biggest issue with Single Instance is the risk arising out of inflexibility to change.
A quote attributed to Darwin states "It is not the strongest of the species that survives, or the most intelligent; it is the one most capable of change"
Maybe I am wrong. But like you I have not been able to find a very convincing arguments in its favour.