Thursday, October 18, 2012

Snowman Architecture Part Two: Economic Benefits

This is the second part of a four part blog about The Snowman Architecture. The first part was The Snowman Architecture: An Overview. In this blog, I will be discussing the economic benefits of the architecture. But don't read this until you have read the overview!

In the next installment (part three) I will discuss The Technical Benefits of the Snowman Architecture. The fourth part, by the way, will be The Criticisms, in which I will describe the many criticisms of the Snowman Architecture and why they are all wrong.

Originally I had planned to cover all of the benefits (economic and technical) in one blog. It turns out there are just too many benefits for one blog so I have had to separate them into those that are more economic in nature (this blog) and those that are more technical in nature (the next blog.)


The Snowman Architecture breaks down a large IT system into small vertically partitioned subsystems called snowmen. These snowmen interact with each other through asynchronous messages. Snowmen are designed to be as autonomous as possible from each other using a design methodology known as Simple Iterative Partitions1 (SIP). Figure 1 shows an IT system designed using the Snowman Architecture.

Figure 1. Snowman Architecture

The Snowman Architecture is in contrast to a traditional architecture that uses a methodology such as TOGAF2 to create a horizontally partitioned system. Figure 2 shows an IT system designed using traditional methodologies.

Figure 2. Traditional Horizontally Partitioned Architecture

Points of Contrast

There are several contrasts that immediately jump out in comparing the Snowman Architecture to the traditional approach. 

The first contrast is in the orientation of the partitioning. The Snowman Architecture uses a strong vertical orientation to the partitioning. The traditional approach uses a weak horizontal orientation to the partitioning.

The second contrast is in the number of subsets in the partition. The Snowman Architecture supports an unlimited number of vertically oriented subsets (snowmen). The transitional approach has exactly three horizontally oriented subsets (business architecture, technical/SOA architecture, and data architecture.)

The third contrast is in the strength of the partitioning. The strength of the partitioning refers to the porosity of the boundaries separating subsets. The more "stuff" that passes between subsets, the greater the porosity. Porosity weakens the partitions, so the greater the porosity, the weaker the partition. The Snowman Architecture partitioning is strong, indicated by the minimal number of connections between subsets. The traditional horizontal architecture partitioning is weak, indicated by the large number of almost random connections between subsets. 

Economic Benefits of Snowman Architecture

Okay, now that you remember the basic overview, let's look at the economic advantages of The Snowman Architecture.

Benefit 1: Linear Versus Exponential Complexity Curve

As an IT system gets larger it gets more complex. This is because complexity is driven both by the amount of functionality in a system and the number of connections in a system3. Both the Snowman Architecture and the traditional architecture gets more complex as the system increases in size but how they increase in complexity is quite different. The complexity of the tranditional system increases exponentially. The complexity of the Snowman Architecture increases linearly

For small IT systems, the difference between an exponential increase and a linear increase of complexity is not important. But as the size of the IT system exceeds $5M in cost, the difference becomes very important. 

Figure 3 show the relationship between complexity and project size of a traditional versus a Snowman Architecture. 

Figure 3. Complexity of Traditional Architecture versus Snowman Architecture

As shown in Figure 3, the complexity of a traditional IT architecture increases exponentially. It starts low and then enters the Risk Zone (the zone in which project failure is likely) when the size hits someplace around $8M. From there it rapidly ascends into the Failure Zone (the zone in which project failure is certain)4.  

In contrast, the complexity of the Snowman Architecture starts low (as does the traditional architecture) and then increases with a shallow linear slope. There is little difference between a shallow linear line and an exponential slope at low numbers. In Figure 3, you can see that at project sizes under $1M, there is effectively no difference between the Snowman Architecture and the traditional approach.

However this changes quickly as the project size increases. Traditional architectues are already in the Danger Zone by the time they hit $8M and by the time they hit $10M they are in the Failure Zone. In contrast, the shallow linear complexity slope allows the size of the Snowman Architecture to remain comfortably  in the Success Zone until well past $100M in project size. In fact, it isn't even clear that there is a size limitation with the Snowman Architecture.

The bottom line: a traditional architecture becomes likely to fail at around $5M whereas a Snowman Architecture has a high probability of success even at $100M.

Benefit 2: Return on Investment (ROI)

To compare the ROI of the Snowman Architecture versus the traditional horizontally partitioned architecture, let's take some reasonable project numbers for, say, a $20M project. 

Using a traditional architectural methodology (e.g. TOGAF) we can reasonably assume the $20M project will go over budget by at least 200% and will cost an additional 400% in lost opportunity costs5

Using the Snowman Architecture we won't be doing a single $20M project, we will be doing some number of smaller project of at most a few $M each. Projects of this size are well within the Success Zone (as shown in Figure 3.) Projects in this zone typically have no overruns and no lost opportunity costs. 

The Snowman approach requires an additional phase in the project life cycle, a pre-planning phase. This is where most of the work is done to design and plan the snowmen. In the worst case, this phase could add 10% to the overall cost of the project.

Of course, these numbers are just best guesses based on what I have seen of industry data. Feel free to plug in actual numbers from your own projects.  But based on these numbers, we can calculate the Snowman ROI.

Without using the Snowman architecture, we expect a total cost of

   $20M (planned cost)
+ $20M (200% overrun)
+ $40M (lost opportunity costs
$80M (total cost)
With the Snowman architecture we expect a total cost of 

  $20M (planned cost)
+ $2M (10% overhead for Snowman preplanning)
$22M (total cost)

The difference between the two approaches is

  $80M (Cost of traditional approach)
- $22M (Cost of Snowman approach)
  $58M (Difference between approaches)

The ROI of using the Snowman approach is thus

  $58M (Difference in Costs) 
/   $2M (Added cost of Snowman Approach) 
X 100
2900% (Calculated ROI)

The bottom line: the Snowman approach returns a 2900% ROI. A 2900% ROI is excellent by any measure.

Benefit 3: Non Tangible Benefits

There are many benefits to delivering a project on time other than eliminating the lost opportunity costs. It is hard to measure these benefits, but they certainly include the following:

  • Predictability of IT deliverables.
  • Increased trust between Business and IT.
  • Better ability to use IT as a strategic asset.
As you can see, there are compelling reasons favoring the Snowman Architecture over traditional approaches. The reduction in complexity is huge and the ROI would make even the most seasoned CFO salivate  But the most compelling reasons favoring the Snowman Architecture may not be economic, they may be technical. But for those benefits, you must wait for the next installment of this blog.


(1) SIP is a patented methodology for autonomy optimized partitioning. It is described in a number of places, including the web short SIP Methodology for Project Optimization.

(2) TOGAF® is a methodology owned by The Object Management Group. It is described on the TOGAF 9.1 On-Line Documentation.

(3) If you are interested in the mathematical relationship between size, connections, and complexity, see my white paper The Mathematics of IT Simplification.

(4) I have written about the relationship between traditional IT project size and failure rates in a number of places including the web short The Relationship Between IT Project Size and Failure Rates.

(5) Unfortunately, we do not have good data on what these number are world-wide. These particular numbers came from averaging a number of large projects discussed in the Victorian Ombudsman Investigation into ICT Enabled Projects (2011).


Snowman picture by CileSuns92


Harry said...

Very nice article and discussions. However, I wonder where would data that will across multiple snowmen sit? For example, enterprise product information may be used by accounting, marketing, sales, CRM. If the data will be owned by 1 snowman, others will need to access the data via its service. This may not be very effective.

DBJDBJ said...

We need to re-use one snowman. For example to re-use the data it encapsulates. But then the whole IT landscape is back to the oriented graph of inter-connected snowman.
Mail server is a snowman. It contains data, components with logic and components implementing services. Mail servers communicate to each other using asynchronous messaging paradigm. The whole e-mail system is oriented graph of snowman nodes. No vertical or any other partitioning.Still the whole system is very reliable and it works very well, as we know.

What is new in Your methodology ?