A brief excerpt from the executive summary of my Vlerick EMC BPM thesis.

Using ABC to design a successful business model

Managing the business becomes easier, less stressful and more profitable.

In a startup company, it is very important to know what comes in and what goes out. Fluctuations in revenue create great pressure on running a tight ship.

We need to be able to manage our costs and apply this information to our strategy and business models. XAOP is still looking for a solid and sustainable business model. Traditional accounting mechanisms do not give us the necessary insight into the real cost of our activities. I have tested Activity Based Costing (ABC) in process and operational management (short term) and also on Business Models for strategy (long term).

bm-abc

In this project we investigated the impact of ABC on a set of different business models that we are experimenting with. XAOP has at least three business models today. The first business model is the software development service, the second (Cloudring) and third (Timetag.it) are product oriented software as a service business models. There is of course an overlap between the different models. Our investment in research can be controlled via ABC and assigned to the correct projects (using a service model or product models).

I learned during the project that ABC is not only valuable for large organizations but that also small software companies like XAOP can benefit from its modern approach to track activity performance with a financial metric.

This work defines the direction that we will take in developing and selling our products and services from a financial perspective for the next two years. This was missing before in our business model and I will highlight the benefits of doing this. But there is also a downside: creating and keeping an ABC model up to date demands a lot of work.

ABC is dependent on gathering data and capturing information on processes and activities in order to give an accurate view. The XAOP ABC application is being run on a dataset of approximately one year’s worth of detailed information. This project has been a catalyst for capturing activities (with Timetag.it) in a more structured way than before.

Excerpt Chapter 2 - Section 2.5.1

Cloud computing services

Most companies are moving towards using a new kind of technology with regards to their servers. They make use of virtualization to run multiple servers on one physical machine. This may seem a bit like magic, but it brings flexibility and cost efficiency by making better use of the capacity of the hardware. Not only does the cost of the hardware/machines go down, but the cost of the energy to power these machines is also reduced.

The question is, does it become more difficult with virtualization to measure server usage between the different projects?

What we have done at XAOP is create specific virtual machines for each project. This allows us to create a modular and separated infrastructure. The administration of these machines is simplified and automated because they all use one standardized base. It is like copying or cloning an existing machine. The new hype around Cloud computing is fueled by virtualization technology. At XAOP we can start in two minutes 100 servers to compute a very big problem in the cloud. We use cloud providers such as Amazon and Heroku (part of Salesforce.com) to accomplish this.

There is a lot of literature about applying ABC in the manufacturing industry. I could not find an equivalent example of the true cost of calculating an IT service. But I do think it is interesting to apply ABC to the IT software as a service business (SAAS) model. We looked into costs such as :

  • asset utilization

  • hardware costs

  • power efficiency

  • enabling redundancy

  • security

  • supply chain management

  • personnel

At XAOP we have had the following experience :

  • When we started analyzing the activities and process of using a cloud service, we discovered that the customer channel and purchase were different and a lot easier compared with buying a traditional server. You only need a credit card to get going with Amazon or Heroku. There is no need to waste time talking to sales people or requesting an offer because all costs are transparent and published.

  • We need to compare apples to apples if we want to compare activities, there are a lot of indirect costs that are not assigned in comparing the move to cloud computing/SAAS model.

    • In small companies like XAOP there is no budget to hire extra people to manage server procurement (like at our customers).

    • There is no internal technical team that needs to assess and test different solutions.

    • There is no negotiation process for buying machines and infrastructure (data center, cooling, energy etc).

    • Maintenance personnel are not required to keep everything in the data center up and running.

    • There is no need to keep track of licenses and contract requirements.

These are all indirect costs that have to be taken into account. The cost to power a machine even when you do not need it remains a fixed cost.

A typical project at XAOP needs three separate environments which have to be bought and maintained: a development environment, an acceptance & staging environment and a production environment.

With the introduction of virtualization we can now offer our customers the possibility to carry out development and acceptance & staging with a virtualized cloud solution. We learned that Amazon and Heroku are ideally suited for applications that are for transient use (e.g. an application that is only used approximately 9 months a year).

2.5.1. ABC Analysis of XAOP Amazon AWS costs

The costs of our cloud infrastructure at Amazon Web Services can be analyzed as follows.

The direct and indirect costs remain the same, but what we tried to do was to more accurately calculate the overhead costs. In the long run the total amount for overheads remains exactly the same, but what has changed is the allocation of the overhead costs between the different projects.

Let’s take a look at the following sample data. From our AWS Amazon environment we have identified the following costs and cost drivers (without our reserved instances fees) :

Table 2.1. Costs relate to May 2011.

We divided these into the different services that we provide on Amazon EC2. Of the 4099 instance hours, Timetag and clouding only used 1 instance server and XYZ used 2 instance servers.

Table 2.2. Provided Services

Cost CenterCostCost DriverDriver Quantity
EC2 Linux Servers$304Instance hours774h ($0.03) + 3.325h ($0.085) = 4,099h
Database$245Instance hours2.232h ($0.11)
AWS Management / Training$350Hours8h
Support$350 Number of users1000

Now we can compute the cost per cost driver for each of the different cost centers.

Table 2.3. EC2 Linux Servers

ServiceInstance hoursInstance hours DBHoursUsers
Timetag.it1024,757442,66800
Cloudringapp1024,757442,66150
XYZ2049,57442,6650

Table 2.4. Databases (RDS)

ServiceHoursPercentageCost
Timetag.it1024,7525%$76
Cloudringapp1024,7525%$76
XYZ2049,550%$152
TOTAL4099100%$304

To simplify this analysis example, I kept the currency in dollars.

Table 2.5. AWS Management / Training

ServiceHoursPercentageCost
Timetag.it74433,3%$81,66
Cloudringapp74433,3%$81,66
XYZ74433,3%$81,66
TOTAL2232h100%$245
ServiceHoursPercentageCost
Timetag.it2,6633,3%$116,66
Cloudringapp2,6633,3%$116,66
XYZ2,6633,3%$116,66
TOTAL8h100%$350
Tip

The cost driver for support is identified as the number of support users.

Table 2.6. Support

Next we add up the costs and come up with a unit cost for each of these services. We can compute the total cost and average cost per user for both of our products (Timetag and Cloudring) and the XYZ service for the period of May 2011.

The next step is to define our units, find out the cost per unit and the number of users served.

Table 2.7. Cost by unit

ServiceUsersPercentageCost
Timetag.it80080%$280
Cloudringapp15015%$52,5
XYZ505%$17,5
TOTAL1000100%$350
Cost centerTimetag.itCloudringappXYZ
EC2 Linux Servers$76$76$152
Databases(RDS)$81,66$81,66$81,66
AWS Management / Training$116,66$116,66$116,66
Support$280$52,5$17,5
TOTAL Cost$554,32$326,82$367,82
/ number of users80015050
Average cost$0,69$2,1$7,35
Note

This analysis shows that on average for every Timetag user we spend only $0,69.

For the XYZ service we spend $7,35 per user and a total of $367,82 per month.

Based on this information, we need to charge our XYZ service users a lot more than our Timetag users. If we charge them the same amount per hour we will probably lose money.

One possibility would be to get rid of XYZ and only do Timetag. However keeping XYZ should also be considered as it is a strategic service with high visibility at one of our key customers. So there are other factors besides financial ones that come into play with this decision. Nevertheless it costs a lot of money to serve XYZ users.

We renewed our contract with customer X for the XYZ service based on this information.

We can do the same calculation by project for other cloud services such as Github for source control management (with the cost driver being the number of commits/transactions per project).

Pivotal tracker is a project management tool that we use and the cost of this tool can be divided by usage i.e. the number of stories to measure the size of the project.

For Cloudring or Timetag this analysis is valuable if we want to define and convert how much we need to charge in a freemium model (assuming 5% of the users are premium and 95% of users are free).

blog comments powered by Disqus

Entries per category

  1. 9 pages are tagged with documentum
  2. 12 pages are tagged with events
  3. 14 pages are tagged with rails
  4. 32 pages are tagged with ruby
  5. 13 pages are tagged with sharepoint

Recent Comments

Popular Threads