How to prove anything with TCO
Translation of “How to lie with cost models” by Brian Madden.
The article was written in 2009, when the VDI hype was emerging, so examples are given specifically from this area. But without exception, all techniques are applicable in any other field.
The desire to save money is one of the key reasons to virtualize desktops. In this case, they usually start writing something similar to a feasibility study (feasibility study), in which some analytics justify the financial effect of the project.
Any engineer who has been in the industry for any time has seen such justifications. Consultants, vendors, integrators, and even your boss have them. Moreover, the degree of complexity of calculations varies from a few lines on a napkin to multi-page Excel tables that can make mathematicians dizzy. And although it is quite rare that such justifications are the key reason for decision-making, it can still be argued that they do play some role.
Summary (albeit scary): when you create a cost model to justify a desktop virtualization project, there is a 100% guarantee that you will be able to tweak the model to justify your chosen solution. Anyone, no problem.
You can take the same company, the same input parameters and create two models: “VDI is an ideal project”, and “There is no project worse than VDI”.
All you need is a little creativity. So here are five basic techniques for creating useful cost models that prove anything you choose.
Technique 1. Pack up soft costs
approx. this is the correct translation, namely soft costs, not software costs
Adding soft costs is the easiest way to get the opportunity to swing a project in any direction you choose. By “soft costs” we mean total costs leading to an outcome that cannot be directly measured in money. For example, such as “employee productivity” or “user satisfaction”.
Of course, since we are talking about cost models, they need to be calculated somehow in order to be included in your table. And how to count them?
To calculate “employee productivity”, logic often suggests the following way:
if we have 3000 calls to technical support per year
15% technical support for desktops
each appeal costs 5,000 rubles (a figure out of thin air, which can also be justified in 30 different ways and swung in any direction)
total cost 3000 * 15% * 5000 = 2,250k rub
This means that if we reduce requests to zero, we will save 2 million 250 thousand rubles per year! And this is just a basic example of “soft cost” manipulation.
You can raise the level of manipulation by adding “lost productivity.”
Let be:
average employee cost 1000 rub/hour
Desktop virtualization will save an average of 2 hours per year
1000 employees
Voila, and another 2 million rubles per year! Of course, in the real world, it is not at all a fact that employees will work in these two saved hours or will work 2 hours more than usual. In the real world, anything can happen, for example, staying 2 unpaid hours in the evening to work on your report (which could not be completed due to a broken desktop).
You can add a third level of manipulation, loading the cost of a working hour. Add taxes, bonuses, training, the cost of electricity and the workplace itself, and administrative costs. Why not add “lost profit” at the same time?
Technique 2: Move trackable costs to non-tracked costs
The easiest way to remove unpleasant costs from a project that break its harmony is to stop taking them into account altogether. Those. move to the block that is not taken into account by your cost model. The first candidates for this are the cost of electricity and cooling. Desktop virtualization allows employees to work from home, which means less electricity is needed in the office itself, and less electricity is needed for heating/cooling the office, which leads to an overall positive economic effect for the company.
Of course, there is a small flaw in this wonderful scheme; the costs themselves have not gone away, it’s just that the employees now pay for them themselves, working from home. You simply pass these costs on to the employees in your model and remove them from yours. You can bet that the actual costs will be lower because employees no longer have to pay for travel, cafeteria meals, or even office clothes. But our task is not to calculate the general theory of everything, but only to prove why desktop virtualization is beneficial for the company, so we will not take all this into account. Also see Technique 5.
Advanced liar consultants can combine Technique 2 with Technique 1 – adding the cost of lost productivity due to lack of face-to-face meetings and discussions.
Technique 3. Make up numbers if you can’t prove them.
Some time ago I had a meeting where the issue of “greenness” of VDI was discussed, whether VDI is more environmentally friendly than regular desktops. One of the key points in the discussion was that the device does not always work as it is written on it. In other words, if the server has two 800 W power supplies, this does not mean that the server consumes 1600 W from the moment it is turned on, it just means that the maximum possible consumption can be UP TO 1600 W, and the real one depends on the server configuration ( how powerful the processors are, how much memory, how many disks).
How can you find out how much each server actually consumes? Of course, there are servers that can measure their consumption, UPS, etc. But in most cases they don't, and most people don't think about electrical power at all. So if you want to kill a project with new servers on the basis of “they consume too much” – just come up with numbers to calculate monthly costs. Even if you are wrong, who will know about it? It is unlikely that an engineering representative will chase you with the correct energy consumption report.
This is what makes Technique 3 different from the fake numbers from Technique 1 or the swept-under-the-rug numbers from Technique 2. Here we know that there is a cost and it’s real, but we can’t (don’t want) to measure it. Therefore, we’ll simply throw it out / come up with something convenient for us.
Technique 4. Justify savings with chips that you know you will never use
Cost models are complex, complex calculations with dozens of components and formulas, some easier to justify than others.
“How much does this server cost and how many users does it support?” is generally a simpler and clearer calculation than “How much more productivity will I get from happy users?”
So what do you do if there is “real” data that gets in the way? It’s not a problem at all to make them work for you! If you have data that helps justify it, add it to the cost model, even if you know for sure that it will never be applicable in your project. The more “self-evident” and “universally accepted” points, the better!
Does your hypervisor have the ability to oversubscribe memory so that you can have more virtual machines per server? Great, add it to the cost model! Even if you never use it in production and tear off the hands of anyone who tries it!
Do you have a protocol that consumes half the bandwidth of a competitor's protocol that you don't want to use in your project? Great, add it to the cost model! And it doesn’t matter that you have a gigabit local network to your workplace, which is essentially “unlimited bandwidth” within the project.
Technique 5. Ignore inconvenient numbers
If there's an important message to this lying technique: if the data/numbers don't support your point, just ignore them!
Considering that there is no generally accepted theory for constructing cost models that dictates what must be included, everything is simple. You can safely ignore any inconvenient facts and figures that interfere with your project. To the point that, according to your model, the company can and will definitely go bankrupt in the near future if it chooses the wrong desktop virtualization product that you offer.
Remember: numbers lie, and liars show numbers. So get out your Excel and start your cost model!
PS If you know any other techniques, feel free to add them in the comments.
