5 problems we faced

5 problems we faced

Hello, Habre! We have already written here about the allocation of IT costs. In short, it’s a way to allocate the cost of a company’s IT spending across its business units. We consider how much our IT resources are worth, who is consuming them and how much, monetize these resources and map them to business units. This helps to more accurately plan the budget and control the consumption of IT resources in the company.

In order to somehow formalize and further apply the results of data collection from all IT services consumed by business units, we build a model that describes the cost of the company’s IT services. But the allocation model is just a set of rules and calculation formulas. For convenience, all this logic of calculations should be transferred to the calculator, hardware and software specifications, the cost of support, rent, etc. should be stitched there. As a result of such an exercise, we will receive either a “simple” Excel calculator, or an automation tool, or both.

Calculating the cost of the service is just the beginning. It must be provided, and then invoiced. Allocation is part of the overall billing process. We solved this task through the ManageIQ private cloud. You can read about one of these projects in this post.

I want to tell in more detail about the process of preparing the allocation model. More precisely, what problems await the analyst on this path and how we solved them on our projects.

Problem 1: We don’t know our infrastructure

The description of each IT service begins with the collection of initial data. We interview the employees responsible for the directions, request diagrams, tables and in general any data sources that will help to get an overall picture of the IT service. In general, we conduct a small survey.

Usually, most of the information is stored by engineers in Excel sheets, and to determine a full-fledged explanatory note for the system is a huge luck. As a result, when we transform disparate data into a single formula and bring it to the customer, a dialogue like this can easily result:

We: “According to your data, we have described your IT service. Did we understand correctly that you have 10 here, 3 here, and none at all?

Customer engineer: “Yes”.

IT director of the customer: “Wait a second, why do we have 10 here, this one 3, and that one not at all?!”

Head of the IT department: “So you yourself told us to do it that year, and then we had to buy more. But they haven’t bought it yet.”

IT director of the customer: “Stop, colleagues! Let’s figure it out.”

And everyone goes together for a month to figure things out. A month later, they come back with the words: “Yes, that’s what you write: we have 10 here, this one 3, and that one is not there at all.”

Therefore, it is extremely important to conduct a survey before starting the construction of the allocation model, even in express mode. The results of this examination must be agreed with all interested parties.

Problem 2: Perfectionism

The situation can be quite the opposite. The customer has a complete order with documentation, all services are described and the steps are counted. But questions may arise during implementation.

On one of the projects, we prepared a model and a calculator, our work was done. But this calculator was not launched internally for several months. The problem was that Gendyr was not satisfied with the IT department’s explanation of why they had a 12 to 1 oversubscription in the test virtual stand. Gendyr himself is a techie in the past and knows how to ask tricky questions.

Regarding the preparation of the allocation model, the phrase “Better is the enemy of good” is more relevant than anywhere else. If you rely on 100% accuracy of calculations, you will definitely depend on either data collection, or the accuracy of the description of the work of IT services, or on model approvals. Let’s not forget the Pareto principle: “20% of the effort gives 80% of the result.” In our experience, 90% accuracy is more than enough.

Problem 3: Not so simple

Above, I described situations when we receive data from the customer without diving into the infrastructure ourselves. But sometimes, when the customer realizes that his infrastructure is poorly documented, instead of an explanatory note on IT systems, there is only a table in Excel, and an old one, a different approach is needed. In this case, it is necessary to carry out a rather in-depth examination and inventory of IT resources with an analysis of the corresponding downloads of monitoring software, scripts and utilities. And this is a completely different schedule.

In our audit projects, we connect 1-2 engineers for each direction of infra or infosecurity. Each of them collects the discharge according to its field of activity. For example, a virtualization engineer collects RVTools and VMware HealthAnalyzer reports, a Linux engineer collects and analyzes SAS reports, and a Microsoft engineer runs MAPT. Network engineers, IBS engineers and database experts have their own scripts. And this is in addition to the built-in monitoring tools.

I currently have an audit project of medium intensity. There are 30 people in the internal work chat, and almost all of them need answers to questions.

We usually try to agree on a certain number of workplaces on the site and access to all places that interest us. At the introductory meeting, the customer actively nods and promises to promptly solve the problems. In practice, there are not five workplaces, but two, and we will not give you access – we are forbidden by the IT experts. And in general, we are very busy here, and you are distracting us from work.

As a result, many of our designers flood the customer’s IT department with questions, and all scripts and utilities are launched by their sysadmin Seryoga, who already had a lot to do. As a result, the customer begins to slip in their current work, and rarely anyone is ready for this.

Therefore, if the situation develops in this way, it is important at the initial stages to discuss and agree on the approach, requirements and definitely confirm the participation of the customer’s employees in this project in as much detail as possible. Ideally, colleagues need to be shown the importance of their direct involvement in the survey.

Problem 4: Need a good solution here and now

I already mentioned at the beginning of the article that the simplest option for the allocation calculator can be an Excel file. It would seem, well, what else is Excel when we are surrounded by private clouds, DevSecOps practices and heavy Enterprise? Somehow it is not serious.

First, at the beginning of the project, the customer does not need more. Excel functionality is enough to prepare calculation formulas and build the first version of the calculator. About 400 functions are already sewn into it, of which in fact you will need a couple of dozen keys plus some VBA code. And when the calculations mature and the final versions are agreed upon, they can be transferred to the web application.

On one of the projects, it was enough for us to write a small billing program in Python. For the customer, it was easier and cheaper than investing in some expensive solution, when it is not yet completely clear whether the project will be developed or it will stop at the piloting stage. It’s the same with Excel.

Secondly, even if the customer does not have money for the web application, it is quite possible to continue working with Excel. Each of us either uses it on a daily basis or has used it at some point. Therefore, the customer himself will be able, if not to develop, at least to fill the Excel calculator with new data or delete old ones.

And, as they say, the last but not least, Excel is conditionally free for the customer, because it is already available.

Problem 5: We are not robots

The devil, as always, is in the details. It was not yet possible to describe the logic of the work of some IT service in the form of a formula. Especially if the formula contains several large blocks and each one has an explanation of what it consists of.

A model for calculating the cost of a virtual machine

But when the formula is transferred to real data scattered on different sheets of an Excel book, the customer may have difficulties with reading it. For example, below is a formula for calculating the cost of backup from one of our projects.

Yes, it’s all one formula in one cell. Moreover, this is also an option with the reduction of values ​​and the removal of common factors in parentheses.

Here, the first block is a “preamble” that blocks the calculation until all the necessary parameters are selected. Other blocks are cost calculation formulas depending on the backup policy selected by the user.

It took me three months and eight meetings to agree on the finished formula with the customer’s architect and SRK engineer. First, on the same sheet, I tried to show what each block of the formula consists of, how I reduced the values ​​in each block and bracketed the common factors, and then how they all add up to the target formula. Then he prepared a separate presentation with comments and step-by-step transformations. At two meetings, we practically colored cells in the same Excel to visualize the described processes.

And the point is not that the customer’s employees are problematic. No, they are competent specialists and know their infrastructure. The problem is solely in several layers of abstractions that inevitably arise when you try to turn informalized processes into a formula that, moreover, evolves as it is filled. And also because few people are used to reading large formulas in Excel. Somewhere they are a little simpler, somewhere a little more complicated, but in general they are always difficult for an untrained person to understand.

Let’s summarize briefly:

  1. Know your infrastructure. Before starting the description of the infra, you need to study it well. It is best to do a preliminary survey/inventory and record the results in a report.

  2. The best is the enemy of the good. 90% accuracy in the description of IT processes and infrastructure is more than enough, otherwise the project terms tend to infinity, and with them the budget.

  3. If you dare to write a model and an allocation calculator, prepare yourself for the fact that this is a full-time job. It will not be possible to do this in parallel with current tasks, allocating 2-3 hours in the evening.

  4. Initially, it is not necessary to invest in expensive or complex data development and visualization tools. Start with what you have, Excel is quite an option.

  5. We are not artificial intelligence to read formulas easily. Detailed step-by-step explanations of calculations will allow you to understand what is happening and move faster with the project.

Building an allocation model is a difficult, long and costly process. But it’s worth it: at the exit, you will receive clear models for calculating the cost of your IT resources and you will be able to determine how much each of your divisions spends on IT. Based on these calculations, you will be able to implement full-fledged billing in your IT infrastructure, if it is implemented according to the private cloud model. We will talk about the experience of such projects in the following articles.

Andriy Filinov

Consultant of the IT audit and consulting department

Related posts