- Introducing AllScale
- Why People
- Why Organizations
- Why Management and Leadership
- Conclusion
This chapter is from the book
This chapter is from the book
This chapter is from the book
Why Organizations
It should follow that if people are important to the scalability of a system, their organizational structure should also be important. If this isn't intuitively obvious, we offer a few things to consider regarding how organizational structure and responsibilities can positively or negatively impact your ability to scale a system.
An important concept to remember when considering organizational design as it relates to scale or any situation is that there rarely is a single right or wrong organizational structure. Once again, this is an art and not really a science. Each organizational structure carries with it pros and cons or benefits and drawbacks relative to the goals you wish to achieve. It's important when considering options on how to structure your organization to tease out the implicit and explicit benefits and drawbacks of the organizational design relative to your specific needs.
Some questions you should ask yourself when developing your organizational design are
- How easily can I add or remove people to/from this organization? Do I need to add them in groups, or can I add individual people?
- Does the organizational structure help or hinder the development of metrics that will help measure work done by the organization?
- How easily is this organization understood by the internal and external stakeholders of the organization (i.e., my customers, clients, vendors, etc.)?
- How does this organizational structure minimize the amount of work I lose on a per person basis as I add people to the organization?
- What conflicts will arise within the organizational structure as a result of the structure and how will those conflicts hinder the accomplishment of my organization's mission?
- Does work flow easily through the organization or is it easily contained within a portion of the organization?
These aren't the only questions one should ask when considering organizational structure, but each has a very real impact to the scalability of the organization. The question of how easily people are added is an obvious one as it is very difficult to significantly increase the amount of work done by an organization if your organizational structure does not allow the addition of people to perform additional or different types of work. Additionally, you want the flexibility of adding people incrementally rather than in large groups and the flexibility of easily removing people as market situations demand, such as a sudden increase in demands on the company or a market recession requiring constriction of expenses.
The question regarding metrics is important because while you often need to be able to scale an organization in size, you also want to ensure that you are measuring the output of both the organization and the individual people within the organization. An important point to remember here is that as you add people, although the total output of the team increases, the average output per person tends to go down slightly. This is the expected result of the overhead associated with communication between people to accomplish their tasks. Each person can only work so many hours in a day and certainly no more than 24. If an organization consisting of a single person were to work the maximum possible hours in a day, constrained either by law or exhaustion, doing his or her primary task and absolutely nothing else, it stands to reason that the same person when required to interface with other people will have less time to accomplish his or her primary task and as a result produce less in the same amount of time. Therefore, the more people with whom an individual needs to interface to complete any given task, the more time it will take for that person to complete that task as increasing amounts of time are spent interfacing and decreasing amounts of time are spent performing the task.
The way to envision this mathematically is that if a single person can produce 1.0 unit of work in a given timeframe, a two-person organization might produce 1.99 units of the same work in the same timeframe. Each person's output was slightly reduced and while the team produced more overall, each person produced slightly less on an individual basis. The resulting relative loss of .01 units of work in the aforementioned timeframe represents the inefficiencies caused by coordination and communication. We will cover this concept in more detail in Chapter 3, Designing Organizations, where we discuss team size and how it impacts productivity, morale, and customer relations.
If the structure of your organization is such that it disallows or makes difficult the establishment of measurements on individual performance, you will not be able to measure output. If you cannot measure the output of individuals and organizations, you can't react to sudden and rapid deteriorations in that output resulting from an increase in size of the organization or a change in organizational structure.
"How easily is this organization understood by the internal and external stakeholders of the organization" addresses the need for intuitive organizational constructs. Written another way, this question becomes "Are you aligned with your stakeholders or do you waste time getting requests from stakeholders to the right teams?" If you want an organization to scale well and easily, you don't want the external teams with which you interface (your customers, vendors, partners, etc.) to be scratching their heads trying to figuring out with whom they need to speak. Worse yet, you don't want to be spending a great deal of time trying to figure out how to parcel work out to the right groups based on some stakeholder request or need. This might mean that you need to develop teams within your organization to handle external communication or it might mean that teams are developed around stakeholder interests and needs so that each external interface only works with a single team.
We discussed the question of "How does this organization structure minimize the amount of work I lose on a per person basis as I add people to the organization?" within our explanation of our question on metrics. You might have been in organizations where you receive hundreds of internal emails a day and potentially dozens of meeting invites/requests a week. If you've been in such a situation, you've no doubt spent time just to eliminate the emails and requests that aren't relevant to your job responsibilities. This is a perfect example of how as you add people, the output of each individual within an organization goes down (refer back to our example of one person producing 1.0 unit of work and 2 producing 1.99 units of work). In the preceding example, as you add people, the email volume grows and time dedicated to reading and discarding irrelevant emails goes up. Figure 1.1 is a depiction of an engineering team attempting to coordinate and communicate and Table 1.1 shows the increase in overall output, but the decrease in individual output between an organization of three individuals and an organization consisting of one individual. In Table 1.1, we show an individual loss of productivity due to communication and coordination of .005, which represents 2.4 minutes a day of coordination activity in an 8-hour day. This isn't a lot of time, and most of us intuitively would expect that three people working on the same project will spend at least 2.4 minutes a day coordinating their activities even with a manager! One person on the other hand need not perform this coordination. So, as individual productivity drops, the team output still increases.
)
Figure 1.1 Coordination Steals Individual Productivity
Table 1.1. Individual Loss of Productivity as Team Size Increases
|
Organization Size |
Communication and Coordination Cost |
Individual Productivity |
Organization Productivity |
|
1 |
0 |
1 |
1 |
|
3 |
0.005 |
0.995 |
2.985 |
You can offset but not completely eliminate this deterioration in a number of ways. One possibility is to add management to limit interpersonal coordination. Another possibility is to limit the interactions between individuals by creating smaller self-sufficient teams. Both of these approaches have benefits and drawbacks that we will discuss in Chapter 3. Many other approaches are possible and anything that increases individual throughput without damaging innovation should be considered.
Another important point in organizational design and structure is that anywhere you create organizational or team boundaries, you create organizational and team conflict. The question "What conflicts will arise within the organizational structure as a result of the structure and how will those conflicts hinder the accomplishment of my organization's mission?" attempts to address this problem, but there is really no way around boundaries causing friction. Your goal then should be to minimize the conflict created by organizational boundaries. The greatest conflict tends to be created when you have organizations with divergent missions, measurements, and goals, and an easy fix to this drawback is to ensure that every organization shares some set of core goals that drive their behaviors. We'll discuss this in more detail in Chapter 3 where we will cover the two basic types of organizational structures and what purposes they serve.
"Does work flow easily through the organization or is it easily contained within a portion of the organization?" is meant to focus on the suitability of your organizational design to the type of work you do. Does work flow through your organization as efficiently as a well-defined assembly line? Does the type of work you do lend itself easily to a pipeline, where one team can start its work at a predefined place marked by where another team completes its work without a lot of communication overhead? Or is the work largely custom and highly intellectual, requiring a single team to work on it from start to finish without interruption? Are the components of what you build or produce capable of operating through a well-defined interface such that two teams can work on subcomponents at the same time?
Let's take a look at our company, AllScale. AllScale recognizes that it has a need to scale the number of people within the engineering team that is supporting the HRM software in order to produce more products. Over the course of the last year, AllScale has added several engineers and now has a total of three managers and ten engineers. Each of the three managers reports to the chief technology officer (CTO) of AllScale. These engineers are broken down into the following teams:
- Two engineers responsible for the provisioning of systems, networking devices, databases, etc. for AllScale's HRM product. This is the Operations team.
- Six engineers responsible for developing the applications that make revenue for AllScale's HRM product. This is the Engineering team.
- Two engineers responsible for testing AllScale's HRM product for defects and other quality related issues. This is the QA team.
)
Figure 1.2 AllScale Org Chart
At a high level, we can intuit a few things from the structure of this organization. The designer of the organization believes that the separation into teams by skill set or functional job responsibility will not have an adverse impact on his or her ability to develop and launch new product functionality. The designer evidently sees great value in dedicating a group of people to testing the product to ensure it conforms to the company's quality standards. Benefits we would expect from such an organization are the ability to recruit top talent with focused skill sets such as software engineering in one or more programming languages, hardware/infrastructure experience, and quality/testing experience. At a high level, it appears that we should be able to relatively easily add engineers, operations/infrastructure engineers, and quality assurance engineers—at least until a manager is saturated with direct reports. This organization should be easily understood by all of the stakeholders as it is structured by relatively easily understood skills. Finally, work would seem to be able to flow easily between the organizations as we should be able to define measurable criteria that will qualify any given work product as being "ready" for the next phase of work. For instance, code might be ready for QA after it has passed a peer review and all unit testing is completed, and it might be ready for launching to the site and the Operations team after all priority one bugs are fixed and at least 90% of all other defects found in the first pass are resolved.
There are some potential drawbacks of such an organizational structure, however. For instance, how are you going to measure the throughput of the teams? Who is responsible for causing a slowdown of new initiative (feature or product) development? Will you measure your operations/infrastructure team by how many new features are launched to the site; if not, what keeps them from slowing down feature development in an attempt to increase a metric they will likely covet such as availability? When do you determine that something is "completed" for the purposes of measuring your engineering throughput? Is it when the feature launches live to site and if so, have you calculated the bug induced rework time in developing the feature?
Will the structure minimize the work loss on a per person basis as you grow the team? To know this, we probably need to dig into exactly how the software engineers are structured but we can probably also guess that coordination across teams is going to be a source of some work. Who will perform this coordination? Are the managers responsible for shepherding something from engineering to QA (Quality Assurance) and finally into the production team (Operations)? Who is responsible for setting the guidelines and criteria for when something moves from one place to another? Should you create a project management team responsible for helping to do this or should you instead reorganize your teams into self-contained teams that have all the skill sets necessary to complete any given task?
There are likely to be a great many conflicts in this proposed structure, many of them across the organizational boundaries we've defined. Operations will likely have concerns over the quality of new code or systems deployed, QA is likely to have concerns over the level of quality initially presented to them by Engineering, and Engineering will complain that Operations does not meet their needs quickly enough with respect to the creation of new systems, installation of new databases, and provisioning of new network devices. Who will be responsible for helping to resolve these conflicts, as each conflict takes time away from doing "real work."
Other, larger questions we might have of such an organizational structure might be "Who is responsible for ensuring that the product or platform has an appropriate level of scale for our needs?" or "Who is responsible for identifying and resolving issues of scale?" When considering the answer to this question, please note that a scale issue might be the result of a network capacity constraint, a database capacity constraint, or a software capacity constraint. Moreover, that constraint isn't going to be easily bucketed into one of these areas every time it comes up.