Practical Software Estimation: Size, Effort, and Scheduling of Projects
This chapter is from the book
This chapter is from the book
This chapter is from the book
Importance of Size
A person intending to build a house typically estimates the overall size of the house in number of square feet. While buying an office table, you may specify the size as Length x Breadth x Height. Almost every object used in daily life can be sized by using one or more parameters. Because software is "soft," it is always quite difficult to size it as accurately as other material products like a house, a car, or a television. Software professionals traditionally have been measuring the size of software applications by using different methods; Lines-Of-Code (LOC) is the oldest and most popular method used.
Whatever the level of approximation, LOC does give some sense of software size. For example, you can differentiate between two applications developed in the same language that have LOC figures of 20,000 and 50,000. Measurement of software size (in LOC or other units) is as important to a software professional as measurement of a building (in square feet) is to a building contractor. All other derived data, including effort to deliver a software project, delivery schedule, and cost of the project, are based on one of its major input elements: software size.
Key Inputs to Software Sizing
In software estimation parlance, scope of work (also expressed in terms of business functionality provided) is one of the key inputs that determine the size of the final product being delivered. Software estimators sometimes confuse size and effort. Size, in a software development context, is the complete set of business functionalities that the end user gets when the product is deployed and in use. And the person months required to produce the software application of a given size is the effort. For example, in an invoicing application, the user gets the capability to prepare invoices for products sold to customers. The features provided in the invoicing application are equivalent to the functionality available; hence, the size of the application. The amount of time spent by the programmers to design and build the invoicing application is the effort. IT organizations should practice recording metrics of past sizing activities and use the information and experience to size future projects. In his technical paper, "Size Does Matter: Continuous Size Estimating and Tracking," Mike Ross observes, "The secret to making progress in sizing these new environments (software applications) is to identify the unit of human thought in the abstraction being used. Next, the organization must go through a calibration process, starting with projects for which the actual size can be determined in terms of that unit of human thought" 1.
Differentiate Functions from Production Effort/Costs
While negotiating the price of a product, both the buyer and the seller normally assume that the final price includes all aspects of the product; that is, features (functions), look and feel (user interface), quality and reliability (performance), and more. When two products of the same type (family) are being compared, very rarely are products 100 percent identical. To illustrate, compare any two products (for example, mobile phones like Nokia and Samsung, laptops like IBM and Toshiba, or software products like Oracle and Microsoft SQL Server); none of these products are the same in all respects. Similarly, each software project has its own unique requirements, and as such, estimating the cost of the project will also be unique if you adopt the process of assessing effort based on software features, such as quality, scalability, and reliability, which are normally unique for each project.
Estimating the effort and cost of software development projects is perhaps much more complex than estimating the production costs of most consumer products as well as other areas of project execution, whether it involves construction, manufacturing, services, or other elements. Table 7.1 provides some insight into the key differences between a typical product development and software development activities.
Table 7.1. Activity Comparison for Products and Software Development
|
# |
Activity Description |
Typical Product/Service |
Software Development |
|
1 |
Functions/Features Clarity |
Very clear, defined, and definite |
Somewhat vague; can and will change during development phase |
|
2 |
Quality attributes |
Accurately measurable |
Measurable but cost of quality varies based on team skills |
|
3 |
Tools availability |
Well-established tools available in the market |
Tools with somewhat ambiguous claims on productivity gains |
|
4 |
Skills/Competency of workers |
Defined |
Defined |
|
5 |
Production effort |
Can be estimated very well |
Quite loosely estimated |
|
6 |
Effort/Cost Estimation |
Definite |
Gut feel + Buffer + Contingency; typically overruled by managers |
|
7 |
Changes during production |
Negligible |
Very frequent |
|
8 |
Specification & Design |
Well-defined, reviewed, and signed-off before start of production |
Loosely defined, keeps changing throughout the lifecycle of product development |
|
9 |
Rework/Improvements |
Almost impossible to modify once the product is delivered |
Always possible to modify even after it goes into production |
Over and above the ambiguity and ever-changing scope (functionality and features) in the software development projects shown in Table 7.1, if you add issues related to the target technology platform on which the software is being developed, the cup of woes would be full to the brim. No wonder that the question that bothers a software development project team is, "Can there be an alternative and better-defined method of estimating project execution effort and costs?" On second thought, wouldn't it be wonderful to have a standard yardstick that can measure different products with different sets of functions against the same measuring scale? This yardstick should provide the user with a true comparison facility. For example, a measuring tape can measure a table size, the number of square feet in a house, the height of a tower, the distance between two locations, and more. All the items are different in nature but the measuring unit (yardstick) is same. Further, if you measure two similar items (such as the distance between locations) using the same measuring tape, you can compare the two measurements and even define a relative size ratio between them.
Function Point Analysis Method
The Function Point Analysis (FPA) methodology-based estimation model designed by Allan Albrecht of IBM in 1979, and now owned and continuously upgraded by IFPUG 2 (International Function Point Users Group), is perhaps the nearest to separating the functions delivered by a product from the technology platform on which the product is developed and hence the path to deriving the total effort and cost to deliver the application. The uniqueness of this FPA method enables the estimator to clearly size the software application (product) based purely on the functions that are expected to be delivered by the application. Perhaps it is due to this unique feature in the FPA method that its popularity and usage, as compared to other estimation methods, is the highest in the software developer community.
To understand the uniqueness of the FPA method, consider the example of a mobile phone, as shown in Figure 7.1. From a mobile phone user's perspective, the various functions the user expects to experience are
- To be able to communicate with contacts, friends, and family at will, irrespective of physical location and environment
- Instant, online access to the directory of contact numbers
- Provision to send SMS (text) messages to anyone, any time
- Internet browsing facility
- Storing and playing music
)
Figure 7.1 Defining size.
The FPA method is built on the premise that the function point (FP) count finally determined is based totally on the user's perspective of the functions expected to be delivered with the final product. In Figure 7.1, the Product Features map clearly to the end user functions that are expected to be delivered from the product (mobile phone). The FP counting method provides key inputs and maps to this aspect of Product Size.
Now consider the other half of the effort and cost-calculation activities (other than Product Sizing), which contribute toward arriving at the overall product pricing. Figure 7.1 shows those activities as:
- Manufacturing techniques/processes
- Skills/Competency of the workers
- Effective usage of right tools
- Raw material
- Quality process
- Pricing
- After sales service
A careful review of these parameters exposes the fact that most of these activities and processes are vendor specific and depend on the team assembled to deliver the project. The Effort and Cost text box contains all activities that point to vendor capabilities. Understanding the two different attributes of a production activity in terms of Product Size and Effort and Cost paves the way for further discussion and focus on Size as the single, critical measurement attribute in software development projects.
Size—The Differentiator
In a typical software development project scenario, the end user (or business user) is the owner (customer) of the application being developed and the IT development team or an external (outsourced) vendor is the supplier. Knowing the size of the application would come in very handy for the customer in situations where an evaluation of multiple vendors, including the internal IT development team, is being done. Here are the key pointers to successful negotiation of software development contracts:
- When the application's size is predetermined, the user can avoid being misled by different vendors on various functional complexities of the application. Instead, the vendors could be asked to provide their quotes for the defined size of the application being developed on a given technology platform.
- The user may not have a deep knowledge of the internals of the application development process or even the technology involved (sometimes the technical architecture and coding of a project is compared to a black box). Despite this situation, the user can still manage all project contract execution activities based on the final size of the product that needs to be delivered and accepted.
- "An experienced estimator can produce reasonably good size estimates by analogy if accurate size values are available for the previous project and if the new project is sufficiently similar to the previous one" 3.
- "Basically, function points are the quantified representation—or size—of the functions (or transactions) and data that go together to operate as a computer application. They provide a rough unit of size that is based upon the software component of the requirement" 4.
- Because functional features are separated through the size model, this opens the opportunity for the customer to choose the technology platform on which the application needs to be developed. Development costs on different technology vary based on skills available, and this leads to better control over cost and budget constraints.
- Function points are perhaps one of the best methods to estimate the size of an application. The method is quite ambiguous and therefore flexible enough to be molded into a variety of estimation needs, such as software development, maintenance, reengineering, enhancement, etc. Source Lines of Code (SLOC) or LOC is a poor alternative.
- "...'Size' refers in a very general way to the total scope of a program. It includes the breadth and depth of the feature set as well as the program's difficulty and complexity" 5.
The Yardstick
The business value delivered by software systems, in the form of functions provided, is one of the key inputs to the size of the system. This size can be used as an effective yardstick to facilitate many needs of an IT organization. In much the same way a measuring tape can be used to measure height, length, width, and depth of a variety of material objects and places, size can be an effective yardstick for software estimation in projects ranging from simple to complex.