Automated Function Points Analysis

Automated function points (AFP) analysis is no longer a pipe dream. Whether you are looking to size your software from written requirements, or directly from code, the proper tools now exist to automate function point analysis. FPA is most useful before the code is written, or in some cases, even before the requirements of a project are made apparent. After the code has been written, we primarily use function points sizing for benchmarking.

The Object Management Group established a method for automated function point analysis from code (you can find the details here). It was subsequently endorsed as an ISO standard, consistent with both IFPUG and repetition. Additionally, it has been freely available for over a decade and officially implemented by Cast Software, examining source code to generate a functional size measure of developed software.

Consistent Results

Any automated functional size estimation process, whether from code or from requirements, must be consistent in order to make decisions that would otherwise be unreliable. We have ensured that ScopeMaster always produces the same result for any given set of written requirements, achieving the golden 100% consistency.

Standards Based

ScopeMaster performs the text analysis necessary to identify the data movements employed by all functional sizing methods. It can then map these movement to each method to determine the COSMIC Function Point size estimate as well as the IFPUG Function Point size. COSMIC is the latest generation of functional sizing method, and as an ISO standard, it is designed to suit and complement modern software patterns. The COSMIC function point count generated by ScopeMaster is an estimate (typically within about 15% of a manual count) whereas the IFPUG estimate is less precise due to characteristics of the methodology itself. ScopeMaster does the “heavy lifting” of manual function point counting, saving considerable time and effort. In both cases, ScopeMaster speeds up the counting process to an incredible degree.

Automated Function Points – Performance

Analyzing about 100 function points per minute, the raw performance of ScopeMaster in sizing speed outperforms manual counting by leaps and bounds. In practical terms, however, a professional FP specialist would typically review automated counts for potential misinterpretations. Based on early feedback from an independent FP professional, ScopeMaster consistently delivers a 400% speed improvement over manual counting.

Language Support

ScopeMaster analyses and performs automated functional sizing in English, French, Spanish, Italian, and Portuguese. Translation to other languages is currently underway.

Accuracy

The accuracy of a functional size count refers to the adherence to the methodology principles and rules so that we can have confidence in the reliability of the size estimate and count. The accuracy of any FP count, manual or automated, is highly dependent on the wording of the requirements. Consistent functional requirements are key to a reliable functional size, as readers can then interpret the results more easily.

• Manual count vs. manual count: less than 5% variation (as compared by two certified counters)
• ScopeMaster COSMIC count vs. manual COSMIC count: less than 20% variation
• ScopeMaster IFPUG count vs. manual IFPUG count: less than 35% variation

We collaborated on a post-doctoral investigation with Erdir Ungan, PhD. and Professor Alain Abrain to validate our approach to automated functional sizing with the power of ScopeMaster, who have both endorsed ScopeMaster whole-heartedly.

Automated COSMIC Function Points

We chose COSMIC Functional sizing as our main sizing methodology for the following reasons:

1. Like its predecessors, it is a valid, consistent, proven, stable and reliable ISO standard measure of software size.
2. It is principle-based, which means it can be used for virtually all types of software without adjustment.
3. It is suited to Agile development where the requirements are not usually fully known upfront. With COSMIC, you can consistently measure a single requirement.
4. The correlation between COSMIC FP and effort has been proven again and again.

Automated IFPUG Function Points & Automated Simple Function Points

In summary: ScopeMaster has cracked the code of automated functional sizing from requirements text.

Automated Simple Function Points

Improvement is a journey, not a destination. We have recently updated ScopeMaster to better report automated simple function point counting.

Benefits of Automated Function Point Estimation (AFP)

AFP is far more reliable than story points. Experienced software professionals fully understand the merits of function points over Agile story points as the primary size metric for a software project. In short, Agile story points tend to vary in size up to 400% from team to team and are principally an indication of effort rather than functional size based on user requirements. Function point counts rarely vary more than 5% from one professional’s count to another. Function points are viable for contractual, benchmarking, or reliable pre-project sizing purposes, whereas story points are not.

No need to learn the functional sizing methodologies. The IFPUG sizing methodology is tough to learn and apply accurately. In practice, it takes at least six months of learning and experience to become fully skilled; COSMIC is considerably more straightforward. With ScopeMaster, there’s no need to learn the methodology in order to hit the ground running.

Comparing Automated Sizing Before and After Coding

In 2021, we worked with CAST Software on the (first ever!) investigation into examining the functional size of a piece of software using tools for automated sizing from requirements (before) and automated sizing from code (after). The overall results showed a very promising proximity of total sizes, leading to the conclusion that automated sizing is now a viable reality throughout the software development life cycle.

Common Uses of Function Points as the Core Metric in Software Projects:

Function point metrics can be used before development, (for estimation), during development (for control) and after a project (for benchmarking).

Common FP-based metrics are…

• functional size in function points or COSMIC function points
• staff requirements, based on known allocation & productivity by FP)
• cost development cost, based on benchmarks for similar industries & projects
• productivity (FP/Dev/mo., for developers, testers, and others)
• quality exposure, to defect potential in defects per FP
• quality achievement, in defects per FP (resolved)
• schedule estimations, in project schedules based on FP size

The single most important measure to take on any software project is a function point size. Don’t just take our word for it; the only ISO standards for measuring software size are function Points! And now that ScopeMaster has automated the process of functional sizing from requirements, there’s no excuse to use anything else!

A small bit of advice: don’t throw away story points just yet. They’re still a productive means of stimulating discussion amongst developers to find better ways of achieving some functionality. For everything else, use CFP.

Other forms of Sizing

Story Points

Story points are a variable, subjective, gameable, non-standard proxy for effort estimates. They simulate “ideal days” for a single person on a given team, at a single point in time. They are unsuitable—even dangerous—to use as a management metric, except as a very loose indicator. Whilst the discussion around why one story might be better than another is valuable, we discourage the use of story points overall.

Story Counts

After analysing hundreds of thousands of stories, we have proven that they tend to vary in size from 0 – 100+ CFP. The size distribution can be quite unpredictable too. It is therefore unwise to use story count as anything but (again) a loose indicator of magnitude.

Gartner Fast Function Points

Gartner FFP was created by Gartner to make functional sizing faster and easier. This methodology is proprietary, and unlike COSMIC and IFPUG, it is not an ISO standard. Gartner created a version 2 and a version 3 of FFP. The two versions are substantially different from each other; v2 counts are not comparable to v3 counts. We are only aware of its use in Italy.