Feature Driven Development (FDD)
What is Feature Driven Development (FDD)?
Feature-driven development (FDD) is an agile method that starts with an overall domain model, then plans and builds software as a list of small, client-valued features, each taking no more than two weeks.
Updated November 2026 · 3 min read
Feature-Driven Development (FDD) is an agile method that plans, builds and tracks software as a list of small, client-valued features, each of which can be designed and built in two weeks or less. It combines up-front domain modeling with short, feature-by-feature delivery, which made it one of the agile methods considered better suited to larger teams in its early years.
FDD was devised by Jeff De Luca in 1997 for a large software project at a bank in Singapore, drawing on Peter Coad's object modeling techniques. It was first published in Java Modeling in Color with UML (1999) by Coad, Eric Lefebvre and De Luca, and described in more detail in A Practical Guide to Feature-Driven Development (2002) by Stephen Palmer and John Felsing.
What counts as a feature in FDD
In FDD a feature is a small piece of functionality that a client would value, written in the form "<action> the <result> <by|for|of|to> a(n) <object>." Examples include "Calculate the total of a sale" and "Validate the password of a user." Features are grouped into feature sets, such as "making a sale," and feature sets into broader subject areas. This naming discipline keeps features small and testable and makes progress reports readable by business stakeholders.
The five processes of FDD
- Develop an overall model. Domain experts and developers build an object model of the problem domain together.
- Build a features list. Decompose the domain into subject areas, feature sets and individual features.
- Plan by feature. Sequence feature sets, assign them to chief programmers and assign classes to class owners.
- Design by feature. A chief programmer forms a small feature team, which produces a design for a group of features and reviews it.
- Build by feature. The team codes, unit tests, inspects and promotes the features to the main build.
The first three processes happen mostly at the start of the project; the last two repeat for every feature.
Roles and progress tracking
FDD defines roles including project manager, chief architect, development manager, chief programmers, class owners and domain experts. Its distinctive idea is class ownership: each class has a named developer responsible for it, and feature teams form around the owners of the classes a feature touches.
Progress is tracked through six milestones per feature, each worth a fixed percentage of completion:
| Milestone | Weight | Cumulative |
|---|---|---|
| Domain walkthrough | 1% | 1% |
| Design | 40% | 41% |
| Design inspection | 3% | 44% |
| Code | 45% | 89% |
| Code inspection | 10% | 99% |
| Promote to build | 1% | 100% |
FDD example: Tracking feature progress
Consider a hypothetical retail software team working on the feature set "Making a sale," which contains three features. "Calculate the total of a sale" has been promoted to the build (100%). "Apply a discount to a sale" has finished coding but not code inspection (89%). "Print the receipt for a sale" has passed design inspection (44%). The feature set is (100 + 89 + 44) ÷ 3 = 233 ÷ 3, or about 78% complete. Because every feature uses the same milestones, the same calculation rolls up to subject areas and the whole project, giving managers a consistent view of progress.
FDD vs. Scrum
Both deliver working software in short increments, but they organize the work differently. Scrum plans work in fixed-length sprints, leaves design to a self-organizing team and prescribes no up-front modeling. FDD starts with an overall domain model, plans and tracks progress by individual feature through fixed milestones, and assigns each class to a named owner. Some teams combine the two, using an FDD-style feature list and progress reporting inside Scrum sprints.
Common pitfalls
- Over-modeling at the start. The overall model should be good enough to plan with, not a complete design. Teams that spend months modeling lose the agile benefit.
- Features that are too big. Anything that will take more than two weeks should be split.
- Class ownership as a bottleneck. If one owner is overloaded, features that touch their classes queue up. Balance ownership and allow backups.
- Treating percentages as precise. The milestone weights are a convention for consistent reporting, not a measure of actual effort.
FDD is part of the wider agile software development family and shares its focus on small, valuable increments with user stories in Scrum. Teams comparing it with other methods often look at Extreme Programming and DSDM.
Frequently asked questions
How does feature-driven development differ from Scrum?
Both deliver working software in short increments, but Scrum plans work in fixed-length sprints, leaves design to a self-organizing team and requires no up-front modeling. Feature-driven development begins with an overall domain model, tracks progress per feature through fixed milestones, and gives each class a named owner. Some teams run an FDD-style feature list inside Scrum sprints.
How is progress measured in feature-driven development?
FDD tracks each feature through six milestones with fixed weights: domain walkthrough 1%, design 40%, design inspection 3%, code 45%, code inspection 10% and promote to build 1%. A feature's completion is the sum of milestones reached, and averaging features gives progress for a feature set, subject area or whole project. The weights are a reporting convention, not a measure of effort.
Who created feature-driven development?
Jeff De Luca devised feature-driven development in 1997 for a large software project at a bank in Singapore, building on Peter Coad's object modeling techniques. The method was first published in Java Modeling in Color with UML in 1999, and later described in more detail by Stephen Palmer and John Felsing in A Practical Guide to Feature-Driven Development.
Related terms
Agile Software Development
Building software through brief iterations and frequent user feedback.
User Story
A plain-language statement of what a user wants to do and why.
Extreme Programming (XP)
An agile method built on intensive engineering practices.
Dynamic Systems Development Method (DSDM)
An agile project framework that fixes time and cost, not scope.