Stories act as a ‘pidgin language,’ where both sides (users and developers) can agree enough to work together effectively.

—Bill Wake, co-inventor of Extreme Programming


Stories are short descriptions of a small piece of desired functionality, written in the user’s language.  Agile Teams implement small, vertical slices of system functionality and are sized so they can be completed in a single Iteration.

Stories are the primary artifact used to define system behavior in Agile. They’re not requirements. Instead, they’re short, simple descriptions of functionality usually told from the user’s perspective and written in their language. Each one is intended to enable the implementation of a small, vertical slice of system behavior that supports incremental development.

Stories provide just enough information for both business and technical people to understand the intent. Details are deferred until the story is ready to be implemented. Through acceptance criteria and acceptance tests, stories get more specific, helping to ensure system quality.

User stories deliver functionality directly to the end user. Enabler stories bring visibility to the work items needed to support exploration, architecture, infrastructure, and compliance.


SAFe describes a four-tier hierarchy of artifacts that outline functional system behavior: Epic, Capability, Feature, and story. Along with Nonfunctional Requirements (NFRs), these Agile backlog items define the system and Solution Intent, model system behavior, and build up the Architectural Runway.

Epics, Capabilities, Features, and enablers are used to describe the larger intended behavior. But the detailed implementation work is described through stories, which make up the Team Backlog. Most stories emerge from business and enabler features in the Program Backlog, but others come from the team’s local context.

Each story is a small, independent behavior that can be implemented incrementally and provides some value to the user or the Solution. It’s a vertical slice of functionality to ensure that every iteration delivers new value. Stories are split into smaller ones so they can be completed in a single iteration (see the splitting stories section).

Often, stories are first written on an index card or sticky note. The physical nature of the card creates a tangible relationship between the team, the story, and the user: it helps engage the entire team in story writing. Sticky notes offer other benefits as well: they help visualize work and can be readily placed on a wall or table, rearranged in sequence, and even passed off when necessary. Stories allow improved understanding of the scope and progress:

  • “Wow, look at all these stories I’m about to sign up for” (scope)
  • “Look at all the stories we accomplished in this iteration” (progress)

While anyone can write stories, approving them into the team backlog and accepting them into the system baseline are the responsibility of the Product Owner. Of course, stickies don’t scale well across the Enterprise, so stories often move quickly into Agile lifecycle management (ALM) tooling.

There are two types of stories in SAFe, user stories and enabler stories, as described below.

Sources of Stories

Stories are typically driven by splitting business and enabler features, as Figure 1 illustrates.

Figure 1. Example of a business feature split into stories

User Stories

User stories are the primary means of expressing needed functionality. They largely replace the traditional requirements specification. (In some cases, however, they serve to explain and develop system behavior that’s later recorded to support compliance, traceability, or other needs.)

Because they focus on the user as the subject of interest, and not the system, user stories are value-centric. To support this, the recommended form of expression is the user-voice form, as follows:

As a (user role), I want (activity) to, so that (business value)

By using this format, the teams are guided to understand who is using the system, what they are doing with it, and why they are doing it. Applying the ‘user voice’ format routinely tends to increase the team’s domain competence; they come to better understand the real business needs of their user. Figure 2 provides an example.

Figure 2. Example user story in user voice form

‘Personas’ describe specific characteristics of representative users that help teams better understand their end user.  Example personas for the rider in Figure 2 could be a thrill-seeker ‘Jane’ and a timid rider ‘Bob’. Stories descriptions then reference these personas (As Jane I want…).

While the user story voice is the common case, not every system interacts with an end user. Sometimes the ‘user’ is a device (e.g., printer) or a system (e.g., transaction server). In these cases, the story can take on the form illustrated in Figure 3.

Figure 3. Example of a user story with a ‘system’ as a user

Enabler Stories

Teams may need to develop the architecture or infrastructure to implement some user stories or support components of the system. In this case, the story may not directly touch any end user. These are enabler stories and they support exploration, architecture, or infrastructure. Enabler stories can be expressed in technical rather than user-centric language, as Figure 4 illustrates.

Figure 4. Example enabler story

Enabler stories may include any of the following:

  • Refactoring and Spikes (as traditionally defined in XP)
  • Building or improving development/deployment infrastructure
  • Running jobs that require human interaction (e.g., index 1 million web pages)
  • Creating required product or component configurations for different purposes
  • Verification of system qualities (e.g., performance and vulnerability testing)

And, of course, enabler stories are demonstrated just like user stories, typically by showing the artifacts produced or via the user interface, stub, or mock-up.

Writing Good Stories

Good stories require multiple perspectives.  In agile, the entire team – Product Owner, developers, and testers – create a shared understanding of what to build to reduce rework and increase throughput. Teams collaborate using Behavior-Driven Development (BDD) to define detailed acceptance tests that definitively describe each story.

The 3Cs: Card, Conversation, Confirmation

Ron Jeffries, one of the inventors of XP, is credited with describing the 3Cs of a story:

  • Card – Represents capturing the statement of intent of the user story on an index card, sticky note, or tool. The use of index cards provides a physical relationship between the team and the story. The card size physically limits story length and premature suggestions for the specificity of system behavior. Cards also help the team ‘feel’ upcoming scope, as there is something materially different about holding ten cards in one’s hand versus looking at ten lines on a spreadsheet.
  • Conversation – Represents a “promise for a conversation” about the story between the team, Customer/user, PO, and other stakeholders. The discussion is necessary to determine more detailed behavior required to implement the intent. The conversation may spawn additional specificity in the form of attachments to the user story (e.g., mock-up, prototype, spreadsheet, algorithm, timing diagram). The conversation spans all steps in the story life cycle:
    • Backlog refinement
    • Planning
    • Implementation
    • Demo

These discussions provide a shared understanding of scope that formal documentation does not provide. Specification by example replaced overly detailed documentation of functionality. Conversations also help uncover gaps in user scenarios and NFRs.

  • Confirmation – The acceptance criteria provide the information needed to ensure that the story is implemented correctly and covers the relevant functional and NFRs. Figure 5 provides an example. Some teams often use the confirmation section of the story card to write down what they will demo.
Figure 5. Story acceptance criteria

Agile teams automate acceptance tests wherever possible, often in business-readable, domain-specific language. Automation creates an executable specification to validate and verify the solution. Automation also provides the ability to quickly regression-test the system, enhancing Continuous Integration, refactoring, and maintenance.

Invest in Good Stories

To remind themselves of the elements of a good story, people often use the INVEST model, developed by Bill Wake [1, 2]:

  •  I – Independent (among other stories)
  • N – Negotiable (a flexible statement of intent, not a contract)
  • V – Valuable (providing a valuable vertical slice to the customer)
  • E – Estimable (small and negotiable)
  • S – Small (fits within an iteration)
  • T – Testable (understood enough to know how to test it)

Estimating Stories

Agile teams use story points and ‘estimating poker’ to value their work [2, 3]. A story point is a singular number that represents a combination of qualities:

  • Volume – How much is there?
  • Complexity – How hard is it?
  • Knowledge – What’s known?
  • Uncertainty – What’s unknown?

Story points are relative, without a connection to any specific unit of measure. The size (effort) of each story is estimated relative to the smallest story, which is assigned a size of ‘one.’  A modified Fibonacci sequence (1, 2, 3, 5, 8, 13, 20, 40, 100) is applied that reflects the inherent uncertainty in estimating, especially large numbers (e.g., 20, 40, 100) [2].

Estimating Poker

Agile teams often use ‘estimating poker,’ which combines expert opinion, analogy, and disaggregation to create quick but reliable estimates. Disaggregation refers to splitting a story or features into smaller, easier to estimate pieces.

(Note that there are a number of other methods used as well.) The rules of estimating poker are:

  • Participants include all team members.
  • Each estimator is given a deck of cards with 1, 2, 3, 5, 8, 13, 20, 40, 100, ∞, and,?
  • The PO participates but does not estimate.
  • The Scrum Master participates but does not estimate, unless they are doing actual development work.
  • For each backlog item to be estimated, the PO reads the description of the story.
  • Questions are asked and answered.
  • Each estimator privately selects an estimating card representing his or her estimate.
  • All cards are turned over at the same time to avoid bias and to make all estimates visible.
  • High and low estimators explain their estimates.
  • After a discussion, each estimator re-estimates by selecting a card.
  • The estimates will likely converge. If not, the process is repeated.

Some amount of preliminary design discussion is appropriate. However, spending too much time on design discussions is often wasted effort. The real value of estimating poker is to come to an agreement on the scope of a story. It’s also fun!


The team’s velocity for an iteration is equal to the sum of the points for all the completed stories that met their Definition of Done (DoD). As the team works together over time, their historical trend of average completed story points per iteration builds a reliable picture of the team’s velocity. Knowing the velocity assists with planning and helps limit Work in Process (WIP), as teams don’t take on more stories than their historical velocity would allow. This measure is also used to estimate how long it takes to deliver epics, features, capabilities, and enablers, which are also forecasted using story points.


Capacity is the portion of the team’s velocity that is actually available for any given iteration. Vacations, training, and other events can make team members unavailable to contribute to an iteration’s goals for some portion of the iteration. This decreases the maximum potential velocity for that team for that iteration. For example, a team that averages 40 points delivered per iteration would have to adjust that maximum velocity to some lower number (for example to 36 points – the team’s actual capacity for that iteration) if a team member is going to be on vacation for one week of the two-week iteration. Knowing this in advance, the team only pulls a maximum of 36 points of stories into the iteration during iteration planning. This also helps during PI Planning to forecast the actual available capacity for each iteration in the PI so the team doesn’t over commit in building their PI Objectives.

Starting Baseline for Estimation

In standard Scrum, each team’s story point estimating—and the resulting velocity—is a local and independent concern. In SAFe, however, story points must share the same starting baseline so that estimates for features or epics that require the support of many teams can be understood.

SAFe uses a starting baseline where one story point is defined roughly the same across all teams. This means that work can be prioritized based on converting story points to cost. Of course, adjustments may be needed to account for the different average labor cost across geographies (e.g. United States, China, India, Europe). After all, there’s no way to determine the potential Return on Investment (ROI) if there is no common ‘currency.’  Normalized story points provide a method for getting to an agreed starting baseline for stories and velocity as follows:

  • Give every developer-tester on the team eight points (adjust for part-timers).
  • Subtract one point for every team member vacation day and holiday.
  • Find a small story that would take about a half-day to code and a half-day to test and validate. Call it a ‘one.’
  • Estimate every other story relative to that ‘one.’

Example: Assuming a six-person team composed of three developers, two testers, and one PO, with no vacations or holidays, the estimated initial velocity = 5 × 8 points = 40 points/iteration. (Note: Adjusting a bit lower may be necessary if one of the developers and testers is also the Scrum Master.)

In this way, story points are somewhat comparable to an ideal developer day, and all teams estimate using a common method. Management can better understand the cost for a story point and more accurately determine the cost for an upcoming feature or epic.

Note: There is no need to recalibrate team estimation or velocity after that point. It is just a starting baseline.

While teams will tend to increase their velocity over time—and that’s a good thing— in reality, the number tends to remain stable. A team’s velocity is far more affected by changing team size and technical context than by productivity variations. If necessary, financial planners can adjust the cost per story point a bit. Experience shows that this is a minor concern, versus the wildly differing velocities that teams of comparable size may have if they don’t set a common starting baseline. That simply doesn’t work at enterprise scale, making it difficult to make economic decisions.

Splitting Stories

Smaller stories allow faster, more reliable implementation, since small things go through a system faster, reducing variability and managing risk. Splitting bigger stories into smaller ones is, thus, a mandatory survival skill for every Agile team. It’s both the art and the science of incremental development. Ten ways to split stories are described in Leffingwell’s Agile Software Requirements [1]. A summary of these techniques follows:

  • Workflow steps
  • Business rule variations
  • Major effort
  • Simple/complex
  • Variations in data
  • Data entry methods
  • Deferred system qualities
  • Operations (ex., Create, Read, Update, Delete [CRUD])
  • Use-case scenarios
  • Break-out spike

Figure 6 illustrates an example of splitting by use-case scenarios.

Figure 6. An example of splitting a big Story into smaller stories

Stories in the SAFe Requirements Model

As described in the SAFe Requirements Model article, the Framework applies an extensive set of artifacts and relationships to manage the definition and testing of complex systems in a Lean and Agile fashion. Figure 7 illustrates the role of stories in this larger picture.

Figure 7. Stories in the SAFe Requirements Model

Figure 7 illustrates how stories are often (but not always) created by new features and how each has a story acceptance test. Further, each story should have a unit test. Unit tests primarily serve to ensure that the technical implementation of the story is correct. Also, this is a critical starting point for test automation, as unit tests are readily automated, as described in the Test-Driven Development (TDD) article.

Note:  Figure 7 uses Unified Modeling Language (UML) notation to represent the relationships between the objects: zero to many (0..*), one to many (1..*), one to one (1), and so on.

Learn More

[1] Leffingwell, Dean. Agile Software Requirements: Lean Requirements Practices for Teams, Programs, and the Enterprise. Addison-Wesley, 2011.

[2] Cohn, Mike. User Stories Applied: For Agile Software Development. Addison-Wesley, 2004.

Last update: 2 October 2018