What needs to be finished first
A user story cannot be completed within one sprint or exceeds agreed estimation thresholds (for example >8 Story Points).
The big picture of the story in context is known, so slices retain their relationship to the journey.
What needs to be ready before start
Story description; whiteboard or Miro with splitting patterns as checklist (Workflow Steps, Variations, Data Variants, Operations, Interface, Defer Performance, Defer Quality, Major Effort, Spike); backlog tool for new stories.
One PO or product-adjacent role holder; one to two engineers; optionally QA; one facilitator if team is new to splitting.
Story draft with acceptance criteria; known technical constraints; Definition of Done; sprint capacity as reference for slice size.
20-45 min per story
Make story visible. Splitting patterns as checklist on side. Rule: every slice delivers independent value or learning, no "backend slice" without outcome.
The one question this method answers
How can the story be split into 2-5 slices, each delivering value or learning, without losing overall benefit?
Marker: Phase
| Step | Duration | Action | Hint |
|---|---|---|---|
1Phase 1: Clarify story | 5-10 min | Briefly refresh story outcome, acceptance criteria and Definition of Done. Formulate main benefit in one sentence. | If outcome cannot be formulated in one sentence, several outcomes are mixed. Splitting starts here with outcome separation. |
2Phase 2: Choose splitting pattern | 10 min | Walk through patterns (Workflow Steps, Variations, Data Variants, Operations, Interface, Defer Performance, Defer Quality, Major Effort, Spike). Several patterns possible, choose best fit. | Workflow Steps and Variations are most common patterns. If none fits, story may be spike or needs a new cut. |
3Phase 3: Create slices | 15-20 min | Formulate slices as own stories. Each slice has outcome, acceptance criteria, sub-Definition of Done. Prioritize order by value-learning mix. | Slices should be independently deployable. If a slice only delivers value as package with others, split is too technical. |
4Phase 4: Estimate and backlog | 10-15 min | Estimate slices (Story Points, T-Shirt, rough hours). Close original story as parent or epic. Put slices into backlog in order. | If each slice is same size as original story, splitting was unsuccessful. Hard maximum per slice as sprint share. |
What comes out at the end
List of new stories in backlog with outcome, acceptance criteria, estimate and order; original story as epic or closed. Splitting note as comment with applied pattern and rationale.
Splitting decision as comment in backlog tool: applied pattern, date, participants. Preserve reference to parent on slice changes.
Compact working template for Story Splitting with context, input, output artifacts, and next step.
Concrete filled scenario, fictional example
Concrete filled scenario, fictional example
Original story: 21 Story Points, too large for one sprint.
Outcome: Solo self-employed users activate their account in under 5 min without Customer Success contact.
Pattern: Workflow Steps plus Variations.
Slices
Order: 01 -> 02 -> 03 -> 05 -> 04 (04 after data available).
Note: Original story WS-1303 closed as epic. Pattern Workflow Steps chosen because onboarding has clear step sequence. Variations possible later (for example team instead of solo).
Recognize symptoms and steer against them
Slices are named "Backend", "API", "Frontend", each without independent value.
Vertical slicing: each slice delivers usable value from frontend to database. If backend-only is needed, it is a spike or technical story, not slice.
Each slice is still 8+ points, sprint risk not reduced.
Repeat splitting. If pattern cannot split further, try different pattern. Several patterns can be combined.
User only sees value after slice 4 of 5.
Ask per slice: what can user do after this slice that they could not do before. If nothing, cut slice differently.
Slices repeat identical acceptance criteria, differences unclear.
Own, specific criteria per slice. Remove duplication. If criteria identical, slice is redundant.
One slice is actually technical research without outcome.
Label spike explicitly as spike, with timebox and clear learning goal. Do not disguise as value slice.
Done signals checkable in under a minute
Go to the profile for purpose, similar methods, and sources or continue to the next method in the catalog.