Ddd Strategic Design

DDD Strategic Design

Comprehensive guidance on Domain-Driven Design, covering both strategic patterns (bounded contexts, context mapping, ubiquitous language) and tactical patterns (aggregates, entities, value objects, domain events, repositories) with practical code examples and workshop techniques.

When to Use This Skill

Use this skill for:

• Modeling a complex business domain with multiple stakeholder groups
• Identifying and defining bounded contexts for a new or existing system
• Drawing context maps to clarify integration relationships
• Designing aggregates with correct consistency boundaries
• Implementing domain events for cross-context communication
• Running event storming workshops to discover domain structure
• Establishing a ubiquitous language shared by developers and domain experts
• Refactoring an anemic domain model into a rich domain model

Trigger phrases: "domain-driven design", "DDD", "bounded context", "aggregate", "domain event", "event storming", "ubiquitous language", "context map", "value object", "anti-corruption layer", "shared kernel"

What This Skill Does

Provides domain modeling patterns including:

• Strategic Patterns: Bounded contexts, context maps, shared kernel, anti-corruption layer, published language, customer-supplier, conformist, open host service
• Tactical Patterns: Aggregates, entities, value objects, domain events, domain services, repositories, factories
• Discovery Techniques: Event storming, domain storytelling, example mapping
• Ubiquitous Language: Glossary construction, language enforcement in code
• Integration Patterns: Context mapping relationships and their implementation
• Code Organization: Package structure aligned with bounded contexts

Instructions

Step 1: Discover the Domain with Event Storming

Event storming is a collaborative workshop for rapidly discovering domain structure.

Workshop Setup:

Materials:  Unlimited wall space, sticky notes (orange, blue, yellow,
            pink, purple, green, red), markers
Duration:   2-4 hours for Big Picture, 4-8 hours for Design Level
Attendees:  Domain experts, developers, product owners, UX designers

Sticky Note Color Code:
  Orange  = Domain Event       (past tense: "Order Placed")
  Blue    = Command            (imperative: "Place Order")
  Yellow  = Actor / User       (who triggers the command)
  Pink    = External System    (integration point)
  Purple  = Policy / Process   ("When X happens, then Y")
  Green   = Read Model / View  (data the actor needs to see)
  Red     = Hot Spot           (question, conflict, unknown)

Big Picture Event Storming Steps:

Phase 1: Chaotic Exploration (30 min)
  - Everyone writes domain events on orange stickies
  - Place them on the wall roughly left-to-right by time
  - No discussion, no filtering, just capture

Phase 2: Timeline Enforcement (20 min)
  - Arrange events in temporal order
  - Identify parallel streams (concurrent processes)
  - Mark hot spots (red stickies) for conflicts or unknowns

Phase 3: Reverse Narrative (30 min)
  - Walk backwards through the timeline
  - For each event, ask: "What caused this?"
  - Add commands (blue) and actors (yellow)

Phase 4: Bounded Context Discovery (30 min)
  - Look for clusters of related events
  - Draw boundaries around clusters with tape or markers
  - Name each boundary (these become bounded context candidates)

Phase 5: Context Relationships (20 min)
  - Draw arrows between contexts for events that cross boundaries
  - Identify upstream/downstream relationships
  - Mark integration hot spots

Example Event Storm Output for E-Commerce:

┌─ Catalog Context ────────────────────────────┐
│  Product Created -> Product Priced ->         │
│  Product Published -> Product Discontinued    │
└──────────────────────────────────────────────┘
        │ Product Published
        ▼
┌─ Shopping Context ───────────────────────────┐
│  Item Added to Cart -> Cart Updated ->        │
│  Checkout Started -> Order Placed             │
└──────────────────────────────────────────────┘
        │ Order Placed
        ▼
┌─ Fulfillment Context ────────────────────────┐
│  Order Accepted -> Inventory Reserved ->      │
│  Shipment Created -> Shipment Dispatched ->   │
│  Delivery Confirmed                           │
└──────────────────────────────────────────────┘
        │ Order Placed
        ▼
┌─ Billing Context ────────────────────────────┐
│  Payment Initiated -> Payment Authorized ->   │
│  Payment Captured -> Refund Issued            │
└──────────────────────────────────────────────┘

Step 2: Define Bounded Contexts and Ubiquitous Language

Each bounded context gets its own ubiquitous language. The same real-world concept may have different names and different shapes in different contexts.

Example: "Product" Across Contexts:

Ubiquitous Language Glossary Template:

## Bounded Context: [Name]

### Glossary

| Term | Definition | Examples | NOT the same as |
|------|-----------|----------|-----------------|
| Order | A confirmed intent to purchase one or more items | Order #12345 | Cart (not yet confirmed) |
| Fulfillment | The process of picking, packing, and shipping | Warehouse operations | Delivery (carrier-side) |
| Backorder | An order accepted but not yet fulfillable | Out-of-stock item ordered | Preorder (not yet released) |

Enforcing Ubiquitous Language in Code:

// BAD: Generic, anemic, no domain language
public class Item {
    private String id;
    private double price;
    private int qty;
    private String type; // what does "type" mean in this domain?
}

// GOOD: Ubiquitous language embedded in types
public class OrderLineItem {
    private final ProductId productId;
    private final Money unitPrice;
    private final Quantity quantity;

    public Money lineTotal() {
        return unitPrice.multiply(quantity.value());
    }
}

Step 3: Draw Context Maps

A context map shows how bounded contexts relate to each other.

Context Map Relationship Types:

Partnership (P)
  Two contexts evolve together; teams coordinate closely.

Shared Kernel (SK)
  A small, shared subset of the domain model maintained by both teams.

Customer-Supplier (C/S)
  Upstream (supplier) provides what downstream (customer) needs.
  Downstream can negotiate.

Conformist (CF)
  Downstream conforms to upstream's model with no negotiation power.

Anti-corruption Layer (ACL)
  Downstream translates upstream's model into its own language.

Open Host Service (OHS)
  Upstream provides a well-defined protocol (API) for many consumers.

Published Language (PL)
  A shared, documented exchange format (e.g., JSON schema, protobuf).

Example Context Map:

┌──────────┐    OHS/PL     ┌───────────┐
│ Catalog  │ ───────────>  │ Shopping  │
│ Context  │               │ Context   │
└──────────┘               └─────┬─────┘
                                 │
                          C/S    │
                                 ▼
                           ┌───────────┐    ACL     ┌───────────┐
                           │ Billing   │ <──────── │ Payment   │
                           │ Context   │            │ Gateway   │
                           └───────────┘            │ (Stripe)  │
                                                    └───────────┘
┌──────────┐    C/S
│ Shopping │ ───────────>  ┌─────────────┐
│ Context  │               │ Fulfillment │
└──────────┘               │ Context     │
                           └──────┬──────┘
                                  │  CF
                                  ▼
                           ┌─────────────┐
                           │ Shipping    │
                           │ Partner API │
                           └─────────────┘

Step 4: Design Aggregates

Aggregates define transactional consistency boundaries. They are the most important tactical pattern.

Aggregate Design Rules:

Rule 1: Protect true invariants within a single aggregate.
Rule 2: Design small aggregates (prefer fewer entities inside).
Rule 3: Reference other aggregates by identity only (not object references).
Rule 4: Use eventual consistency across aggregate boundaries.
Rule 5: One transaction = one aggregate mutation.

Example: Order Aggregate (Python):

# domain/order/aggregate.py
from dataclasses import dataclass, field
from datetime import datetime
from typing import List
from uuid import UUID, uuid4

from domain.order.events import (
    OrderPlaced, OrderLineAdded, OrderConfirmed, OrderCancelled,
)
from domain.shared.value_objects import Money, Quantity

@dataclass(frozen=True)
class OrderLineItem:
    """Value object representing a line in an order."""
    product_id: UUID
    product_name: str
    unit_price: Money
    quantity: Quantity

    @property
    def line_total(self) -> Money:
        return self.unit_price * self.quantity.value

@dataclass
class Order:
    """Aggregate root for the Order bounded context."""
    id: UUID = field(default_factory=uuid4)
    customer_id: UUID = field(default=None)
    lines: List[OrderLineItem] = field(default_factory=list)
    status: str = field(default="draft")
    placed_at: datetime | None = field(default=None)
    _events: List = field(default_factory=list, repr=False)

    # --- Invariant-protecting methods ---

    def add_line(self, product_id: UUID, name: str,
                 price: Money, qty: Quantity) -> None:
        if self.status != "draft":
            raise InvalidOrderOperation(
                "Cannot add lines to a non-draft order"
            )
        if qty.value <= 0:
            raise ValueError("Quantity must be positive")
        line = OrderLineItem(product_id, name, price, qty)
        self.lines.append(line)
        self._events.append(OrderLineAdded(
            order_id=self.id, product_id=product_id, quantity=qty.value,
        ))

    def place(self) -> None:
        if self.status != "draft":
            raise InvalidOrderOperation("Order already placed")
        if not self.lines:
            raise InvalidOrderOperation("Cannot place an empty order")
        self.status = "placed"
        self.placed_at = datetime.utcnow()
        self._events.append(OrderPlaced(
            order_id=self.id,
            customer_id=self.customer_id,
            total=self.total.amount,
        ))

    def cancel(self, reason: str) -> None:
        if self.status not in ("draft", "placed"):
            raise InvalidOrderOperation(
                f"Cannot cancel order in status '{self.status}'"
            )
        self.status = "cancelled"
        self._events.append(OrderCancelled(
            order_id=self.id, reason=reason,
        ))

    @property
    def total(self) -> Money:
        return sum(
            (line.line_total for line in self.lines),
            Money(0, "USD"),
        )

    def collect_events(self) -> List:
        events = list(self._events)
        self._events.clear()
        return events

class InvalidOrderOperation(Exception):
    pass

Example: Value Objects:

# domain/shared/value_objects.py
from dataclasses import dataclass

@dataclass(frozen=True)
class Money:
    amount: int      # Store in smallest unit (cents)
    currency: str

    def __post_init__(self):
        if self.amount < 0:
            raise ValueError("Money amount cannot be negative")
        if len(self.currency) != 3:
            raise ValueError("Currency must be a 3-letter ISO code")

    def __add__(self, other: "Money") -> "Money":
        if self.currency != other.currency:
            raise ValueError("Cannot add different currencies")
        return Money(self.amount + other.amount, self.currency)

    def __mul__(self, factor: int) -> "Money":
        return Money(self.amount * factor, self.currency)

@dataclass(frozen=True)
class Quantity:
    value: int

    def __post_init__(self):
        if self.value < 0:
            raise ValueError("Quantity cannot be negative")

@dataclass(frozen=True)
class EmailAddress:
    address: str

    def __post_init__(self):
        if "@" not in self.address or "." not in self.address:
            raise ValueError(f"Invalid email: {self.address}")
        # Normalize to lowercase
        object.__setattr__(self, "address", self.address.lower().strip())

Example: Order Aggregate (C#):

// Domain/Orders/Order.cs
public class Order : AggregateRoot
{
    private readonly List<OrderLine> _lines = new();
    public IReadOnlyList<OrderLine> Lines => _lines.AsReadOnly();
    public CustomerId CustomerId { get; private set; }
    public OrderStatus Status { get; private set; } = OrderStatus.Draft;
    public Money Total => _lines.Aggregate(
        Money.Zero("USD"),
        (sum, line) => sum + line.LineTotal);

    public void AddLine(ProductId productId, string name,
                        Money unitPrice, Quantity qty)
    {
        if (Status != OrderStatus.Draft)
            throw new InvalidOrderOperationException(
                "Cannot modify a placed order.");

        _lines.Add(new OrderLine(productId, name, unitPrice, qty));
        AddDomainEvent(new OrderLineAddedEvent(Id, productId, qty.Value));
    }

    public void Place()
    {
        if (Status != OrderStatus.Draft)
            throw new InvalidOrderOperationException("Already placed.");
        if (!_lines.Any())
            throw new InvalidOrderOperationException("Order is empty.");

        Status = OrderStatus.Placed;
        AddDomainEvent(new OrderPlacedEvent(
            Id, CustomerId, Total.Amount));
    }
}

Step 5: Implement Domain Events

Domain events represent something important that happened in the domain.

Domain Event Definition:

# domain/order/events.py
from dataclasses import dataclass
from datetime import datetime
from uuid import UUID, uuid4

@dataclass(frozen=True)
class DomainEvent:
    event_id: UUID = field(default_factory=uuid4)
    occurred_at: datetime = field(default_factory=datetime.utcnow)

@dataclass(frozen=True)
class OrderPlaced(DomainEvent):
    order_id: UUID = None
    customer_id: UUID = None
    total: int = 0

@dataclass(frozen=True)
class OrderLineAdded(DomainEvent):
    order_id: UUID = None
    product_id: UUID = None
    quantity: int = 0

@dataclass(frozen=True)
class OrderCancelled(DomainEvent):
    order_id: UUID = None
    reason: str = ""

Event Dispatcher (in-process):

# infrastructure/event_dispatcher.py
from collections import defaultdict
from typing import Callable, Type

class EventDispatcher:
    def __init__(self):
        self._handlers: dict[Type, list[Callable]] = defaultdict(list)

    def register(self, event_type: Type, handler: Callable) -> None:
        self._handlers[event_type].append(handler)

    def dispatch(self, event) -> None:
        for handler in self._handlers[type(event)]:
            handler(event)

# Usage in application service
class PlaceOrderUseCase:
    def __init__(self, repo: OrderRepository,
                 dispatcher: EventDispatcher):
        self._repo = repo
        self._dispatcher = dispatcher

    def execute(self, command: PlaceOrderCommand) -> UUID:
        order = self._repo.find_by_id(command.order_id)
        order.place()
        self._repo.save(order)
        for event in order.collect_events():
            self._dispatcher.dispatch(event)
        return order.id

Step 6: Structure Code by Bounded Context

src/
├── catalog/                    # Bounded Context: Catalog
│   ├── domain/
│   │   ├── product.py          # Aggregate root
│   │   ├── events.py           # Domain events
│   │   └── value_objects.py    # Price, SKU, etc.
│   ├── application/
│   │   ├── commands.py         # CreateProduct, PublishProduct
│   │   └── handlers.py         # Command handlers
│   └── infrastructure/
│       ├── postgres_repo.py    # Repository implementation
│       └── search_index.py     # Elasticsearch adapter
│
├── ordering/                   # Bounded Context: Ordering
│   ├── domain/
│   │   ├── order.py            # Aggregate root
│   │   ├── events.py
│   │   └── value_objects.py    # Money, Quantity, etc.
│   ├── application/
│   │   ├── commands.py
│   │   └── handlers.py
│   └── infrastructure/
│       ├── postgres_repo.py
│       └── payment_client.py   # ACL to payment gateway
│
├── fulfillment/                # Bounded Context: Fulfillment
│   ├── domain/
│   ├── application/
│   └── infrastructure/
│
└── shared_kernel/              # Shared Kernel (minimal!)
    ├── domain_event.py         # Base event class
    ├── aggregate_root.py       # Base aggregate
    └── entity.py               # Base entity with identity

Best Practices

• Start with strategic design before tactical - Bounded contexts matter more than aggregate internals
• One ubiquitous language per bounded context - Do not force a single model across the entire system
• Keep aggregates small - Large aggregates cause contention and performance issues
• Reference aggregates by ID, not object - Prevents coupling and enables distribution
• Use eventual consistency across aggregates - Immediate consistency is only needed within one aggregate
• Domain events are facts, not commands - Name them in past tense ("OrderPlaced", not "PlaceOrder")
• Protect invariants in the aggregate - Never let external code reach into an aggregate to set state
• Avoid "god aggregates" - If an aggregate has 15+ fields, it probably needs splitting
• Value objects should be immutable - Use frozen=True in Python, record in C#/Java
• Run event storming before writing code - Discovery is faster with sticky notes than refactoring

Common Patterns

Pattern 1: Anti-corruption Layer Between Contexts

When integrating with a context you do not control (legacy system, third-party API), translate at the boundary:

# ordering/infrastructure/catalog_acl.py
from ordering.domain.value_objects import ProductInfo, Money
from catalog_client import CatalogAPIClient

class CatalogAntiCorruptionLayer:
    """Translates Catalog context models into Ordering context models."""

    def __init__(self, client: CatalogAPIClient):
        self._client = client

    def get_product_info(self, product_id: str) -> ProductInfo:
        raw = self._client.get_product(product_id)
        return ProductInfo(
            product_id=raw["id"],
            name=raw["title"],              # "title" in catalog, "name" in ordering
            price=Money(
                amount=int(raw["price_cents"]),
                currency=raw["currency"],
            ),
        )

Pattern 2: Saga for Cross-Aggregate Coordination

When a business process spans multiple aggregates, use a saga (process manager):

# ordering/application/sagas/order_fulfillment_saga.py
class OrderFulfillmentSaga:
    """Coordinates Order, Payment, and Inventory across contexts."""

    def __init__(self, orders, payments, inventory, events):
        self._orders = orders
        self._payments = payments
        self._inventory = inventory
        self._events = events

    def handle_order_placed(self, event: OrderPlaced) -> None:
        # Step 1: Reserve inventory
        try:
            self._inventory.reserve(event.order_id, event.line_items)
        except InsufficientStock:
            self._orders.reject(event.order_id, "Insufficient stock")
            return

        # Step 2: Authorize payment
        try:
            self._payments.authorize(event.order_id, event.total)
        except PaymentDeclined:
            self._inventory.release(event.order_id)  # Compensate
            self._orders.reject(event.order_id, "Payment declined")
            return

        # Step 3: Confirm order
        self._orders.confirm(event.order_id)

    def handle_order_cancelled(self, event: OrderCancelled) -> None:
        # Compensating actions
        self._payments.void(event.order_id)
        self._inventory.release(event.order_id)

Pattern 3: Specification Pattern for Complex Queries

Encapsulate business rules as composable specifications:

# domain/shared/specification.py
from abc import ABC, abstractmethod

class Specification(ABC):
    @abstractmethod
    def is_satisfied_by(self, candidate) -> bool: ...

    def and_(self, other: "Specification") -> "Specification":
        return AndSpecification(self, other)

    def or_(self, other: "Specification") -> "Specification":
        return OrSpecification(self, other)

    def not_(self) -> "Specification":
        return NotSpecification(self)

class AndSpecification(Specification):
    def __init__(self, left, right):
        self._left, self._right = left, right
    def is_satisfied_by(self, candidate) -> bool:
        return (self._left.is_satisfied_by(candidate)
                and self._right.is_satisfied_by(candidate))

# Usage
class HighValueOrder(Specification):
    def is_satisfied_by(self, order) -> bool:
        return order.total.amount >= 10000

class RecentOrder(Specification):
    def __init__(self, days: int = 30):
        self._days = days
    def is_satisfied_by(self, order) -> bool:
        cutoff = datetime.utcnow() - timedelta(days=self._days)
        return order.placed_at >= cutoff

# Compose specifications
vip_candidates = HighValueOrder().and_(RecentOrder(days=90))

Quality Checklist

• Bounded contexts identified and named using ubiquitous language
• Context map drawn showing all relationships (ACL, OHS, C/S, etc.)
• Ubiquitous language glossary maintained per bounded context
• Aggregates designed small (ideally 1-3 entities per aggregate)
• Aggregates reference each other by ID only, not by object
• All invariants protected within aggregate methods (no public setters)
• Domain events defined for all state transitions that cross contexts
• Value objects used for descriptive attributes (Money, Email, Quantity)
• Repository interface defined in the domain layer, implemented in infrastructure
• No framework dependencies in the domain layer
• Event storming conducted with domain experts before implementation
• Sagas or process managers handle cross-aggregate workflows

Related Skills

• architecture-design - System-level decomposition and trade-off analysis
• api-design - Designing published language and open host services
• microservices-patterns - Implementing bounded contexts as services
• event-driven-architecture - Event infrastructure and messaging patterns
• refactoring-expert - Refactoring toward a rich domain model

Version: 1.0.0 Last Updated: March 2026

Iterative Refinement Strategy

This skill is optimized for an iterative approach:

1. Execute: Perform the core steps defined above.
2. Review: Critically analyze the output (coverage, quality, completeness).
3. Refine: If targets aren't met, repeat the specific implementation steps with improved context.
4. Loop: Continue until the definition of done is satisfied.