An architecturally significant requirement (ASR) is one that directly shapes a system’s architecture: missing it or getting it wrong would force an architectural change. Its significance is not the same as its prominence in a feature list. A requirement can be important to users yet leave the foundational design intact; another can appear routine but make the proposed design unsuitable. section 2.1.3 section 2.1
Identify ASRs while defining the problem, before committing to a solution. Engineers need to understand the user’s job and business context well enough to question requirements and expose constraints. Otherwise, an appealing technology can become the answer to a problem that has not yet been established. section 2.0 section 2.1
The “what would break this?” test keeps attention on architectural limits:
- Sketch a reasonable minimal design for the requirements currently understood. section 2.1.3
- Ask which additional functional requirement would force the team to abandon that design. section 2.1.3
- Flag requirements that cross that threshold as ASRs; leave details that only affect implementation for later decisions. section 2.1.3 section 2.5
Customer-record search
Backoffice staff need to find the right customer record without repeatedly narrowing broad results by hand. Filtering by date, name and type fits a straightforward initial design. A requirement for full-text search across customers, orders and interactions changes that assessment: it is flagged as architecture-shaping, while the exact ordering of filters is not. The distinction depends on the design under consideration, not on a rule that every search feature is an ASR. section 2.1.3 section 2.4
Table 2.6 Identified requirements
|
Type
|
Requirement
|
Notes
|
|
Functional
|
Full-text search across multiple entities (customers,
orders, interactions).
|
Requires capable search engine; architecture-shaping.
|
|
Functional
|
Filter by date, name, type and other attributes.
|
All filter details are usually not an ASR. We acknowledge
the need but do not spend much time debating or analysing a specific sort
order.
|
|
Availability
|
Standard tier – 99.9% uptime for core search.
|
Multi-node failover should be planned.
|
|
Durability
|
Moderate is acceptable for search index.
|
Indexes can be rebuilt from source, no need to worry about
data loss.
|
|
Latency
|
Acceptable tier – search response under 1s.
|
Indexing and solid query design is required however
without excessive optimisations.
|
|
Throughput
|
Under 200 concurrent users; under 50 requests per second.
|
Not a number to worry about however relevant for choosing
compute provisioning; deprioritises a
need for autoscaling complexity.
|
|
Volatility
|
Ranking and filter logic: Campaign tier. Indexing and
overall search functionality design: Foundational.
|
Initially, filters might be of high volatility while the
problem is still in the early stage. Decoupling it or making it config-driven
could be an idea.
|
|
Security
|
Service-to-service communication must be authenticated and
encrypted
|
Likely control to go with: mutual TLS, organisation
default
|
|
Legal
|
GDPR right to erasure and retention rules
|
Must be designed in, especially the deletion flow.
|
|
Architecture
|
Deployment to existing regions on the adopted cloud
provider.
|
Organisational standard to follow, functionality likely
will be just a part of existing larger infrastructure
|
The requirements table places full-text search beside ordinary filter details and records availability, durability, latency and throughput expectations. It shows how an ASR can be flagged without treating every requirement as a foundational design decision. section 2.4
An ASR is one part of a problem description ready for solution discussion. Record it alongside the user’s job, non-functional tiers, compliance constraints and applicable standards. Unanswered questions need not all be resolved first: make them visible and decide whether design work or a prototype is the better next step. section 2.4

