SR&ED for Software Development Companies (2024)

Software development companies are among the largest claimants of SR&ED tax credits in Canada, yet many tech companies either don't claim or significantly underestimate their eligible work. This guide helps software companies understand what qualifies and maximize their SR&ED claims.

Why Software Companies Should Claim SR&ED

The Opportunity

Software companies often have 60-80% of expenses as salaries (highly eligible for SR&ED)
Much development work involves genuine technological uncertainty
Claims of $100K-$500K+ are common for established tech companies
Refundable credits provide cash flow critical for growth

Common Misconception

"We just build software—that's not R&D."

Wrong. Software development frequently involves:

Algorithm development with uncertain outcomes
Performance optimization requiring experimentation
Architecture decisions without clear best practices
Integration challenges with no documented solutions

What Software Work Qualifies

Eligible Software SR&ED

Algorithm Development:

Machine learning models with uncertain accuracy
Novel optimization algorithms
New approaches to data processing
Innovative search or recommendation systems

Architecture Innovation:

Scalability solutions for unprecedented loads
Novel approaches to distributed systems
New database architectures or query optimization
Real-time processing systems

Performance Optimization:

Speed improvements requiring technical experimentation
Memory optimization with uncertain outcomes
Novel caching or compression strategies

Security Innovation:

New encryption or authentication methods
Novel approaches to threat detection
Innovative security architectures

Integration Challenges:

Complex API integrations with technical unknowns
Data transformation with no existing solutions
Novel protocol implementations

Not Eligible

Routine development:

Standard CRUD applications
Well-documented API implementations
Minor feature additions without technical challenge
Bug fixes (unless significant technical investigation)
Routine maintenance and updates

Non-technical work:

UI/UX design without technical innovation
Project management
Requirements gathering
Testing of known functionality

The Key Test

Ask: "Did we face technical uncertainty that required experimentation?"

YES: Tried multiple approaches, analyzed results, iterated
NO: Knew how to do it, just took time to build

Examples of Eligible Software SR&ED

Example 1: AI/ML Development

Project: Build recommendation engine for e-commerce platform

Uncertainty: Existing collaborative filtering models achieved only 12% CTR. Needed to reach 20% with cold-start users and sparse data.

Experimentation: Tested 8 different model architectures (collaborative filtering, content-based, hybrid approaches, neural networks). Experimented with feature engineering, data augmentation techniques. Developed novel ensemble approach.

Advancement: Achieved 23% CTR through hybrid model combining user embeddings with contextual features. Proved viability of transfer learning for cold-start problem.

Claim value: $185,000

Example 2: Scalability Challenge

Project: Scale real-time messaging platform from 10K to 1M concurrent users

Uncertainty: Standard approaches (vertical scaling, traditional message queues) couldn't achieve required latency at scale. No existing architecture handled this combination of throughput and latency.

Experimentation: Tested 5 message broker configurations, developed custom sharding algorithm, experimented with connection pooling strategies, built prototypes with different consistency models.

Advancement: Developed novel multi-tier architecture achieving 50ms latency at 1M concurrent connections. Created new approach to eventual consistency for messaging applications.

Claim value: $120,000

Example 3: Data Processing Innovation

Project: Process and analyze 50TB daily data with 15-minute SLA

Uncertainty: Traditional ETL pipelines couldn't meet SLA at this volume. Existing streaming frameworks had memory limitations.

Experimentation: Tested Apache Spark configurations, built custom streaming processor, experimented with partitioning strategies, developed novel checkpointing mechanism.

Advancement: Created hybrid batch-streaming architecture achieving 12-minute processing time. Developed memory-efficient streaming approach reusable for similar problems.

Claim value: $95,000

Documentation for Software SR&ED

Technical Documentation

What to keep:

Design documents showing technical decisions
Architecture diagrams and iterations
Algorithm pseudocode and analysis
Performance benchmarks and test results
Git commit history with meaningful messages
Code review notes discussing technical challenges
Technical debt/spike tickets

Example documentation:

Commit: "Experiment with LSTM vs Transformer for sequence prediction"
- Hypothesis: Transformer attention mechanism will improve accuracy on long sequences
- Results: 15% accuracy improvement but 3x inference time
- Next: Test distilled model and quantization

Time Tracking

Approaches for software teams:

Project management tools (Jira, Linear) with SR&ED tags
Sprint allocation by project
Git analytics showing commits per project
Team estimates by project area

What to allocate:

Development time on SR&ED projects
Architecture planning and design
Technical experimentation
Code review for SR&ED work
Testing and analysis

Common Software Documentation Gaps

No record of technical decisions - Why did you choose approach A over B?
Missing experimentation evidence - What did you try that didn't work?
No performance metrics - How did you measure advancement?
Vague commit messages - "Fixed stuff" vs. "Tested CNN architecture, accuracy 72%"

Expenditure Considerations

Salaries (Usually 80%+ of Software Claims)

Eligible roles:

Software engineers/developers
Data scientists
DevOps (when building experimental infrastructure)
QA (when testing SR&ED work)
Tech leads/architects

Allocation approach:

What % of sprint time was on SR&ED projects?
What projects involved genuine experimentation?
Typically ranges from 40-80% for development teams

Cloud Computing Costs

Eligible:

Compute costs for experimentation
Training ML models
Performance testing environments
Development/staging for SR&ED

Not eligible:

Production infrastructure
Commercial customer usage
General business applications

Contractors

Common for:

Specialized expertise (ML consultants, security experts)
Additional development capacity
Specific technical challenges

Requirement: Contract must specify SR&ED work (not general development support)

Common Mistakes Software Companies Make

1. Claiming Routine Work

Wrong: Claiming all development because "software is R&D" Right: Claiming projects with genuine technological uncertainty

2. Poor Technical Narratives

Wrong: "We built a recommendation engine" Right: "We faced uncertainty in achieving 20% CTR with sparse user data. We tested 8 model architectures and developed a novel ensemble approach..."

3. Ignoring Failed Projects

Wrong: Only claiming successful projects Right: Failed experiments are valuable SR&ED evidence

4. Missing Cloud Costs

Wrong: Only claiming salaries Right: Including eligible cloud computing for experimentation

5. Under-claiming Time

Wrong: Conservative 20% allocation when 60% was SR&ED Right: Reasonable allocation based on actual project work

Maximizing Software SR&ED Claims

1. Identify All Eligible Work

Review:

Every sprint for technical challenges
Architecture decisions requiring research
Performance work with uncertain outcomes
Integration challenges
Failed experiments

2. Document Contemporaneously

Simple approach:

Tag tickets as SR&ED in your project tool
Write meaningful commit messages
Keep design docs and technical notes
Record test results and benchmarks

3. Allocate Time Reasonably

Consider:

Sprint allocation by project
Senior vs. junior time on experimental work
Architecture and design time
Testing and analysis time

4. Include All Eligible Costs

Salaries (don't forget the proxy overhead)
Cloud computing for R&D
Contractors for technical work
Tools for experimentation

5. Work with Tech-Savvy Consultants

Look for:

Software industry experience
Understanding of modern development practices
Ability to write compelling technical narratives
Knowledge of cloud cost eligibility

See our detailed guide on How to Choose an SR&ED Consultant →

Browse Consultants by Service

Software companies typically need:

Claim Preparation Services - End-to-end claim preparation
Technical Documentation Services - Building compelling technical narratives
Financial Analysis Services - Optimizing your expenditure claims
CRA Audit Support - Defense during CRA review

Find Software Industry SR&ED Consultants →

Frequently Asked Questions

Is building a SaaS product SR&ED?

Parts of it likely qualify—algorithm development, scalability challenges, novel features with technical uncertainty. Routine feature implementation does not.

What about open source work?

If you're solving technical challenges not addressed by existing open source (or extending it significantly), this can qualify.

Do we need to patent our work?

No. SR&ED doesn't require patents. Most software SR&ED is never patented.

Is refactoring eligible?

Generally no—unless refactoring involves genuine technical uncertainty and experimentation (not just code cleanup).

Can mobile app development qualify?

Yes, when it involves technical innovation—novel UI frameworks, performance optimization, hardware integration challenges. Standard app development doesn't qualify.

What if we use existing frameworks like TensorFlow?

Using frameworks doesn't disqualify you. The question is whether you faced uncertainty in applying them to your specific problem.

Next Steps

Review recent sprints for technically challenging work
Identify projects with experimentation and uncertain outcomes
Improve documentation practices for future claims
Estimate claim value based on eligible developer time
Connect with a consultant experienced in software SR&ED

Find Software SR&ED Consultants →

Last updated: November 2024. Consult a qualified SR&ED professional for your specific situation.