How We Fixed 5,000+ Duplicate Time Entries in a QuickBooks API Integration
A misconfigured pagination parameter caused 5,000+ duplicate time entries in a QuickBooks automation. Here's how we diagnosed and fixed the STARTPOSITION bug.
Crawl, Walk, Run: The Smart Way to Start Automation Projects
Learn the crawl-walk-run methodology for automation projects. Discover why starting simple beats building the perfect system, with real consulting examples from 400+ implementations.
TABLE OF CONTENTS
The fastest way to fail at automation is trying to build the perfect system on day one. After helping over 400 companies implement automation, we've learned that the "crawl, walk, run" approach consistently outperforms ambitious, complex builds. Here's why starting simple works, and how to apply this methodology to your next automation project.
What is the Crawl, Walk, Run Methodology?
The crawl-walk-run framework breaks complex automation projects into three distinct phases, each building on the previous one. Instead of attempting to automate everything at once, you start with the simplest version that works, prove the value, then expand.
Crawl: Build the minimum viable automation with basic functionality
Walk: Add enhancements, handle edge cases, improve user experience
Run: Scale the system, optimize performance, add advanced features
This isn't about lowering your standards. It's about managing risk and proving value before investing heavily in infrastructure you might not need.1
Why "Start Simple" Beats "Build Perfect"
In a recent consulting session, a client came to us with a detailed seven-phase plan generated by AI. The proposed tech stack included nine different tools: Supabase, Next.js, Vercel, Cursor, and five others. The timeline? At least six months.
The actual business need? Automatically register domains, set up hosting, and clone websites from form submissions.1
We simplified the stack to four tools: GoDaddy for domain registration, Cloudflare for DNS and hosting, GitHub for code repositories, and Zoho for form handling. The new timeline? One week to a working prototype.1
The difference between six months and one week isn't corner-cutting. It's about eliminating complexity that doesn't serve the immediate goal.
The Hidden Cost of Over-Engineering
When you start with a complex tech stack, you're betting that:
- You've correctly predicted all future requirements
- The tools you choose today will still be relevant in six months
- You won't learn anything during implementation that changes your approach
- Your users will actually need all the features you're building
In reality, none of these assumptions hold up. The tools landscape changes rapidly. User needs evolve. And you always learn critical details during implementation that change your design.2
Starting simple gives you the flexibility to adapt as you learn.
How to Apply Crawl-Walk-Run to Your Automation
Phase 1: Crawl - Build the Deterministic Core
Identify the workflow steps that are completely predictable and rule-based. For the website cloning project, this meant:
- Capture form submission from Zoho
- Register domain via GoDaddy API
- Configure DNS in Cloudflare
- Create GitHub repository
- Set up Cloudflare Pages hosting1
These steps are deterministic—they work the same way every time. Build and test these first. Don't touch the complex parts yet.
What to avoid: Adding "nice to have" features, building for scale you don't have yet, or trying to handle every possible edge case.
Phase 2: Walk - Add Intelligence and Flexibility
Once your deterministic core works reliably, layer in the parts that require judgment or variability.
For the website cloning example, the content generation piece—scraping the target site and producing HTML/JavaScript—was separated into phase two. This allowed the team to test different approaches (Cursor vs Claude Code) while the infrastructure was already operational.1
What to focus on: Handling the 80% use case well, improving error messages, and adding monitoring so you know when things break.
Phase 3: Run - Scale and Optimize
Only after proving the value in phases one and two should you invest in optimization, advanced features, and scale infrastructure.
This is when you add:
- Performance improvements
- Advanced error handling
- Multiple integration options
- Sophisticated reporting
- Scale-appropriate infrastructure
Real-World Application: Daily Sales Briefings
A director of operations came to us needing an automated daily briefing for their sales team. He'd already spent six hours trying different approaches that didn't work.2
Crawl phase (1-2 hours):
- Connect HubSpot and Google Calendar
- Filter for demo appointments
- Send daily Slack message at 8am
That's it. No AI summarization, no fancy formatting, no predictive scoring. Just the core workflow.
Walk phase (next session):
- Add company information lookups
- Format the message for readability
- Include HubSpot links for quick access
Run phase (future):
- Add AI-generated briefing notes
- Integrate with additional data sources
- Personalize by sales rep
The key insight: They got immediate value from the crawl phase. The walk and run phases became optional enhancements, not prerequisites for launch.
Common Mistakes That Slow You Down
Mistake 1: Planning for Problems You Don't Have Yet
"What if we need to handle 10,000 requests per day?"
"What if the API goes down?"
"What if we need to support multiple languages?"
These are all valid concerns—eventually. But if you're processing 50 requests per day today, optimize for that reality. Add complexity when you actually need it.3
Mistake 2: Choosing Tools for Their Features, Not Your Needs
The most powerful automation platform isn't the one with the most features. It's the one that solves your specific problem with the least overhead.
In the website cloning example, the client was "adamant" about using Cursor and Supabase because they'd researched these tools extensively. But the recommendation was to start with simpler tools that could deliver results in one week rather than six months.1
You can always migrate to more sophisticated tools later. You can't get back six months of waiting.
Mistake 3: Combining Multiple Problems Into One Project
Automation projects fail when they try to solve everything at once. Separate your concerns:
- Infrastructure setup (deterministic)
- Content generation (variable)
- Error handling (iterative improvement)
- User experience (evolves with feedback)
Build and test each piece independently. This makes debugging easier and gives you flexibility to iterate on each component separately.1
How to Know When to Move to the Next Phase
Don't rush to the walk phase until you've validated the crawl phase with real usage. Look for these signals:
Move from crawl to walk when:
- The core workflow runs reliably for at least a week
- You have at least 10-20 successful runs
- Users are actually using it (not just testing)
- You've identified the most common edge cases from real usage
Move from walk to run when:
- You're hitting infrastructure limits
- Manual interventions are taking significant time
- Users are explicitly requesting advanced features
- The business case for optimization is clear
The Speed Advantage Compounds
Starting simple doesn't just get you to launch faster. It creates a compounding advantage:
- Faster feedback: Real users provide better insights than planning sessions
- Lower switching costs: Simpler systems are easier to modify
- Proven value: Each phase justifies investment in the next
- Reduced risk: Small bets fail fast and cheap
- Team confidence: Quick wins build momentum
As one consultant explained: "I made that mistake a couple times, and I learned that it's better like this, so that way you have control of each individual project that is being created."4
Your First Steps
If you're starting an automation project this week:
- List your requirements in plain English, following the format: "When [this happens], do [this action]"
- Identify the deterministic core—the steps that are completely predictable
- Pick the simplest tools that can handle your current volume (not your imagined future volume)
- Set a one-week deadline to have something working, even if it's basic
- Plan your walk phase only after the crawl phase is working
The most successful automation projects aren't the most technically impressive. They're the ones that deliver value quickly, prove their worth, and evolve based on real-world usage.
Start crawling today. You can always learn to run later.
Ready to Start Your Automation Project?
If you're tired of planning the perfect system and ready to build something that works, book a consulting call to discuss your automation project using the crawl-walk-run approach.
Book a free discovery call to verify if this methodology will work for your specific situation.
Related Posts
Focused Directions - Beyond Email: Modern Client Communication + Client Portal Case Study
Stop using email to manage client workflows. See what modern clients expect and how a boutique advisory firm used a client portal to cut follow-ups and scale.
Automating Employee Onboarding on Growing Teams with Tally, Zapier, and Notion
Art gallery automated onboarding for 300+ employees using Tally forms, Zapier webhooks, and Notion database. Department-specific workflows included.
Made with Bullet