Software Development Case Studies You Can Check Yourself

Software development case studies are only useful when you can check them. Most agency portfolios are anonymous stories with round numbers you have to take on trust. Parkk Technology’s portfolio is the opposite: named products, public source code, published benchmarks, and a commit history anyone can read. This post walks through that portfolio, what each project measured, and how to verify every claim yourself before you hire anyone, including me.

I am Jason Ratzlaff, and Parkk Technology is my practice. The same projects appear on the case studies page in short form. Here is the long form, with the evidence for each.

What Makes a Software Development Case Study Credible?

Three things: a named subject, a number with a method behind it, and a way for an outsider to check. “A mid-market retailer cut stockouts by a third” has none of them. “FastMagento renders the home page of a 513,000-product catalog in 775 ms instead of 2,911 ms, and here is the benchmark script” has all three.

Case Study: FastMagento, a Faster Way to Serve Magento

FastMagento is an open-source Magento 2 extension that serves storefront product reads (product pages, search results, cart and checkout) from OpenSearch instead of Magento’s EAV database tables. It is benchmarked in public on a purpose-built catalog of 513,000 simple products, cold render, no page cache.

The scale-invariant number is SQL queries: the home page drops from 10,090 queries to 80. Two companion extensions extend the same approach: FastMagento Checkout and FastMagento Personalization. The full story is in why Magento is still slow after Varnish.

How to verify: the benchmark tables and method are in the README; the code is open.

Case Study: Magento at Scale for Named Brands

As Magento lead across 24+ stores I drove more than $65M in client revenue for brands including Yamaha USA, Rubie’s, simplehuman, eos, The Coffee Bean & Tea Leaf, HerRoom and The House of Staunton, plus a commercial Magento extension line that sold more than $1M. That extension line came from my former role as a partner and lead engineer at Strattas; some of those early extensions are still visible in the parkktech GitHub organization.

How to verify: the stores are public; ask me which work was mine on any of them.

Case Study: Counter-UAS Defense Hardware

This is a counter-UAS RF-detection product. I was founding engineer and am an equity partner, and I took it from concept through CAD, PCB design, firmware and the operator interface. It is deployed and in real-world use. It is also the one project here where the code is private, for the reasons you would expect with defense hardware; what I can show publicly is the tooling it produced, like fluxplace, a signal-flow-aware component placer for KiCad.

How to verify: ask in a conversation; some details are only shared under the right agreements.

Case Study: Fintech at Phillips & Cohen

As VP of Engineering at Phillips & Cohen I built a 20-developer organization and launched five fintech products into Westpac, Best Egg, Merrick and E.ON Next, through penetration tests and international compliance reviews. In financial software a bug is measured in dollars, which is why the audit trails and authorization tests in my newer work look the way they do.

Case Study: SpendifiAI, Built in Public

SpendifiAI is an AI personal finance platform on Laravel, React and PostgreSQL with Plaid bank connections, and its entire history is public: 218 commits, with the research, plans and verification notes for every phase committed next to the code.

• Accountant accounts and audited client impersonation, in one 5,086-line commit.
• A tax deduction finder that checks 69 deductions and credits.
• A tax document vault with AI extraction for 25 form types and a hash-chained audit log.

How to verify: clone the repository and run the tests.

Case Study: Tooling Built in Public

Some of the most useful evidence is the tooling a practice builds for itself, because nobody writes a marketing page for it. Two examples are public.

fluxplace: component placement for KiCad

fluxplace (188 commits) reads a circuit board’s communication graph, leaves power and ground out so they do not tie everything together, and re-places components to minimize weighted wirelength: connected parts pull together, decoupling capacitors collapse onto their chip’s power pins, and buses line up. It runs inside the KiCad PCB editor as a plugin. It came out of that counter-UAS hardware work.

Shrinkage: keeping AI-written code small

Shrinkage (121 commits) is a plugin for Claude Code and Copilot with one job: make the AI write less and better code, and safely delete code that is no longer needed. It exists because AI-assisted development, which is how much of SpendifiAI was built, tends to add code faster than anyone removes it.

How to verify: both are open source with their full commit histories.

Case Study: EdgeTrades, AI Market Analysis

EdgeTrades is an LLM-driven market-analysis platform on Laravel and PostgreSQL, where models do real analytical work rather than sitting in a chat box on top of a dashboard. It is a private product, so it is described here and on the case studies page without numbers.

What These Projects Have in Common

Different industries, the same habits:

1. Measure first FastMagento’s work began with a benchmark of stock Magento, not with a rewrite.
2. Keep what works FastMagento leaves category grids on MySQL, FastMagento Checkout falls back to the stock checkout if anything fails, and SpendifiAI’s accountant exports reuse the same tax service as individual users.
3. Write down decisions SpendifiAI’s plans and verification notes sit in the repo next to the code.
4. Publish the result With the method, so someone else can reproduce it.

How to Choose a Software Development Company Using Its Case Studies

Whoever you are evaluating, ask these questions of every case study:

• Is the client or product named, and can you find it?
• Is there a number, and does it say how it was measured?
• Is any of the code or its history public?
• Does the case study admit a limit, or does everything work perfectly?
• Can the person you would work with explain the technical decisions themselves?

Frequently Asked Questions

How do I verify a software company's portfolio?

Look for public code, public benchmarks and named clients you can look up. Read the commit history if it is public. Ask the engineer who did the work to explain a specific technical decision in detail.

Is FastMagento's benchmark independently reproducible?

The benchmark method, catalog size and results are published in the FastMagento README, and the extension is open source, so anyone can run it on their own hardware. Results will vary with hardware and catalog.