How to tell when a recurring workaround is costly enough to justify a project.
ABOUT / EDISON
I write about the work behind the result.
Project stories often end up as a single results slide. The harder part—finding the real problem, earning cooperation, and agreeing on the rules—disappears. This site keeps a record of that work.
Why I write about both software and business
I started in enterprise backend development and later moved into customer-facing delivery. Customers rarely arrived with a complete requirements document. They had a business goal; engineers needed clear inputs and constraints; the business team needed something it could test quickly.
I learned to ask how the work gets done today, where people rely on manual workarounds, and what would count as a useful first result. Only then do I decide what to build and who needs to confirm each rule.
On this site, forward-deployed engineering (FDE) names that way of working: stay close to the operating context, define the problem with the people involved, and follow the change through to everyday use.
Working in a capital-intensive business changed how I evaluate plans
In 2022 I worked on a solar-energy venture, coordinating with local authorities, commercial partners, and suppliers. It was a capital-heavy business shaped by policy, partners, and field conditions at the same time.
That experience clarified where I can contribute most: manufacturing settings where I can combine software, digital channels, commercial judgment, and cross-functional coordination instead of working in one narrow specialty.
Four questions I keep working on
Start with the part of the workflow that fails, then decide what kind of system—if any—belongs there.
Connect shop-floor input, calculation rules, and management reporting without making the work harder for operators.
Connect the website and buyer conversations to reliable product information, logistics, and quotation work.
How I report evidence
I separate what happened, what I inferred and what has not been measured. A search-index snapshot is not a sales result. An inquiry is not an order. Without a dependable before-and-after measure, I do not publish a percentage improvement.
That makes the case less dramatic, but more useful if you want to know whether the method could work elsewhere.
Where to start
Read: what I look for when nobody has asked for a system ↗
Case: connecting the data behind WeChat reporting ↗
If you have a concrete problem, see how we can work together ↗