Three Standards Left After Reviewing 30 Project Retrospectives
Standards drawn from looking back at an education platform, e-contracts, travel recommendations, and mobile apps across 30 posts: state boundaries, recovery after failure, and completion backed by evidence.
Joohwan Yu · Founder, Full Stack Developer
Toris · Published
Contents
From August 12 to September 10, I wrote up my project experience one post per date. This series is not a work log of finishing a new feature each day, but a retrospective that rereads earlier work, one topic a day. Looking back from education content to contracts, travel recommendations, and mobile apps, similar questions remained under very different technologies.
First, Set the Boundaries Before the Values
In the education platform, I had to separate the original from the solving state. A scene’s ID and its display number also had different roles. In e-contracts, even when each user’s screen differed, the meaning of the contract state had to stay the same. The trouble often started not because a value was missing but because several meanings had been assigned to the same value.
On the next project, before choosing a store or a library, I will write down who changes the data, when it disappears, and what it connects to. If I can answer those three questions, the reasons for implementation choices become easier to explain.
Second, Look at What Comes After Failure, Not After Success
The save timing of the Drawer, re-entry into match records, and partial failure in photo sharing were all tied to whether users could pick their work back up. Data being kept was not the same as the user being able to reopen it. A request succeeding was also not the same as another user being able to see the result.
So I will put the scenario of leaving midway and coming back into the next completion criteria. Rather than adding tests indiscriminately, I will prioritize the paths where data, permissions, and external requests meet. The state after a real interruption reveals gaps in the design better than a single normal run.
Third, Leave Evidence to Explain the Result
A performance record needs its measurement conditions, and a recommendation improvement needs inputs to compare. A deployment fact and a user effect are also not the same result. I did not turn the shortcomings in testing and documentation left in past retrospectives into things that never happened in this post. It was more useful to leave improvements not yet made as standards for next time.
What I guarded against most while compiling this series was turning experience into a plausible success story. The choices had constraints at the time, and some parts look regrettable now. More than a list of what I built, why I judged as I did under those conditions was more useful for the next round of development.
When I start the next piece of work, I want to ask these three things again. Can I explain the boundaries of the data? Can I resume even if it fails midway? Do I have evidence to say it is done? A good retrospective was less a post that wraps up the past nicely than a record that makes the next choice a little more precise.
First post of the series: State Boundaries in Education Content
Advertisement