I distribute four applications through the Microsoft Store. One gets far more page views than the others.
It is YuGPT.
The application I have been working on most enthusiastically is YuHex, my Windows hex editor for inspecting and editing the bytes inside files. Yet in the Store figures I am looking at, YuGPT gets roughly two orders of magnitude more page views than YuHex.
Does that mean YuGPT sells well? No. It barely sells at all.
People visit. Very few buy.
As the developer, I find that hard to ignore. The app attracting attention is not the one receiving most of my effort, and the sales tell a different story from the views. The numbers do not explain what people thought when they arrived.
So I went back to the problem that made me build YuGPT in the first place. This time, the changes would not all be in the code.
I only wanted to find that answer again
I made YuGPT because I needed it myself. It is a Windows application for backing up ChatGPT and Codex history locally and searching across the saved conversations.
Imagine remembering that you already looked into a problem. You recall part of the subject, perhaps a keyword, but not the conversation title. You might not even remember which service you used. You want the answer you read before, together with the exchange that led to it.
Moving between places to search is a nuisance. I wanted to keep the history and search it from one place. That was the starting point.
YuGPT does not generate new AI answers. It helps you revisit conversations you have saved. Finding an answer is useful; seeing what question produced it can be useful too.
Backup sets and the conversations saved within them, together in one library.
This did not begin with an impressive market-research exercise. I had a problem I wanted to solve.
When you build a tool for yourself, you already know what it does and how you intend to use it. Someone opening its Store page for the first time has none of that background. What is obvious to the developer has not necessarily been explained to the visitor.
A page view does not tell me why someone visited
Why does YuGPT attract so many views?
Perhaps people want a way to keep their ChatGPT and Codex history. Perhaps the name suggests a different kind of application. Someone might arrive through a search and decide that it does not fit their setup or their needs.
Those are possibilities, not findings. I have not asked those visitors why they came or why they left.
Page views are not buyers. They are not a count of people who downloaded and used the software, either. The figures I have do not tell me what someone expected or where the page failed to meet that expectation.
It is tempting to settle on a simple explanation: AI-related applications attract attention, or the price puts people off. Filling a gap in the evidence with a convenient answer would not help me decide what to change.
One possibility did seem worth addressing, though.
What if people thought YuGPT was available only as a paid app?
Like YuHex, YuGPT uses a freemium model with paid and free editions. But seeing the paid edition on the Microsoft Store does not mean a visitor knows a free one exists.
I knew it existed. That is not quite the same thing as having told someone else.
If there is a free edition, say so clearly
I revised the Microsoft Store listing and the website. I also made the Free Edition available on GitHub.
Before asking someone to buy, I wanted to make it clear that they could find out whether the tool suited them.
Putting the word “free” somewhere prominent is only part of that job. People also need to know what they can do with the Free Edition and what changes if they choose the Store Edition.
In the current editions, the free version can import an official ChatGPT export ZIP and local Codex history from the PC. It supports searching and reading those conversations. Its limit is five backup sets, not five individual conversations.
The Store Edition adds direct ChatGPT backup after sign-in, removes the Free Edition's backup-set limit, and allows changing the archive's storage folder. The Free Edition is a separate, limited edition distributed on GitHub, not a time-limited Microsoft Store trial.
Someone evaluating the app may want to answer fairly practical questions: can it find the conversation I need, and is the saved material comfortable to read? Trying those tasks gives them something more useful than a long list of feature names.
If the Free Edition is enough, I am happy for people to use it. If direct backup or the other differences matter to them, the Store Edition is another option. Either way, I want the choice to be informed.
Of course I would like the app to sell. But discovering a mismatch before buying is better for the customer, and probably for its developer, than discovering it afterward.
Five months later, there was a bug to fix too
Improving the description is not much help if a problem inside the application gets in the way next.
I updated YuGPT for the first time in five months. There was a bug.
One of the fixes concerned the relationship between a search result and the query that produced it. If someone edited the search field after running a search, then opened an earlier result, the app used the newly typed query when locating the matching passage in the conversation.
For example, imagine searching for “backup,” getting a list of results, and then replacing the text in the search field with “archive.” Until another search runs, the displayed results still belong to “backup.” Opening one should use the query that produced that result list.
The text currently in an input field and the query associated with displayed results are not necessarily the same. I changed the app to retain the search term belonging to those results.
It does not look like a major new feature. But in a search application, getting back to the place you found matters. Helping someone locate an old answer and then confusing them when they open it rather defeats the purpose.
Between getting a visitor to open the page and having someone feel glad they used the tool, there are small fixes like this one. A clearer description does not repair the application by itself.
I have not solved the sales mystery yet
I made the free option clearer, provided a download route, and fixed a problem in the app.
That does not let me conclude, “Here was the cause, and now it sells.” I do not have the measurements or feedback needed to establish what visitors knew about the editions, let alone whether that was their reason for not buying.
What I changed was the part I could inspect and improve, rather than the unexplained number itself.
The app I most want to work on is not necessarily the app people most want to investigate. Independent development has a way of refusing to follow the story I had in mind. Still, if people stop to look at YuGPT, I would like to explain what it is and how they can try it a little better.
I built it to reduce my own “where was that?” moments. It would be nice if it reached other people with the same problem.
Before adding more features, I needed to reconsider the entrance: had I made the first decision unnecessarily difficult? Then I needed to check what happened after someone came inside. This update involved both.
I was the person closest to the tool and may still have overlooked how to introduce it. The application got a bug fix. Apparently, its developer needs an occasional update too.
If you would like to try YuGPT
The Free Edition on GitHub lets you try importing, searching, and reading your history. The Microsoft Store Edition is available if you want direct backup and its other features.
The YuGPT website explains the differences and includes screenshots.
If it saves someone another search for an answer they have already found once, I will have done something useful. I would be happy for it to become a tool someone remembers when they need it.
Top comments (0)