April 2026 · Lazy Developer EP.16

3 Patterns Apsity Kept Catching Across My 12 Apps

It's been two months since I built Apsity. I built the dashboard in EP.02 and put the AI growth agent on top in EP.03. Now I just use it.

At first I thought more features meant better. Two months in, I figured out feature count doesn't really matter. What matters is what you catch.

This is a writeup of 3 patterns Apsity actually kept catching across my 12 apps. Not one-off incidents — things that happened multiple times over two months. For each pattern I wrote down the symptom, the cause, and what I did. Honestly.

Pattern 1 — When Downloads from a Specific Country Suddenly Spike

Apsity's Countries menu tracks downloads by country using a 7-day moving average. When the current day's value exceeds twice the moving average, an alert fires. The formula is simple. But simple works.

One time, an alert fired in App A's Countries showing '+312% from Brazil vs. last week.' I thought it was a bug at first. Brazil wasn't even in the top 10 countries for weekly downloads on that app.

I cross-checked against the raw App Store Connect numbers first. It was real. That's when I started looking for the reason. Reddit search, YouTube search, TikTok search. Usually the answer shows up within a week. That time, it was a single TikTok video in Portuguese.

This kind of thing happens one or two times a month. Doesn't sound often, but if you miss it, a trend that started for a reason quietly dies out. If you react right away, you can turn a country you never targeted into a key market. Brazil ended up in the top 5 countries for that app.

Pattern 2 — When Conversion Rate Drops Alone Right After an Update

This is the one that gets me most often. And the one I catch latest. The reason is simple: I was only watching rank.

Deploy a new build and one day later the numbers show up in App Store Connect. Impressions (views) unchanged, Installs down. Rank unchanged. If I'd had a rank-only habit, I would have missed this entirely.

Apsity's Overview first screen puts rank, impressions, and conversion rate side by side in one row. I made it that way on purpose. When I first built it I wanted to put 4, 5 metrics — eventually cut it to 3. You need them side by side to immediately catch when one metric moves alone.

This actually happened with App B. The day after an update I checked the Overview: rank unchanged, impressions unchanged, conversion -18%. I went straight to diagnosing the cause. Didn't look like an algorithm change. Algorithm changes usually move impressions too. When conversion moves alone, it's almost always my fault.

Turned out I'd swapped the first screenshot during the update. The new one I thought was an improvement. I reverted it. Conversion was back to the original number within 48 hours.

This pattern gave me one new habit. For 48 hours after deploying an update, I check the Overview morning and evening. Normally I check just once in the morning.

Pattern 3 — When the Review Rating Drops 0.3 Points in a Week

Reviews are the most direct signal. Also the most annoying one to read. App C normally held a 4.6 average. One Monday morning I checked and it was 4.3. Down 0.3 points in a week.

0.3 points is a lot. It means the new reviews over 7 days skewed heavily toward 1 or 2 stars. Normally 5-star reviews are the majority keeping the average up — a sudden skew means there's a common issue.

Too many to read manually. Just the last 7 days was about 80 reviews. So I went straight to Apsity's Reviews menu and ran clustering with Claude. The prompt went roughly like this.

Results came back in under 3 minutes. The top category was 'app freezes on a specific screen after the latest update.' Reading just two of the quotes, both pointed to the same screen. The bug location was basically confirmed.

I opened that screen and tried to reproduce it. Reproduced in 5 minutes. It was a layout bug that only occurred on a specific device (iPhone 12 mini). I pushed the hotfix that evening, and the following Monday the 4.3 was back to 4.5.

What the 3 Patterns Have in Common

After going through all three patterns, I learned something. Rank is a result. If you wait until rank drops before moving, it's already too late. To see the cause, you need other metrics. Countries, Impressions, Conversion, Reviews.

One more thing. The alert structure matters. Instead of me going through 9 menus every day, I need a structure that only notifies me 'when something unusual happens.' People go numb when they look at the same dashboard every day. Alerts prevent that.

So the three places I open most in Apsity are: Overview (3 metrics), Countries (spike alerts), Reviews (clustering). The other 6 menus I open only when needed. Checking all 9 every day was a fantasy.

FAQ

Q1. Can you catch these patterns even with just a few apps?
Yes. With 1 or 2 apps, manually checking every day might actually be faster. Past 4 apps, you can't keep up manually. I built the dashboard when I crossed 5 apps.

Q2. Can you catch these patterns with free tools?
App Store Connect + Google Sheets gets you partway there. But checking every day eventually wears you out. A structure where you 'only look when an alert comes in' holds up longer. That's why I built Apsity.

Q3. How did you decide the threshold for country download spikes?
If the current day's value exceeds twice the 7-day moving average, an alert fires. Simple formula, but it catches things reliably. I started at 3x and missed too much, so I lowered it. Going below 2x produced too many false positives, so I stopped there.

Q4. Can you use a model other than Claude for review clustering?
Yes. GPT-4o and Gemini produce similar quality. But for reliability in producing consistent JSON format, Claude Sonnet is the best. I pick Claude for the same reason elsewhere too.

Closing

Building Apsity was just the start. Two months of use taught me that what you catch matters far more than how many features a dashboard has. And you only figure out 'what to catch' by using it, not at the design stage.

All 3 patterns above weren't things I thought about during the design phase. I hit them repeatedly over two months and realized 'this always happens.' The dashboard has gradually evolved to catch these 3 things well.

In the next EP, I plan to dig deeper into implementing one of these (review clustering). What the prompt that processes 80 reviews in 3 minutes looks like, and how I validate accuracy.

If this was useful, you can try it yourself at Apsity. There's a 14-day free trial.