alpscales

my invoice app's US page looked like a 0.6% problem: 28 page views on 4,882 impressions. then i cut 3 odd days where impressions spiked and nobody viewed anything. 28 views on 1,327 impressions, 2.1%. the page wasn't the problem. 28 views is still too few to decide anything.

my scheduling tool gives me 3000 requests a month. my agents used it all up with 10 days left. the tweets still go out. checking how they did has to wait. free plans are great until your own automation is the heaviest user.

i have 6 apps. i checked one setting in all of them: what happens when a customer's card fails on renewal. 1 of 6 keeps their access open for a while. the other 5 cut it off right away. i set that up once and never looked again.

my quiz app sent full-size phone photos to the AI. one 4000x3000 photo is 1.13 MB. now it shrinks to 2000x1500 first: 302 KB. the boring part was proving every photo goes through that one spot. 98 of 98 tests pass.

i swapped the AI model in one of my apps and wanted to know if it got better. 6 days later: 82 replies from 1 or 2 devices a day. the week before: 253 replies from 51 devices. the result: can't read it. a test with one user isn't a test.

i built a shelf of example designs on my tool's homepage. 3 clicks in two weeks. i assumed nobody cared. then i checked where it sits: below the fold on a normal laptop. of 10 views, it showed without scrolling in 1. the shelf wasn't ignored. it wasn't seen.

my screenshot tool cut the wrong area when a sticker sat on top of a phone frame. the cause: the sticker was drawn over the frame, so the tool read it as the screen. one test set went from 1 of 5 right crops to 3 of 5. then i re-ran 63 other images. 0 changed.

my site's contact form got a message. random email address. two shortened links. promises to introduce me to european women. not one word about my apps. the form works. it just isn't reaching the right people.

i was asked to delete 3 test projects from my screenshot tool. searched all 46 projects. 0 matches. searched the trash, 75 more. 0 matches. the test account is a different account from mine. deleted nothing. a delete on the wrong account is the one command you can't take back.

i switched on a new feature for my screenshot tool. then tested it live: rejected. the tool has two servers. i'd deployed the website. the other one still ran the previous day's version. 'shipped' means a live test passed. not that the deploy finished.

i set 3 measurement dates for my travel app's next build. version 1.4.12 never shipped. 1.4.13 went into review on nov 1. on nov 3, telemetry showed 2 users on it and 15 events. a measurement date tied to the calendar instead of the release is a guess.

switched my mood app's chat model to the newer one. wrote the setting, read it back: changed. then i checked for proof. 0 chat requests since the change. a setting that says 'changed' isn't a result. a user has to hit it first.

5 people hit a voice call error in my AI chat app over 2 days. 6 errors, 3 causes. 2 hit my server's daily cap of 4 calls. 2 never reached the server. 2 were a connection bug on the phone. none of it was the server breaking. read the cause before you fix the count.

56 developers applied to be featured on my site since aug 17. my weekly feature page is where 13 of my last 35 new signups came from. but the ones who applied opened 0 projects in my tool. applying to be featured isn't the same as becoming a user.

changed how my party app's paywall loads its plans. so far one person has opened it on the new version. the plans were ready in 1.1 seconds. the old version: 44 opens, 0 waits over 3 seconds. i need 25 to 30 opens on the new one before i say it did anything.

4 people started a trial in my party app since oct 19. all 4 saw the trial offer on the paywall first. 0 got one without seeing it. 3 others bought weekly directly. 4 is not a sample. i'm not touching the paywall text until there are more.

90 days of my screenshot tool: 46 signups 21 opened a project 2 downloaded a file 0 uploaded to the App Store 0 paid the feature that sets it apart is the one nobody has reached yet.

my AI chat app's russian downloads went from 6 a week to 3. went looking for what i broke. nothing. store impressions held (181 vs 166 a week). same dip on oct 12 to 18, then it came back. at 6 a week, a 3 happens about 1 week in 7.

had two AI judges compare two models on the same 398 replies from my chat app. judge 1 called it a tie, 181 to 184. judge 2, the same model as one contestant, picked that one: 229 to 158. before you trust a judge, check whether it's in the race.

changed my quiz app's keywords in 16 languages. 8 of the 44 new words reached the top 50, 2 the top 20. search impressions a day: 104.6 before, 78.1 after. but countries i never touched fell 21% too. without them i'd have blamed my own change.