How a Solo Founder Built a $30K/Month App with 5.2M Installs

Skip high-popularity head terms (e.g. 'bus times') that VC-backed rivals bid on; instead bid on city-level variants ('Chicago bus times', 'New York bus times'). Drop keywords with no impressions and ones with competitive bids; keep the middle. Exact match only (broad match wasted money). 3-4 related keywords per ad group, each ad group with its own custom product page. Start at $0.50-0.60 cost per tap and raise until ~100% impression share while ads run. Took a few months to find the profitable set.

For each ASA ad group (3-4 related keywords), create a custom product page that copies the standard 10 screenshots and swaps only the first, e.g. 'Live Bus Times' becomes 'Live Chicago Bus Times'. A screenshot template system keeps this fast.

Use Switchr (small web app saved to the home screen) to change the phone's App Store region; browse the Navigation/Travel top lists for that country and note the public-transit keywords popular apps use; type a few characters into App Store search and collect autocomplete suggestions (most people tap a suggestion rather than typing the full query); check candidates in Astro; then load them into Search Ads to test.

Instead of an MMP (AppsFlyer/Adjust), a self-built job (written with ChatGPT/Codex) runs daily: pulls Apple Search Ads reports via the API and RevenueCat's daily bulk export (RevenueCat uploads to an S3 bucket; one line of code in the app passes ASA attribution to RevenueCat), joins them on IDs and keyword names, and reports per keyword: impressions, average cost per tap, taps -> trials -> paid. A keyword qualifies for scaling if $1 of spend returns at least $1.20 within 3 months (3 months chosen because of mixed weekly/annual plans).

Rewrote the App Store description (up to 4,000 characters) to include a Q&A section, because the App Store web listing is the most authoritative page about the app and chatbots readily lift Q&A text. Questions are written as users ask them, per city ('What's the best way to find my bus in Chicago?'), and differ by storefront (US: New York, Chicago, SF, LA; Canada: Toronto, Edmonton, Vancouver).

Launched livebustimes.org (Sept 2025) giving away live bus/train times and route maps from the app's data. Hand-built plain HTML/CSS/JS (no frameworks) for speed and reliability, with invisible structured data on each page for search engines and chatbots. Submitted to Google Search Console and Bing (which ChatGPT search uses); a daily job regenerates the sitemap for millions of pages. Pages show ads for the app (seeing your bus on a map). Measured ChatGPT-User visits in Cloudflare's free analytics.

Through 2025, replaced ~25 near-identical translated screenshot sets with per-country versions: study each storefront's popular apps (via Switchr), copy elements, run each A/B test for 2 weeks to a month. Findings: Italy - widgets and live activities as the first screenshot; Netherlands - laurel wreath and testimonial; Germany - bigger text, shorter message; Japan - dense, information-packed screenshots with little white space.

Used Subkeep (web app) to reset prices per country using its blended average index and sometimes the Spotify Index (Big Mac Index also available), rounded to clean numbers, pushed straight to App Store Connect. A/B test for 3 months: 50% of users in every country saw Apple's default prices, 50% the adjusted prices. US, Canada and UK prices unchanged.

Set the price to $25.00 instead of $24.99 in App Store Connect and strip the '.00' in the app so it reads '$25'. Only for double-digit prices; single-digit ones like '$5' look odd, so the weekly stays $4.99.

Users install at a bus or train station, so onboarding must be short. After ~30 A/B tests, screens talk about benefits ('Get there on time, stress free', 'Never miss your stop again', 'Join millions saving time') with minimal cartoony UI, social proof (laurel wreath, numbers) on every screen, and a fifth screen that warms users up to the premium tier before the paywall. A long onboarding with many questions caused drop-off.

In August 2025, A/B tested annual/weekly/lifetime against the old annual/monthly/lifetime. Weekly ($4.99) is labeled 'Pay as you go, no commitment' and anchors the annual plan, which can then be shown as 90% off versus 52 weeks of weekly.

Anyone who dismisses the onboarding paywall still gets 7 days of premium (widgets, live activities, delay alerts). A post-onboarding flow lets them customize notifications and theme, and a Pro button (top right) offers video tutorials. Day 5: push 'you're on day five of your 7-day trial, tap to learn about the features'. Day 7: final push with a discount (on annual). Works because costs are mostly fixed servers; AI apps with per-user costs may prefer ads to cover free users.

A Pro button stays in the top-right of the home screen; it opens videos walking through the ~5-step widget install (home screen, plus button, search for the app, customize colors). Video beats written instructions.

Users can share their ETA/journey (e.g. a kid's bus to school with a parent) via a universal link. Recipients without the app see a website mirroring that feature, with an ad to download the app.

A RevenueCat webhook notifies his server when a subscriber cancels; the server sends push notifications 5 days and 2 days before the subscription ends, showing features they would lose (including ones added during the year they may not know about).

Not during onboarding or first launch: show the rating prompt the moment a user taps a bus and sees it moving on the map for the first time.

Locales like Mexican Spanish and Vietnamese are also indexed in the US store. Keywords are not combined across locales ('bus' in one and 'times' in the other won't index as 'bus times'). When the US English title/subtitle are full, move a medium-popularity keyword from the 100-character keyword field into the title/subtitle of an unused locale (e.g. Vietnamese, where the app isn't translated) to boost its US ranking. Stop doing this once you actually target that country.

Purple icon and bright purple screenshots, chosen because transit-category listings are mostly white/blue backgrounds and green icons; the goal is to stop people scrolling for a second. Screenshots are checked on a real phone in the App Store.

After ~7 years ad-based (AdMob, ~$10K/month pre-pandemic, then CPMs crashed at the same impressions), moved to subscriptions in 2021; ads remain only to cover free users' costs.

Use a promo campaign such as Indie App Santa / AppAdvice to get a burst of downloads, and show the rating prompt at the right moment, e.g. on the 2nd or 3rd use after the user completes a task, to collect genuine ratings.

Targeted long-tail phrases like 'buy old clothes' with popularity 5 (maybe 1-2 searches a month each) and reached #1 on 90 of them; a quantity game.

A paid-upfront app scaled Apple Search Ads using niche, highly relevant keywords. Whether the $1K-$10K/month refers to spend or revenue is unclear.