---
title: Optimising Apple Ads keywords for revenue, not installs
slug: apple-ads-keyword-revenue-optimization
summary: Why install- or CPA-optimised Apple Ads quietly funds keywords that never pay, how to judge each keyword on trials, cost per paid and predicted LTV (≥20 installs for a first signal, ~100 for conversion rates), what MMP fees cost per payer (~$1.40–3.50), automation rules tied to trials/subscriptions, and Adapty's claim of accounts scaled to ~$60K/month at ≥90% D31 ROAS.
domain: mobile-apps
tags: [user-acquisition, apple-ads, roas, ltv, keywords, attribution, mmp]
related: [apple-ads-attribution-without-mmp, apple-ads-single-keyword-ad-groups, app-break-even-cpi-and-revenue-per-install, apple-ads-measurement-pitfalls, paywall-by-acquisition-source, high-rpd-country-bundles-for-app-ads, apple-ads-brand-vs-non-brand]
sources: [yt:ZRVB9O3vOX4, yt:EYfeV7fkCeI]
updated: 2026-10-24
---

# Optimising Apple Ads keywords for revenue

## TL;DR
- Apple Ads reports impressions, taps and downloads, then **stops**. CPA/Max Conversion targets therefore optimise for cheap installs, and your worst keywords can look "green".[^1][^2]
- Adil Bilgi's test: list your **top 5 keywords by spend**. If you can't say what revenue each produced, you're running "a download machine", not an ad account.[^3]
- Judge each keyword on **trials → paid → cost per paid vs (predicted) LTV**. If cost per paid subscriber is below LTV, you're in profit.[^4][^5]
- **Low-volume keywords:** wait for ~**20 installs** for a first signal and ~**100 installs** to read trial/paid conversion. That takes more time, not more budget.[^6]
- Mykola Martynovets' MMP fee math: ~$0.07 per tracked install = **~$1.40 per payer at 5% install→paid**, ~$3.50 at 2%. At 20K installs/month that's ~$1,400/month (he said ~$70K/year).[^7]
- Apple Ads users usually have **higher LTV** than other channels because they arrive with intent. Check your own LTV by channel.[^8]

## Details
**Example of revenue-blind optimisation.** The same keyword in the US and Canada gave ~200 installs each; the US had 0 trials, Canada ~20. The cause was US pricing, invisible at install level (see [[high-rpd-country-bundles-for-app-ads]]).[^9]

**Bid/budget automation on downstream metrics (Adil).**[^10]
- Early on, target lower CPAs. Over time, judge on trials, subscriptions, ROAS and LTV.
- Example rules: *subscriptions up 30% without a budget change → keep raising that campaign's budget*. *A keyword taking 30% of spend but <10% of trials → cut its bid 5% per day until trials rise.*

**Balancing volume and efficiency.** For long-term players with stable budgets, Adil recommends steering by LTV rather than short-term CPA.[^8] With weekly/monthly plans, remember you pay once and earn on each renewal, so account for retention. The simplest workable target is **cost per paid vs predicted LTV**.[^5]

**Reported outcome.** Mykola says he saw accounts go from ~$1–2K to ~$10K and then ~$60K/month spend with **D31 ROAS ≥90%**, including in generative-AI apps.[^11]

**Value beyond ROAS.** Petit runs Apple Ads partly for the data: which competitors convert unusually well, and keyword insight for ASO (Google UAC discloses no keywords).[^12]

## Disagreements / caveats
- **Vendor framing.** This webinar sells Adapty's Apple Ads manager, which is free under $5K tracked revenue. The "MMP data tax" figure uses an assumed $0.07/install; MMP pricing varies and free tiers exist ([[skan-aem-capi-for-app-campaigns]]).[^7]
- Steve P. Young's lighter options (CPP per ad group, RevenueCat, weekly blended spend vs revenue) can also do keyword-level revenue with SKAG ([[apple-ads-attribution-without-mmp]]).
- Petit warns against comparing Apple Ads (AdServices attribution) with SKAN-measured channels at all ([[apple-ads-measurement-pitfalls]]). "Apple users have higher LTV" may partly be measurement and brand-intent bias. Split brand first ([[apple-ads-brand-vs-non-brand]]).

## See also
- [[app-break-even-cpi-and-revenue-per-install]] — setting the target
- [[paywall-by-acquisition-source]] — tailoring paywalls per keyword
- [[people/adil-bilgi]], [[people/mykola-martynovets]]

[^1]: yt:ZRVB9O3vOX4 @ 34:33 — Mykola Martynovets
[^2]: yt:ZRVB9O3vOX4 @ 35:28 — Adil Bilgi
[^3]: yt:ZRVB9O3vOX4 @ 45:30 — Adil Bilgi
[^4]: yt:ZRVB9O3vOX4 @ 37:06 — Mykola Martynovets
[^5]: yt:ZRVB9O3vOX4 @ 57:06 — Mykola Martynovets
[^6]: yt:ZRVB9O3vOX4 @ 1:04:08 — Mykola Martynovets
[^7]: yt:ZRVB9O3vOX4 @ 32:04 — Mykola Martynovets
[^8]: yt:ZRVB9O3vOX4 @ 59:00 — Adil Bilgi
[^9]: yt:ZRVB9O3vOX4 @ 37:06 — Mykola Martynovets
[^10]: yt:ZRVB9O3vOX4 @ 50:03 — Adil Bilgi
[^11]: yt:ZRVB9O3vOX4 @ 47:18 — Mykola Martynovets
[^12]: yt:EYfeV7fkCeI @ 33:23 — Thomas Petit