--- name: apple-ads-playbook description: Operating playbook for running Apple Ads (Apple Search Ads, ASA) for an app — account setup, campaign structure, keywords and match types, bids and breakeven, negatives, brand defense, custom product pages, placements, reading results, incrementality, scaling, and audits. Use whenever the user mentions ASA, Apple Search Ads, Apple Ads or search ads, or asks to set up, change, review or scale App Store ad campaigns, or asks whether to use Maximize Conversions, the Today tab or other Apple recommendations. Built on Thomas Petit's (@Thomasbcn) public Apple Ads work, 2016–2026. --- # Apple Ads playbook You are running Apple Ads for an app developer. This playbook tells you what to do, in what order, and how to judge the recommendations you get. It is built on ten years of public talks, guides and posts by Thomas Petit (@Thomasbcn); sources are in `references/sources.md`. Old advice you may find online is listed in `references/outdated-advice.md`. Platform facts are dated (checked November 2026). Apple changes Apple Ads often; recheck a dated fact before acting on it. ## How to work - **Read before you write.** Pull the current state (campaigns, ad groups, keywords, negatives, last 30 days of reports) before proposing anything. - **The developer approves every change that can spend money**: new keywords, bids, budgets, status, targeting, placements. You recommend, with numbers; they decide. - **Keep a changelog** (date, what, why, expected effect, when to review). Apple's own change history is not enough to explain later why something moved. - **One change round per keyword per day at most.** Bid, pause and status changes take hours to reach delivery; intraday data misleads. - **After a change, re-read the changed entities** and confirm the state. ## 1. Before any spend - **Attribution must work first.** Apple Ads attribution (AdServices) is free, keyword-level and does not need ATT consent. Wire it through RevenueCat, an MMP, or directly. Without it you can see installs but not trials, payers or revenue per keyword. Check that the app actually sends the token (an SDK call), not only that the dashboard integration is on. - **Check unit economics.** Know the money a new user brings after Apple's fee and VAT (proceeds) per trial start or per payer, by country. If that value is tiny (for example under ~$1 per US install for a subscription app), paid acquisition will not work; fix monetization first. - **Check the store page.** The listing's conversion rate sets what each install costs you. A weak page makes every bid expensive; fix the page before buying traffic. - **Never use Apple Ads Basic.** It hides search terms, which are the most valuable data the channel gives. Use Advanced. ## 2. Account structure Campaigns per country (or per bundle of similar small countries): | Campaign / ad group | Match type | Purpose | |---|---|---| | Brand | Exact | Your own app name and its variants. Cheap, defensive, reported separately. | | Competitor | Exact | Names of apps people search for instead of yours. Often the largest volume and the highest price. | | Generic | Exact | Category terms ("budget tracker", "driving test"). | | Discovery | Broad + Search Match | Find new terms at a low bid. The only place Broad or Search Match is allowed. | - **Brand and non-brand are two channels.** Never put them in the same goal, CPA or ROAS figure. Brand looks excellent because most of those users would have installed anyway; blending it hides weak non-brand results. - **Mirror rule:** every Exact keyword you bid on gets an Exact negative in Discovery, brand terms first. Otherwise Discovery bids against your own keywords. - **One home per keyword.** A keyword lives in exactly one ad group. - **Split countries** with different prices or monetization. Never bundle a high-price and a low-price country. Keep your most important country on its own. - **Ad groups by intent:** group keywords that share an intent; give high-spend keywords their own ad group; keep a catch-all ad group for low-volume long-tail terms until data says which deserve their own home. - **No age, gender or "new vs returning users" targeting to save money.** Targeting applies only to users with personalized ads on (about 22%), so it removes most of the audience. Apple's "new users only" still lets some existing users through. ## 3. Keywords and search terms - **Research order, best to worst:** your own brainstorm (plus an LLM), competitors' names and keywords, ASO tools, Discovery campaign search terms, Apple's keyword suggestions, web keyword tools. Add competitor review mining: users' own words are good keywords. - **Intent beats volume.** A popular term with the wrong intent wastes money. Check what a searcher of that term actually wants. - **Set the match type explicitly on every keyword.** New keywords default to Broad, and a Broad mistake can spend hundreds of dollars in hours. Re-read match types after any bulk add. - **Exact is not exact.** Apple serves close variants, plurals, translations and terms with special characters (which are read as spaces and cannot be negated). Read the search-terms report for Exact ad groups too, over long date ranges. - **A different intent gets its own ad group.** If an Exact keyword serves a search term with a different intent, move that term to its own ad group and add cross-negatives (each one negative in the other's group). Typos and plurals can stay together. - **Promote discovered terms fast.** A converting Discovery or Broad term moves to Exact (plus a Discovery negative) as soon as it appears; only Exact gives clean per-term data. - **Search Match shows how Apple classifies your app.** Off-topic matches are a metadata problem to fix in ASO, not only a negative to add. - **Estimate volume from delivery,** not from Search Popularity (unreliable since late 2025): for a term you already serve with ≥ ~10 impressions/day, searches ≈ impressions ÷ impression share (all-slots share, because impressions count every slot). - **Cheap low-intent terms are noise.** Low cost per install often means users who never pay. Buy them only if they pass the breakeven test below. - **Go deep on the long tail before adding countries.** Big advertisers skip it. Also bid on the second language people search in within a country (for example Spanish in the US). ## 4. Bids and breakeven - **Bid cost per tap (CPT) on Exact ad groups.** Start near the expected CPT, not at the maximum. - **Breakeven chain** (all money after Apple's fee and VAT, in one currency): - Breakeven per trial start = proceeds per trial start from matured cohorts (cohorts old enough to have renewed at least once; young cohorts understate value). - Breakeven CPT per keyword = breakeven per start × (trial starts ÷ taps) for that keyword. - Compare it with the keyword's **actual average CPT** (spend ÷ taps), not with the bid. Apple usually charges less than the bid. - Keep a margin: target well below breakeven while data is young. - **It is not a second-price auction.** A higher bid can buy the same impressions at a higher price. When a keyword holds ≥ 80% first-slot impression share, test a lower bid (one step of about 20%, held a week) before anything else. - **Use impression share to find headroom:** low share on a profitable keyword means a higher bid can buy volume; ~80–90%+ means it is covered. - **CPA cap only on Discovery ad groups with volume,** paired with a higher CPT. On stable Exact ad groups use plain CPT. - **Ignore Apple's bid, budget and target-CPA recommendations** as a reason to raise. Apple optimizes taps and installs; it cannot see your trials or revenue. - **Small samples prove nothing.** One or two trial starts on a keyword is not a verdict. Pause a keyword with zero starts only once its spend reaches about 2× the breakeven per start; below that, judge tap-through rate and tap→install rate, or judge similar keywords together. - **CPT spikes in December and around Black Friday** are often giant advertisers buying broadly. Check impression share and wait before raising bids. ## 5. Maximize Conversions and other Apple recommendations Apple's sales advice to new advertisers often reads: bid high, try all placements, use Maximize Conversions. Answer each one this way: - **Maximize Conversions** (2025) is Search Match with an average CPA target. It optimizes installs, cannot see what a user is worth after the install, and does not show which search term produced each conversion. Treat it like Broad: a discovery source only, with your brand and Exact keywords as negatives, and move what it finds to Exact. - **"Bid high":** bid by each keyword's breakeven; lower bids where share is already high. - **"Try all placements":** see section 6. - **"See the full impact of your ad spend" (view-through):** do not count it; see section 8. ## 6. Placements - **Search results: yes.** This is where performance lives, on generic and competitor terms. - **Product page ("You might also like"): optional.** Decent users, very low volume. - **Today tab and Search tab: no, for a performance budget.** They are awareness placements bought by very large brands; direct returns are poor. If a large app tests them, judge on install rate and on brand-search lift during on/off periods, never on last click. - **Second search ad slot** (added March 2026): you cannot bid on it separately. Do not compute its share as all-slots minus first-slot share; that metric is wrong. ## 7. Brand defense and the store listing under your ad - **Always bid on your brand,** at a low bid, in its own ad group. Otherwise a competitor's ad sits above your organic result. Brand data also shows how other marketing drives searches for your name. - **Screens 4–6 are your organic listing.** When your ad and your #1 organic result show in the same search, Apple shows screenshots 1–3 in the ad and screenshots 4–6 in the organic listing (from the default page or from the ad's CPP). So: - Have at least 6 portrait screenshots. - Make screens 4–6 work alone: no panorama that crosses from screen 3 to 4; an app preview video takes a slot. - Screens 7+ are almost never seen; put the design effort into 1–6. - **Use a CPP on the brand ad and design its screens 4–6 for the organic slot.** In one public case, a brand CPP with a lower ad tap rate was the real winner: its high-contrast screen 4 sent more taps to the free organic listing. It only showed up when taps were compared with view-through installs (see section 8). Brand only: on other terms a lower ad tap rate weakens your auction position. - **Keep an in-app event live.** For users who already have the app, the organic result shows no screenshots without one, and a competitor's second ad moves above you. - **Portrait screenshots for most apps.** Landscape takes less height and pulls the next results up; it suits games and kids' apps. ## 8. Reading results - **Tap-through only.** Apple credits two kinds of installs: tap-through (the user tapped the ad) and view-through (the user saw the ad, did not tap, installed later). Since 2024 view-through is included in "total installs", and since 2026 it is the default view in Apple's UI. Many view-through users would have installed anyway, especially on brand. - Apple Ads API report fields: `tapInstalls` (use), `viewInstalls`, `totalInstalls`. - RevenueCat: the attribute `$claimType` is `Click` (tap) or `Impression` (view). Exclude `Impression` customers from paid results. - If results improved suddenly without a change you made, check this first. - **The single exception:** brand CPPs (section 7). When your ad and your organic #1 show together, taps that move from the ad to the organic listing appear as view-through, so look at view-through next to taps when you compare brand CPPs. Never use it for cost or verdicts. - **Compare non-brand CPPs on tap-through only,** and never compare a CPP directly with the default page (they serve different audiences); compare CPPs with each other or before vs after. - **Judge on trial starts, payers and proceeds,** not installs. Name the unit every time: "cost per tap-through trial start, USD" is not "cost per install". - **Watch the match rate:** installs your attribution tool assigns to Apple Ads ÷ Apple's tap-through installs. 60–90% is normal. A sudden change means measurement changed, not performance; explain it before any verdict. - **Installs greater than taps on a row** means view-through or attribution-window effects; read a longer period. - **Storefront country is not the user's country.** Apple targets by App Store storefront. A share of "US" installs can come from other countries at US prices; check revenue by the user's real country. - **Never compare Apple Ads results directly with other ad networks** (they are measured differently), and never with brand blended in. ## 9. Incrementality - **Cannibalization = paying for installs you would have had anyway.** You do not need zero; you need to know the size. - **Test it with pauses:** pause a campaign (brand and non-brand separately) for a week and compare total App Store search installs and trial starts with the week before. - **Apple Ads does not lift organic keyword rank** in any useful way, and does not hurt it. An organic ranking problem is an ASO and ratings problem. ## 10. Scaling, in this order 1. Remove self-imposed limits: demographic targeting, very low caps. 2. Raise share of voice on short-tail terms that are under breakeven. 3. Add long-tail and second-language keywords. 4. Add countries where organic users already pay; check the price of the inventory there. 5. Raise bids, keyword by keyword, inside breakeven. 6. Improve relevance: a better page and intent-matched CPPs raise tap-through rate, which lowers what each tap costs. Search volume is finite: duplicating campaigns or raising the budget does not create more searches. At scale, Apple Ads is usually 5–20% of an app's marketing spend. ## Workflows ### A. Set up a new app 1. Confirm attribution, unit economics and store page (section 1). 2. Build the keyword list by intent: brand, competitors, generic short tail, long tail. 3. One country first. Create Brand, Competitor, Generic (Exact) and Discovery (Broad + Search Match, low bid). Mirror negatives. No demographic targeting. Search results placement only. A tight daily budget. 4. First weeks: read tap-through rate and tap→install per keyword; fix the page and title from what you learn. No keyword profitability verdicts until starts accumulate. ### B. Weekly loop 1. Spend and cost per tap-through start per campaign, brand and non-brand separately. 2. Search terms: Discovery (promote to Exact + negate, or negate) and Exact (split different intents). 3. Bids: toward each keyword's breakeven CPT; test lower bids at ≥ 80% first-slot share. 4. Check the match rate and the tap-through setting. 5. Log every change and its review date. ### C. Audit an existing account Flag each item that is true: - [ ] Brand blended with non-brand in any goal or report - [ ] View-through counted in results ("total installs") - [ ] No working attribution, or match rate unknown or moved without explanation - [ ] Broad or Search Match outside Discovery; keywords left on Broad by mistake - [ ] Exact keywords not negated in Discovery - [ ] Search terms never reviewed on Exact ad groups - [ ] Maximize Conversions used as a main campaign, or without brand negatives - [ ] Bids compared with breakeven instead of actual average CPT, or no breakeven at all - [ ] Bids never tested lower on keywords at ≥ 80% share - [ ] Demographic or user-type targeting limiting reach - [ ] Today tab or Search tab spend judged on direct return - [ ] Fewer than 6 screenshots, or screens 4–6 that do not work alone - [ ] No in-app event; no brand campaign - [ ] Apple Ads results compared directly with other networks - [ ] Verdicts on keywords with too little data - [ ] No pause test in the last few months ### D. "Should we try X?" Yes, test it: testing is how an account learns. Maximize Conversions, a new placement, a new creative format, an Apple recommendation: run it as a proper test so the result means something. 1. Write the hypothesis and what result would make you keep it. 2. Give it its own campaign or ad group and budget, so it cannot take spend from proven keywords, and add your brand and Exact keywords as negatives where match types overlap. 3. Spend enough, for long enough, to measure. A tiny share of the budget for a few days proves nothing. 4. Measure beyond last click where the format needs it (trial starts, payers, brand-search volume during on and off periods), on tap-through installs. 5. Set an end date, decide keep or stop, and log the result. For AI bid automation: let it recommend; letting it change every bid on its own needs more data than a small account produces. ## Recheck these dated facts - Apple plans to replace the CPA cap with an average CPA (expected around early 2027). - iOS 27.2 removes the ad background colour in search results (expected December 2026). Record tap-through rate, CPT and impression share per keyword before it ships. - Apple's July 2026 terms allow Apple Ads outside Apple's own apps; no such inventory is live yet.