Apple Search Ads bid on day-30 revenue per install

Where the money goes: On the Apple Search Ads accounts I have inherited, brand and brand-misspelling keywords carried between 45 and 70 percent of spend. Nobody argues with that half. The argument is the other half, where cost per install reads acceptable and day-30 revenue per install does not, and the account owner cannot tell the two apart because Apple counts installs while finance counts subscriptions. I bid against day-30 revenue per install at keyword-set level, rebuilt from your MMP or your own postback endpoint, and I do not start until that number exists. How the account is built: Structure follows attribution, not a keyword taxonomy someone inherited. Brand sits in its own campaign so it can never flatter a return read on discovery. Exact match campaigns are grouped by the intent behind the search rather than by theme, because someone typing a competitor name is a different buyer from someone typing the job your app does, and they deserve different destinations. Search Match runs in one contained discovery campaign whose only output is next week's keyword and negative lists. Custom product pages are assigned per intent group, and the assignment is the bet: if the offline-use intent group lands on a page whose first two screenshots are not about offline use, the bid is beside the point. The measurement problem, specifically: SKAdNetwork and AdAttributionKit give you coarse or fine conversion values inside a postback window, and most schemas I am handed spend all 64 fine values on install-plus-registration. I write the encoding: which revenue and retention buckets occupy which values, which window each tier locks, and what the coarse fallback means when volume falls under the privacy threshold. You get the mapping table as a document your engineers implement and your analysts read. Where volume allows, lift is read from a geo holdout, a matched set of countries or states with store search off for a defined period, because the alternative is quoting the platform's own attribution back to your board. What I do not do: No ASO. I do not write or test store metadata, screenshots or preview videos, and I will say plainly when a keyword is losing to a screenshot rather than to a bid. No Google App Campaigns, which is a different auction with a different measurement surface, and I do not pretend one skill covers both. No creative production. No fee as a percentage of media spend, because that fee argues against the recommendation to switch a campaign off. Who should not buy this: Apps under roughly 3,000 monthly installs across all channels, where the privacy thresholds will hide the data this method needs and you would be paying me to read noise. Apps with no in-app revenue event and no MMP, where install counting has cheaper practitioners. Teams who want daily bid changes as a deliverable: I change bids weekly, sometimes not at all, and I write down why either way. How the engagement runs: The first two weeks are diagnosis: conversion value schema, structure rebuild, negative baseline, and a written list of the keyword sets I intend to stop funding with the revenue evidence behind each. After that it is a weekly decision cycle, one bid and budget pass, one keyword and negative pass, one written note of what changed and what it was expected to do. Monthly, a read on whether store search is still the marginal channel worth the money. When it stops being that, I say so.