Guide · Apple Ads · ASO

Which Apple Ads keywords actually make money?

A keyword can bring installs without bringing paying customers. To decide where your next ad dollar should go, review organic rank, ad spend, and revenue over the same period.

This guide shows how to organise that review and choose a test: increase a bid, try a new keyword, or pause an expense that needs a closer look. Keep attributed revenue separate from estimates, and allow for trials that have not converted yet.

What you'll walk away with

1. A simple three-signal model: rank × ads × revenue
2. Four verdicts you can apply manually this week (even in a spreadsheet)
3. How cannibalization actually shows up in Apple Ads
4. A practical weekly review ritual
5. Where native tools (and Veldo) fit if you want the join automated

The trap: optimizing keywords in three different tabs

A normal indie workflow looks like this:

The problem is comparing different date ranges and levels of detail.

• Rank is daily and storefront-specific.
• Ads are campaign/ad-group/keyword and often looked at on 7- or 30-day CPA.
• Revenue is delayed (trials, refunds, tax) and rarely attributed back to the search term that started the story.

So people do the rational-but-broken thing:

• Scale keywords with cheap CPAs that never subscribe.
• Keep exact-match ads on terms where they already sit at #1–#3 organically.
• Cut keywords that look “expensive” but produce the only high-LTV users in the set.

Put the numbers on one row so you can review visibility, cost, and revenue together.

The three-signal model

Exact match first, because it is the only match type where the spend belongs to a single keyword and the row below can mean anything. If your campaigns are not split that way yet, start withmatch types and the campaign structure that uses them.

For every keyword (exact match first), collect one row:

Label revenue estimates clearly. App-level revenue divided by spend share is useful context, but it does not prove a particular keyword generated that revenue.

The only equation that matters (conceptually)

Keyword health ≈ f(visibility, demand, paid efficiency, paid→revenue)

• Visibility → organic rank trajectory
• Demand → popularity + impression share / tap volume
• Paid efficiency → spend vs installs
• Paid→revenue → spend vs attributed trials/subs

Rank alone can't tell you to pause. CPA alone can't tell you you're cannibalizing. Revenue alone can't tell you the keyword is even discoverable.

Four verdicts (use these labels)

Steal the language. Put it on a column. Act weekly.

1. Double down

Pattern

• Attributed revenue ≥ spend (or clearly on a path you accept)
• Organic rank stable or improving
• Demand real (popularity / volume not dead)

Action

• Raise test budget in controlled steps
• Protect the keyword in metadata (title/subtitle/keyword field; use a character counter)
• Consider a custom product page aligned to that intent (CPP + keywords)

2. Bid opportunity

Pattern

• Popularity high / clear demand
• Organic rank weak or missing
• Little or no exact campaign live

Action

• Launch exact match with a tight bid cap
• 7–14 day test, not a forever campaign
• Write metadata that matches the query intent before you scale

3. Cannibalizing

Pattern

• Organic rank already near the top (roughly top 3–5 for competitive terms; top 10 for long-tail; use judgment)
• Exact ad still spending meaningfully
• Total installs barely move when you pause a short test (or ROAS is mediocre because you're paying for users who'd tap organic)

Action

• Pause exact match for 7–14 days
• Watch total installs and trials, not only paid
• If total holds, keep it off or move budget to discovery/conquest terms
• If total drops hard, the ad was incremental; restore with a lower bid

Cannibalization is not a moral failure. It's a budget allocation problem. Paying Apple for a tap you would have won organically is optional.

4. Money pit

Pattern

• Spend continues
• Attributed revenue ~0 over a fair window (mind trial length)
• Rank not improving enough to justify “brand building” stories

Action

• Cut or slash bid
• Reallocate to Double down / Bid opportunity
• Only keep if you have a deliberate non-revenue goal (e.g., category defense) written down; vibes don't count

Sample ledger (illustrative)

This is the table Veldo is built around: same join, on your Mac, with a verdict instead of a hunch. You can start in Numbers today; automation is optional until the habit sticks.

Organic ASO and Apple Ads are one system

Metadata feeds both

Your title, subtitle, and 100-character keyword field influence:

• Organic ranking eligibility
• Which searches your product page is relevant for
• How Search Match and discovery features behave in ads

If ads teach you that breathing exercises converts and organics never show you for it, that's a metadata + creatives job, not only a bid job.

Custom product pages change the end of the funnel

With keyword assignment to CPPs, a converting paid/organic query can land on screenshots that match intent. That's conversion rate work sitting next to keyword work, so treat them as one loop. (See ourCPP keywords note.)

Brand vs non-brand

• Brand terms: usually protect with low bids; watch cannibalization vs pure defense.
• Category generics: where verdicts pay for themselves.
• Competitor conquesting: often ugly CPA; only keep if revenue (or strategic share) clears a bar you defined in advance.

A weekly 45-minute ritual (indie-sized)

Every Monday (or after weekend spend settles):

1. Export or open last 30 days of keyword spend (exact).
2. Refresh organic ranks for the same list + storefronts you care about.
3. Pull attributed revenue if you have it (RevenueCat cohorts are fine).
4. Tag each keyword with one verdict.
5. Execute at most three changes:
   • 1 pause test (cannibalization or money pit)
   • 1 bid increase (double down)
   • 1 new exact test (bid opportunity)
6. Write one line in a log: what you changed and what you expect in 14 days.

Indies lose less to bad strategy than to unreviewed automation. Cap the changes so you can learn.

Check what your revenue source measures

Before drawing a conclusion, check the reporting setup:

• Use customer attribution data when measuring keyword payback; campaign names alone do not attribute purchases.
• Conservative windows that respect your trial length (7-day trial ⇒ don't judge revenue on day 3)
• Separate “probation” for brand new keywords vs mature ones

If revenue join is impossible this month, still run verdicts onrank + spend + paid installs, and promote revenue join as soon as you have a source you understand.Read how RevenueCat attribution and Veldo estimates differ.

Tooling: spreadsheet vs suite vs native Mac

If you only need ranks, a focused tracker is enough. If you run Apple Ads and subscriptions, the join is the product.

For a wider map: Best ASO tools for indie developers (2026). For why native matters:ASO tools for Mac.

How Veldo encodes this playbook

Veldo is a native Mac app that:

1. Tracks organic ranks across 60+ storefronts
2. Pulls Apple Ads spend, installs, and Apple's popularity
3. Optionally joins RevenueCat revenue per keyword
4. Emits a first-match verdict with the evidence under it
5. Notifies when a verdict flips (watchdog)
6. Keeps acquisition data on your Mac (SQLite + Keychain)
7. Exposes read-only CLI/MCP so agents can answer “what am I wasting money on?” without mutating campaigns

Pricing on purpose: $8.99/mo or $99/yr. Pay for the tool, not a percentage of ad spend.

FAQ

Do Apple Ads help organic rankings?

Apple has said relevance and user behavior matter for search. Paid traffic can indirectly help if users install and engage, but buying a keyword is not a guaranteed organic boost. Treat ads as acquisition first; treat any organic lift as a bonus you measure, not a promise you bank on.

Should I pause exact match if I rank #1?

Often worth a timed test. If total installs and trials hold, you were renting your own demand. If they fall, the ad was incremental; bring it back at a saner bid.

How long before I call something a money pit?

At least one full trial + conversion cycle, and enough spend that noise isn't the story. For many subscription apps, 14–30 days of honest data beats gut feel. Brand-new apps may need longer for statistics, but still set a max burn.

Do I need RevenueCat?

No for rank/spend/install verdicts. Yes (or equivalent) if you want revenue-aware double-down vs vanity installs. Veldo works without it; revenue columns stay empty until connected.

Exact vs broad match?

Run verdicts on exact first. Broad is a discovery hose; it will muddy keyword-level truth. Promote winners from search terms reports into exact tests, and keep discovery and performance in separate campaigns so they stop bidding against each other.

How is this different from CPA optimization in Apple Ads Manager?

CPA optimization ignores organic rank and often ignores subscription revenue quality. A low-CPA keyword that never pays, or a mid-CPA keyword you already own organically, both look “fine” in Ads alone.

Next steps

1. Build the four-column ledger for your top 25 keywords this week.
2. Run one cannibalization pause test and one bid-opportunity exact.
3. Pack your metadata with a keyword field counter.
4. If you want the join automated on your Mac, get Veldo.
5. Prefer agents in your editor? See Veldo MCP.

Every keyword needs a next move. Rank is a signal. Revenue is the scoreboard.