Search Terms Triage for Apple Search Ads: a Weekly Workflow That Actually Improves CPI

Running Apple Search Ads well isn’t about “finding a keyword.” It’s about continuously curating what queries you’re willing to pay for. The Search Terms report is your evidence—but only if you sort it into actions you can execute.

Below is a simple, repeatable weekly triage workflow designed for indie iOS teams using Search Results placements and a mix of Exact, Broad, and Discovery/Search Match.

H2: Start with a query “intent” rubric (so you don’t react emotionally)

When you open Search Terms, you’ll see lots of strings that look similar. The trick is to categorize each query by what the user seems to be trying to do.

Use four buckets:

• A) Buy/Install intent (high relevance): queries that include category + “app” phrasing, clear problem terms, or direct use-case language.
• B) Compare/alternative intent: queries that look like “X vs Y”, competitor names, or “best X” style discovery.
• C) Research/ambient intent (low relevance): broad category terms without clear action, vague needs, or tool/feature words with no “app” intent.
• D) Mist/No intent: spelling variants, unrelated categories, or queries that clearly don’t match your value prop.

You don’t need perfection. You just need consistency so your bid changes aren’t random.

Quick rule of thumb

If your screenshots and product page don’t immediately satisfy the “why this app?” of the query, it’s probably not A.

H2: Add two “cost signals” before you touch bids

Search Terms give you clicks, but clicks aren’t the full story. For each row (query), look at:

• TTR (taps ÷ impressions): Are you earning taps from relevant-looking traffic?
• Conversion rate (installs ÷ taps): Is your store turning those taps into installs?

Then combine that with CPI/CPA (depending on your tracked event). You’re trying to answer:

• Is the query getting taps efficiently but converting poorly? (store mismatch)
• Is it getting impressions/taps but CPI is high? (auction pressure + relevance)
• Is it not earning taps? (irrelevant intent or weak ad/product alignment)

Don’t immediately change bids just because you see a single bad week. Triage is about direction, not punishment.

H2: Weekly workflow (30–45 minutes)

Step 1: Pick your review window

Use a rolling window like the last 7 days (or 10 if you have low volume). Then focus on queries with at least one of these:

• a minimum number of impressions (so TTR isn’t a guess)
• a minimum number of taps (so conversion rate isn’t a coin flip)

If a query has near-zero volume, park it until it gathers data.

Step 2: For each query, decide one action

Use the rubric + signals to choose exactly one of these actions:

1. Promote (keep + raise intent alignment): move spend toward better relevance by increasing bid (or shifting match type toward Exact)
2. Maintain (no change): it’s doing its job or doesn’t have enough data
3. Reduce (cap CPT bid): lower bids to reduce wasted taps while still learning
4. Block (add as negative): stop paying for this query entirely

A good practical setup is:

• Bucket A: promote or maintain
• Bucket B: maintain first; promote only if conversion supports it
• Bucket C: reduce bids (or add negatives if it persistently underperforms)
• Bucket D: add negatives quickly

Step 3: Change bids conservatively so you don’t “nuke volume”

For every CPT adjustment, prefer small steps and guardrails:

• Lower bids by a modest percentage (so you don’t accidentally starve the campaign)
• For promotions, raise bids only for queries that show at least reasonable conversion

Because Apple Ads uses a CPT auction, lowering too aggressively can remove you from relevant auctions entirely. If you’re profitable, preserve learning volume.

Step 4: Add negatives for waste-flowing queries (with match design)

Negatives matter most for Broad keywords and Search Match (Discovery) where match expansion can drift.

When you add a negative, do two things:

• Use the right negative match type based on how you want to block behavior.

  • If the intent is truly unrelated, block broadly enough to stop the leak.
  • If you want to block only a specific phrasing but keep close variants, use a narrower negative match.

• Avoid over-blocking competitor or category terms unless you’re sure they’re not converting.

Practical tip: If a query consistently lands in Bucket D (no intent), add it as a negative for the campaign/ad group where it’s being generated. Don’t sprinkle it everywhere.

Step 5: Re-bucket queries into Exact over time (only when they deserve it)

A common improvement loop is:

1. Let Broad/Search Match discover queries.
2. Once a query reliably shows good conversion, promote it by creating an Exact keyword targeting that same query.
3. Reduce reliance on Broad for that behavior.

Why this works:

• Exact gives you tighter control.
• You keep Discovery for new opportunities.

Keep the learning system stable: don’t delete everything that’s currently getting volume—just shift bids gradually.

H2: The “three sanity checks” to avoid misguided optimization

Check 1: Is the problem actually the store?

If a query has:

• decent TTR, but
• weak install conversion

then you’re often looking at a product page issue (messaging mismatch, missing screenshots for the query’s promise, or pricing/subscription friction).

Before you cut bids, verify your product page answers the query in the first few scrolls.

Check 2: Are you comparing apples to apples on metrics?

Make sure you’re looking at the same attribution window and your revenue mapping is stable. If you’re using RevenueCat (or similar) to connect installs → purchases, confirm that your purchase mapping isn’t lagging or changing.

Check 3: Are you accidentally “learning” with one ad group and spending with another?

Remember: one campaign targets one country/region; ad groups hold keywords + bids. If you adjust an ad group but the query keeps showing up, you might have overlapping keyword coverage across ad groups.

In that case, your bid changes might not be affecting the query generator you think they are.

H2: How to structure your keywords so triage is easier

You’ll get better results if your keyword setup makes the Search Terms report more actionable.

A simple structure:

• One ad group for Exact winners (queries you trust)
• One ad group for Broad exploration (bounded by negative keywords)
• One ad group for Search Match/Discovery (for expansion)

Then triage actions map cleanly:

• promote → move to Exact ad group
• waste → add negatives to Broad/Discovery ad group(s)
• maintain → leave bids alone

H2: What to do when a query flips from good to bad

Sometimes queries decay after you increase bids or the auction landscape changes.

When you see a flip:

1. Check TTR first. If TTR drops, the traffic mix changed (relevance/auction winners changed).
2. If TTR stays similar but conversion drops, your store experience likely isn’t meeting the current audience.
3. Only then adjust CPT bids.

Don’t immediately add negatives during the first sign of trouble—sometimes it’s temporary auction noise.

H2: Keep a short “action log” (so you don’t repeat the same mistakes)

Create a note/table with:

• date range reviewed
• query
• bucket (A/B/C/D)
• action taken (bid down, bid up, Exact promotion, negative added)
• resulting CPI/TTR trend next week

After a month, you’ll notice which actions reliably help your account.

H2: Tie this back to what matters: ROAS is downstream of installs

One key limitation: Apple Ads doesn’t provide per-keyword revenue directly. You’ll typically infer quality through the install→purchase chain (often via AdServices token mapping into RevenueCat or your analytics).

So your Search Terms triage should optimize the leading indicators:

• TTR (relevance)
• conversion rate (store + user intent match)
• CPI (cost control)

Then confirm profitability using your downstream purchase metrics and cohorts.

H2: If you want help turning Search Terms into daily actions

If you’re already drowning in the report, that’s exactly where advisory tools can help. AdsBuddy’s workflow is to read your Apple Search Ads performance alongside your revenue signals and return a short, prioritized list of changes you approve (so you still control the “learning pace”).

Closing takeaway

Treat Search Terms like a weekly curation process, not a post-mortem. Sort queries by intent, use TTR + conversion to diagnose whether the problem is relevance or the store, then make one safe action per query—promote, maintain, reduce, or block with negatives. If you do this consistently, CPI stabilizes and profitable volume stops leaking into low-intent auctions.