How did ASAPilot identify the CPI drift cause?

The first audit revealed three converging issues: 38 zero-conversion keywords had accumulated across the 4 apps over 8 weeks (none had been added as negatives), Search Match exposure had drifted on 2 Brand campaigns (responsible for 22% of "unexplained" spend), and the Discovery ad groups still pointed at default product pages instead of game-specific CPPs (10-15 point CR opportunity).

Were the changes made automatically?

No. ASAPilot is read-only by default. It surfaced findings with prioritized impact estimates; the studio operator reviewed each and executed the changes in the Apple Search Ads dashboard manually. Total operator time across the 7 days: ~4 hours, vs the 30+ hours that manual analysis would have required.

Did the CPI improvement hold?

Yes — 60 days post-rollout, average CPI was stable at $3.20-$3.40, well below the original $4.85. The studio attributed sustained performance to the new automated routines: Daily Summary + Weekly Zero-Conversion Sweep + CPI Anomaly Alert, which together caught two further drift events within 12 hours of onset.

How a Mid-Size Mobile Game Studio Dropped CPI by 32% in 7 Days With ASAPilot

Note: This case study describes a representative workflow pattern observed in multiple ASAPilot customer accounts. Studio identity, app names, and exact figures have been anonymized at the customer’s request.

The studio

A mid-size mobile games studio based in Northern Europe, 12 people total — 1 UA operator, 3 game designers, 5 engineers, plus product/design support. The studio had 4 live casual games on iOS and Android, with combined ASA spend averaging $85K/month across the iOS portfolio.

The UA operator was solo on paid ASA. The team had previously tried a campaign-management tool with autonomous write mode and got burned — a misconfigured rule paused a critical ad group during a holiday event. They had reverted to manual management and were running ASA from the native dashboard.

The problem

Q1 2026 performance review revealed CPI drift across the entire iOS portfolio:

Quarter	Average CPI	Total spend	Installs
Q4 2025	$3.45	$245K	71,014
Q1 2026 (through March)	$4.85	$222K	45,773

The same total spend produced 36% fewer installs in Q1. The lead operator had spent the prior 6 weeks attempting manual diagnosis with no clear root cause — every individual campaign looked “normal” in isolation, but the aggregate had drifted.

The decision was forced: cut budget by 40% (which meant slowing the user acquisition ramp), or find a tool to find what manual review was missing.

The implementation

The studio adopted ASAPilot on the Growth plan ($99/month) on March 28, 2026. The 7-day intensive rollout:

Day 1: Connect + first audit

Connected all 4 apps via ASA OAuth (4 × ~2 minutes).
Ran the app-first parallel audit. Total time: 47 seconds across the portfolio.

The audit surfaced 7 prioritized findings ranked by estimated monthly impact:

#	Finding	Estimated impact
1	38 zero-conversion keywords accumulated, never added as negatives	$4,200/month
2	Search Match enabled on Brand campaigns of App A and App C	$2,800/month
3	Discovery ad groups pointing at default product pages	$1,900/month
4	Same keyword in 2-3 ad groups (internal competition)	$1,200/month
5	Today Tab campaign with stale creative	$900/month
6	4 keywords with CPT >> ad-group average and CR <25%	$850/month
7	Audience refinement over-constraining App B’s Discovery	$600/month

Total estimated monthly waste: ~$12,450.

Day 2-3: Negative keyword cleanup

Operator added all 38 zero-conversion keywords as negatives across the relevant ad groups (2 hours). Apple Search Ads dashboard interface; ASAPilot provided the prioritized list.

Day 4: Brand campaign Search Match cleanup

Operator disabled Search Match on the 2 Brand campaigns of App A and App C (45 minutes including verification). Replaced with dedicated Discovery ad groups that absorbed the exploration role.

Day 5: CPP build + assignment

The studio’s design team built 3 game-specific Custom Product Pages — one per casual sub-vertical (puzzle, match-3, idle). Submitted for App Store review.

CPPs cleared review by Day 7 evening. Operator assigned them to the relevant Discovery ad groups (60 minutes).

Day 6: Internal competition cleanup

Operator deduplicated keywords appearing in multiple ad groups — chose the higher-CR location for each, removed from the other. ~30 keywords affected (1 hour).

Day 7: Top-CPT keyword bid adjustment

4 keywords with CPT ≥2× ad-group average and CR <25% had their bids lowered by 25%. One was paused (relevance too weak).

Day 7 evening: Automations enabled

Three automations went live:

Daily Summary at 9am — pacing + flagged anomalies across all 4 apps
CPI Anomaly Alert — sigma-based per app, Slack delivery
Weekly Zero-Conversion Sweep — Mondays, surfaces negative candidates

The outcome (7-day measurement)

By Day 8, the portfolio metrics:

Metric	Day 1 (baseline)	Day 8	Change
Average CPI	$4.85	$3.30	-32%
Daily spend (steady-state)	$2,800/day	$2,750/day	-2%
Daily installs	577	833	+44%
Search Match share of spend	38%	19%	-50% relative
Discovery CR	28%	41%	+46% relative

The Discovery CR lift was the largest single contributor — CPPs aligned to keyword intent lifted tap-to-install conversion from 28% to 41%, which compounded with the negative cleanup to produce the CPI drop.

60-day sustained results

The studio maintained the new automation cadence for 60 days. Sustained metrics:

Window	Average CPI	Note
First 7 days post-rollout	$3.30	Initial impact
Days 8-30	$3.25	Held steady
Days 31-60	$3.20	Slight further improvement

Two anomaly detections during the 60-day window:

Day 22: CPI Anomaly Alert fired on App B’s Discovery (CPI jumped 2.4σ above baseline overnight). Cause: a new competitor entered the auction on 3 core keywords. Operator lowered bids 15% within 4 hours; CPI normalized within 36 hours.
Day 47: Zero-Conversion Sweep surfaced 9 new negative candidates from Search Match drift. Operator added them; Search Match share dropped 8 points.

Neither incident would have been caught within the first 24 hours under manual review.

What worked

Audit prioritization was the unlock. The studio had been doing the right kinds of analysis manually but had no way to rank findings by impact. The audit’s ranked list let the operator allocate the limited 4 hours of execution time to the highest-impact issues first.

Custom Product Pages were the biggest single lever. The 13-point CR lift on Discovery ad groups contributed more to CPI improvement than the negative cleanup. The studio had been planning to build CPPs “soon” for 4 months without a forcing function.

Automations enforced the routine. Manual weekly Search Term Report reviews had been slipping; the Weekly Sweep automation made them automatic.

What was harder

CPP design took longer than expected. App Store review for 3 new CPPs ran 48-72 hours per page. The studio submitted on Day 5 and assignment happened Day 7-8.
First week of CPI alerts was noisy. Sigma thresholds had to be tuned per app — default 2σ was too sensitive for low-volume App D. The studio standardized at 2.5σ for apps below $20K/month.
Operator training overhead. Learning the AI chat interface took ~3 days. After that, query speed exceeded dashboard navigation by 2-3×.

Recommended for

This pattern works best for:

Game studios spending $30K-$200K/month on ASA — high enough volume to make 30% CPI shifts meaningful, low enough to be manageable by a single operator.
Solo or 2-person UA teams managing 3-10 apps.
Teams that have plateaued on manual optimization and need an external lens.

It is less applicable for single-app studios under $5K/month (free tier sufficient) or agency-scale 80+ account portfolios (Agency plan + different workflows).

See pricing for the Growth plan used here, or read the audit guide for the structure of the Day-1 audit findings.