The Right (and Wrong) Way to Grow with Google App Campaigns — Ashley Black, Candid Consulting

US starting budgets: $200-300/day on Android, at least ~$500/day on iOS (installs, inventory and CPMs are pricier). One campaign per geography; do not run duplicate campaigns doing the same thing (Google support blames cannibalization). Only split into separate campaigns when optimizing for different events (e.g. installs vs trial starts), since the optimization event is set per campaign. Give it 2-3 weeks to see how it will trend.

Google engineers' rule: no more than 20% change in any day to bids or budgets, up or down. Guest is more conservative: at most 20% every 2-3 days, because daily changes keep kicking the campaign back into learning mode. When she pushed beyond it, the system started spending wildly to find new inventory.

Google won't let you set budgets per ad group or A/B test creatives, so create ~5 ad groups per campaign, each built around one use case or feature (e.g. for a golf app: swing performance, caddies), with all creatives in the group matching that theme. Watch which groups get traffic and engagement. Common mistake: one campaign with one ad group, then concluding Google doesn't work.

YouTube has several hidden inventory pockets, so supply: a 6-second video (its own YouTube format), 10-15 seconds, ~20 seconds (YouTube Shorts), and 30+ seconds (skippable in-stream). Default to portrait, especially on iOS, but on Android include landscape since many people watch YouTube in landscape.

On Android aim for ~80% landscape images (landscape gets more traffic), 20% square/portrait. Skip fully designed image ads with text, CTA and rating: Google already overlays the app icon, an install button and often the store rating. Use free stock images found while building ad groups (or Google's AI asset generator) and see what gets traffic. Only design full images if CTR is really low.

Text assets are the most crucial part: search ads show only text, and Play Store placements combine the app icon with your text. Work within character limits and write them yourself; don't paste Google's recommended text (which ignores normal punctuation and grammar).

Use the 'networks' segment to see, per asset and ad group, the share of traffic going to search vs display (AdMob, mobile web, Play browse) vs YouTube; search partners should be small. Add CPM as a custom column: when performance drops, check whether CPM also dropped, since cheaper CPMs often mean worse traffic. Ignore Google's asset performance rating (it reflects scale and CTR, not your CPA) and don't hold IPM to one bar across networks. Check placement reports and block spammy sites/apps (blocks apply account-wide). Judge ultimately on CPA and down-funnel ROAS.

A few videos drove most of a client's volume but had very high CPA and poor ROAS. Cutting them all at once risks the campaign getting no conversions and dying, so she removed them slowly, waited a week, then pulled the rest.

Even in the best ad groups, ads fatigue. A tiny change, e.g. 'Buy it today' -> 'Buy today', is treated by Google as a new asset, sending it into fresh learning and new inventory, so the ad is less fatigued.

When a campaign won't spend more even with very high budgets (some set $100K/day), duplicate it: Google treats it as new and tests it, sometimes finding more scale than the steady original. Otherwise scale slowly; when one market is maxed out without aggressive bids, expand to other markets likely to perform.

New themes such as holiday creatives go in a dedicated, labeled ad group so they are easy to find and pause. For apps whose traffic spikes with natural disasters, run separate ad groups for hurricanes, tornadoes or avalanche season. New ad groups often get little volume because Google favors the first ad group created, so leave them on longer before judging.

For subscription apps don't optimize campaigns for purchases or trial completion: with a typical 7-day trial the signal reaches Google a week late. Optimize for an early event, usually start trial, and experiment with other early events.

Google pushes value-based bidding (tROAS), designed for IAP (and a variant for ad-monetized apps), not subscriptions. If your rep pushes it, run a short test rather than switching wholesale.

Reps are assigned by spend and by the billing country on the Google Ads account (thresholds are highest in the US, lower in smaller markets; e.g. a US founder billed via the UK gets a Dublin rep). Assignments change at quarterly 'book refreshes', so expect contact at the start of a quarter. Once assigned, make specific asks: reps can show breakouts such as google.com vs Play Store search traffic that the UI hides. Inbound support forms go to vendor teams that are rarely helpful.