Intent signals that predict ASA conversion — an analysis of 2M taps

We pulled 2 million anonymised Apple Search Ads taps across 14 markets and 6 app verticals to understand which search intent signals best predict first-purchase conversion — not just install.

Key Findings

Commercial intent beats branded intent for LTV. Users who search for generic category terms (e.g. "photo editor" vs "Adobe Lightroom") convert to paying users at a 23% higher rate on average. They're shopping, not brand-loyal.

Query length is a strong positive signal. Queries of 3+ words convert 41% better than single-word queries in our dataset. Longer queries indicate a user who knows what they want.

Negative space matters. Queries containing words like "free," "trial," or "alternative" predict lower LTV even when install CVR is high. We recommend separating these into a lower-bid ad group.

Market Variations

The signals vary significantly by region. In JP and KR, brand-name queries actually outperform generic queries for LTV — brand loyalty is stronger in those markets. In US and UK, generic intent wins.

Implications for Bidding

These findings directly inform how Keenbid classifies intent. Our model now uses query length, word-level commercial signals, and market-specific calibration to score every keyword before applying a bid recommendation.