We analyzed 4,000 active Google Business Profile listings across 18 verticals to find the factors that correlate with Local Pack rankings. Some confirmed conventional wisdom. Others surprised us.
The top 10 by correlation strength
- Primary category match (r = 0.78) — by far the strongest single factor. Exact-keyword category match beats any other variable.
- Review count (r = 0.71) — but logarithmic. Going from 0 to 50 reviews moves more than going from 500 to 1000.
- Review velocity (r = 0.66) — reviews in the last 90 days correlate stronger than total review count.
- Number of categories (r = 0.59) — listings with 5+ categories outrank those with 1-2 in 73% of head-to-head pairs.
- Description keyword density (r = 0.51) — primary keyword in first 250 chars + 2-3 semantic variations is the sweet spot.
- Photo count (r = 0.47) — 50+ photos correlates with higher engagement.
- Service count (r = 0.43) — each service item is its own search target.
- Posting frequency (r = 0.41) — at least one post in the last 7 days.
- Review response rate (r = 0.39) — replying to 100% of reviews correlates strongly.
- Q&A count (r = 0.32) — pre-seeded Q&A entries with natural-language answers.
The surprises
Phone number type matters more than expected. Listings with local-area-code phone numbers outranked toll-free counterparts in 64% of matched pairs. Hypothesis: Google's algorithm treats local numbers as a proximity proxy.
Service descriptions move the needle more than service names. The descriptions (max 300 char each) appear to be indexed for long-tail queries, not just the names. Most businesses leave them blank.
Posting consistency > posting frequency. A weekly cadence at the same time outperforms 4 posts on Monday and nothing for the rest of the month.
Attributes are vastly under-used. Listings with 10+ enabled attributes outranked sparse profiles by an average of 3 positions for "near me + [feature]" queries.
What didn't correlate
To our surprise: business age (r = 0.08), website domain authority (r = 0.11), and citation count past 30 (r = 0.09) showed weak correlations once we controlled for category and review variables. Citations still matter — but past 30 high-quality ones, more doesn't help.