arontg asked up there whether reviews affect lead quality and not just rankings. I've got numbers for that, so let me just put them down.
I pulled every roofing company I could find in Dallas, Houston and Austin on Friday night. 12 searches, four different phrasings per city, 506 unique listings after dedupe.
Ratings are useless as a filter. 88.3% sit between 4.5 and 5.0 and another 4.5% are 4.0-4.4, so 92.8% of the market is inside one star band. That's why WebProfitLab's thing up there works. When everything shows 4.8, the only part left worth reading is the one to three star reviews and what the owner wrote back, and that's a much smaller sample than the rating column pretends it is.
Review count is where it actually splits:
- under 50 - 272 (53.8%)
- 50-99 - 86 (17.0%)
- 100-399 - 112 (22.1%)
- 400+ - 36 (7.1%)
Median is 39. Half this market is small enough that the owner still picks up his own phone, and the ones past 400 have an agency or somebody doing marketing, so you'd be somewhere around the fourth person ringing them that week.
So yes, a company with 50 reviews probably does close better, but not because of the reviews. They're a proxy for company size wearing a review-shaped costume, and what you're really buying is an owner who still answers his own phone.
What I don't have is MoMarketing's recency angle. Counts, not dates. Recency probably beats everything I measured and I can't see it without opening profiles one at a time, which at 506 is not happening. If somebody here pulls first and last review dates in bulk I'd like to know how.