84,435 Reviews Across New York's Injury Firms, 2.4 Stars Apart

The ratings barely differ

Sorted by star rating, the 235 firms we measured run from 5 down to 2.6. 162 of them sit at 4.8 or above, and 62 share the same rating once rounded to a single decimal.

A rating is a real signal. A firm sitting at 3.2 would tell you something loudly.

But within a competitive practice in a large city, almost every firm that stays in business clusters at the top of the scale. The ordering inside that cluster is mostly noise.

The unweighted average of those ratings is 4.75. Weighted by how many reviews each rating rests on, it is 4.85, slightly higher, because the firms with the most reviews are also rated a little above the middle. Both numbers are so close to the top of the scale that the useful question is not which firm is rated highest but which rating is standing on enough reviews to mean anything.

The counts differ enormously

That is where the spread actually is. TopDog Law has 6,289 reviews, and The Lenhardt Law Firm has 1.

One is 6289 times the other, and both firms take the same kind of case in the same city.

Most of that gap is offices, not popularity. A Google review attaches to a business listing, not to a firm, so a practice with 15 listings accumulates reviews at 15 addresses at once.

Across this set, the 68 firms with more than one listing average 722 reviews. The 167 with a single listing average 212.

A firm that has never opened a satellite office is not worse than one that has. It is counted differently. Any directory that ranks on raw review volume is quietly ranking on office count.

This is also why our per-firm totals are sums. We add every listing a firm operates and weight the rating by each listing's count, so a firm is described by all of its reviews rather than by whichever office Google decides to show you. The source line on each profile says how many listings went into the number.

What the score does about it

Client experience is pillar C of the LFL Certification Score, worth 20 of 100, and reviews feed the first two sub-factors. They are deliberately built to resist both problems above.

C1, authentic volume, is banded rather than linear. Past a few hundred reviews the sub-factor is maxed out, so the firm with 6,289 and the firm with a tenth of that score the same. Volume past the point of statistical confidence buys nothing.

C2, rating quality, is a Bayesian average rather than the raw star rating. Each firm's rating is pulled toward the count-weighted mean of the cohort, 4.85, with the prior weighted as 100 reviews.

A firm with thousands of reviews barely moves. A firm with forty moves a lot.

The arithmetic is the familiar shrinkage form, (v·R + m·C) / (v + m), and it means three reviews at a perfect 5.0 cannot outrank four hundred at 4.8. Every profile prints the result of that calculation next to the inputs, so you can check it.

What to read instead

If the ratings are all the same and the counts measure offices, the text is what is left, and it is more informative than the number sitting above it.

• Read the one- and two-star reviews first. Not to count them, since every firm with hundreds of clients has some, but because they are where the recurring complaint shows up. A pattern of "nobody returned my calls for three months" across several years is a finding. One furious review about an outcome nobody could have won is not. You will have to go looking for them: we measured what Google hands back when asked for a firm's reviews, and the middle and bottom of the star scale are almost absent from it.
• Check whether the firm replies. A firm that answers criticism publicly is a firm that reads its own reviews. That is our C4 sub-factor, and it is not yet assessed for any firm here, but you can see it yourself in ten seconds.
• Look at the dates. Two hundred reviews spread over eight years describes a stable practice. Two hundred inside one quarter describes a campaign. That is C3, recency and consistency, unassessed for 1 of the 235 firms in the directory, which is exactly why we say so on the profile rather than scoring it anyway. Where it is assessed, it is read from the sample Google returns, and that sample skews recent for every firm whatever the firm did.
• Read what the case was. A firm rated highly for handling straightforward rear-end collisions has told you nothing about how it handles a birth injury or a trucking case, which need different experts and much larger case budgets.

What this does not cover

One platform. Every figure above comes from Google Business Profile, because it is the only review source we have collected so far. Our own methodology asks for a cross-platform picture, and a single-platform average is weaker evidence than that, so the evidence line on every profile says "single platform, Google only" rather than leaving you to assume otherwise.

And reviews are not outcomes. A five-star rating records how clients felt about being treated, which is not the same as whether the firm won.

Outcomes sit in pillar B, which scores what a firm publishes about its own results rather than what anyone has checked: 93 of the 235 firms here publish a results page we could read, and none of those results has been matched to a court record, so no amount from them appears on this site. We counted what those pages carry, and that is the harder half of the work: it is where the directory is still thin.

Review counts and ratings come from the Google Places API (New), summed across every business listing we can attribute to the firm's own domain. A firm we cannot match to a listing keeps these sub-factors pending rather than receiving an estimate.