Business Club Varna2 years, 140+ members, zero ad spend
Ad cost per signed case from $6,750 to $2,750 for a Brooklyn injury firm with spend down $3,500 a month
Halloran & Reyes, a 4-attorney injury firm in Downtown Brooklyn, had a site a legal-marketing vendor built in 2022, $9,000 a month in Google Ads pointed at it, and a case count that had not moved in 2 years. Between Q1 2025 and Q1 2026, signed cases traced to Google Ads went from 4 to 6, the ad cost per signed case fell from $6,750 to $2,750, and the firm spends $3,500 a month less. The answer was in 2 places nobody had looked: a tag that counted page views as phone calls, and a mobile speed score on the last line of every quarterly report.
About the client
Halloran & Reyes Injury Law is a 4-attorney personal-injury firm on Court Street in Downtown Brooklyn, founded in 2016, taking car, truck and slip-and-fall cases on contingency from Brooklyn, Queens and Staten Island, with a signed case worth low five figures in fees a year or more after intake. Google Ads had run since 2022 through the vendor that built the site, at $9,000 a month of media (our fee sits outside every number here), which in a market where a click on "car accident lawyer Brooklyn" costs $100 to $150 is a modest budget. The vendor's report said leads were up; its last line every quarter carried a mobile PageSpeed score of 23 that nobody at the firm had read, and the 2 intake staff privately thought the lead count was too high.
Personal injury law
Brooklyn, New York
6 people
2016
Our analysis of the situation
Intake speed was the first suspect, the usual leak in law, and the CallRail log answered it: calls answered within 2 rings, an answering service after 7 p.m., every voicemail returned the same day. Search terms came next and held up, with off-topic queries at about 9% of clicks, and geography, 3 counties, presence only, had nothing to hide.
So we counted the leads ourselves, matching Google's 81 reported conversions for the quarter against CallRail and Clio Grow, and got 60: the phone-call tag fired on page load rather than on tap, so for 2 years every visitor who opened a page with the number on it had been logged as a call, and the vendor's report and the platform agreed because they were the same number. At $27,000 for the quarter that was $450 a real lead and, with 4 of 60 signing, $6,750 a case. Then the site: mobile PageSpeed scored a 23, mostly an autoplay video and a 4 MB hero image on shared hosting, while 7 in 10 ad clicks came from phones, and every search landed on the same page about the firm rather than the visitor's accident.
The figure we would be judged on was cost per signed case, reconciled monthly against Clio Grow, and we left intake alone.
Supporting research
Clio's Legal Trends Report prices a 5-minute response at roughly 8 times the conversion of a 30-minute one [1], the standard intake was checked against. Pareto Legal's 2026 benchmark found 84% of 13 plaintiff-side firms could not attribute 3 quarters of their signed cases to a channel [2], the polite version of the same problem. Portent's 2019 site-speed study across 100 million page views found B2B lead-generation pages converting about 3 times better at one second than at 5 [3].
Our strategy to solve the problem
On 7 April the phone conversion tag in Google Tag Manager was rebuilt to fire on a tap, not on page load; Fluent Forms submissions were routed through a server-side GTM container; GA4 conversions were kept as a separate count; UTM parameters were standardised; and a monthly reconciliation against CallRail and Clio Grow began. The tracking fix did more than anything else we changed, and it changed nothing a visitor could see. Once the tag fired on a tap and the forms came through the server, the account was fed 20 real leads a month instead of 27 mixed with page views, and Smart Bidding stopped chasing the visitors most likely to open a page and started chasing the ones who called. It took about 6 weeks to relearn on that little data, and through late April the account looked worse than the one we had inherited.
A week later, on 14 April, the autoplay video came off the homepage, images went through ShortPixel, WP Rocket ($59 a year) took over caching and deferred scripts, and hosting moved from GoDaddy shared to Kinsta ($35 a month). Mobile PageSpeed went from 23 to 61.
On 19 May 12 neighbourhood-and-case pages went live, built in Elementor on the same install: car, truck and slip-and-fall for Bay Ridge, Flatbush, Sunset Park and Williamsburg, with the ad groups matched to pages one to one. A redirect map went live the same day, and the 2 pages that already ranked kept their URL, H1 and load time and were retested after launch, so the traffic the old site had earned was kept.
The pages were the second change. Each one opens with the visitor's situation: what happens after a crash on the Belt Parkway, what the insurer's adjuster will say in the first call, what a free case review involves, and then who the attorneys are. None of them quotes a settlement or a verdict. New York's Rule 7.1 allows prior results only with the disclaimer that they do not guarantee a similar outcome, and testimonials only within limits the firm did not want to spend its budget policing, so the pages make no claim about outcomes at all. Every page carries the "Attorney Advertising" label, and we keep dated copies of each version, as the rule requires. Marisol Reyes, the other name on the door, read every page against 7.1 before it went live.
On 2 June the contact form was split: a 4-field "start my free case review" form for new enquiries and a separate contact link for existing clients, so a new visitor is no longer asked for a case number they do not have.
When the pages were in place we paused the head terms, in August: "car accident lawyer Brooklyn" and its variants at $100 to $150 a click, with the budget going from $9,000 to $7,000. A person searching "hit by a car Flatbush" at 11 p.m. lands on a page about Flatbush and calls, and that click costs a third of the one from the generic search. In November, after the lead count had held for 2 months at the lower spend, the budget went to $5,500.
From the firm we needed read access to CallRail and Clio Grow, a list each month of which leads had been signed, Brendan's sign-off to remove the video he had chosen, and Marisol's time on the compliance read.
Before and after
Before is January to March 2025; after is January to March 2026. Winter is the firm's busiest quarter for car cases, so a summer-to-winter comparison would have credited the ice to the work.
| Metric | Q1 2025 (before) | Q1 2026 (after) | Change |
|---|---|---|---|
| Monthly ad spend | $9,000 | $5,500 | −39% |
| Ad spend, quarter | $27,000 | $16,500 | −39% |
| Leads as Google Ads reported them | 81 | 66 | not comparable |
| Real leads (unique call, chat or form, verified in CallRail and Clio Grow) | 60 | 66 | +10% |
| Cost per real lead | $450 | $250 | −44% |
| Lead-to-signed rate | 7% (4 of 60) | 9% (6 of 66) | +2 points |
| Signed cases traced to Google Ads | 4 | 6 | +50% |
| Ad cost per signed case | $6,750 | $2,750 | −59% |
| Mobile PageSpeed score | 23 | 61 | +38 points |
| Sessions from ads | index 100 | index 97 | flat |
Sessions held flat, and that was deliberate. The money bought roughly the same number of clicks from people whose search already named their neighbourhood and their accident, and a page that loaded before they gave up. 2 cautions. The counts are small: 4 and 6 signed cases is the right order of magnitude for a firm at this budget in this market, and a single case either way changes the rate by a point and a half. And we cannot separate the speed work from the pages; speed went live in April and the pages in May, and holding the pages back 3 months to get a clean read would have cost the firm a quarter of its budget on the old page.
In the before quarter the intake desk and the vendor's report disagreed every month with nobody to arbitrate, and the firm paid $6,750 in ads for each case it signed, some months signing one. Now the report and the phone log say the same thing, the 2 intake staff take about the same number of calls, and each signed case costs the firm less than half what it did on $3,500 a month less.
Summary
A tag that fired on page load had inflated the lead count by a quarter for 2 years and trained the bidding on the wrong people, while a 23 on mobile speed meant most of those visitors on phones left before the page finished, and the ones who stayed got a page about the firm instead of their accident.
We ruled out intake and keyword targeting with the firm's own call log and search-term report, then found the gap by matching every reported conversion to a real call or form by hand.
Honest counting, a faster site on Kinsta, 12 neighbourhood-and-case pages with the ad groups matched to them, and the head terms paused.
6 signed cases in the quarter instead of 4, at $2,750 in ads each instead of $6,750, on $5,500 a month instead of $9,000.
References
- Clio, Legal Trends Report. clio.com
- Pareto Legal, "Law firm PPC statistics 2026". pareto.legal
- Portent, "Site speed is (still) impacting your conversion rate" (2019). portent.com
Final word from the client
2 more cases a quarter, fine. What gets me is the 23. It was on the last line of the report for 2 years and I never asked what it was. I'd stopped asking about the cases because the answer was always that leads were up, and it turns out a quarter of the leads were people looking at our phone number.