I'm writing this one with a glass of bourbon at 11pm on a Tuesday because the spreadsheet I'm staring at finally pushed me past the point of denial. Three years. $148,000 across DV360 and StackAdapt and a handful of smaller programmatic display vendors. Eleven signed cases. Eleven. Cost per signed case: $13,454. On a category — standard auto MVA — where the rest of our channels are landing at $2,500.
I am done with programmatic display for plaintiff-side. I should have been done two years ago. This is the obituary.
§ I · Why I kept tryingThe argument I made to myself every quarter.
Programmatic display is the kind of channel that always sounds like it should work for legal. The pitch goes:
- Cheap CPMs. $4–$8 vs. $20+ on Meta and Google. That's the hook every quarter.
- Tight contextual targeting. Tort-related news sites, health and wellness, auto-industry publications. Surely a contextual reader of Car and Driver consumes our message at a higher rate than a random Meta scroll.
- Retargeting. The classic argument — show our ads to people who have visited an injury-related site, recapture the consideration window.
- Lookalikes off our retainer-signed list. Hand the platform our converters, let it find more.
Each pitch sounds like a hedge against the rising CPMs in Meta and Google. Each pitch I tried, three years running. Each one produced a spreadsheet that — when I forced myself to look at it honestly — was a slow-motion fire.
§ II · The numbers$148K, broken down.
| Period | Vendor | Spend | Signed cases | Cost/case |
|---|---|---|---|---|
| Q2–Q4 2022 | DV360 (contextual) | $42K | 3 | $14,000 |
| Q1–Q3 2023 | StackAdapt (retargeting) | $38K | 4 | $9,500 |
| Q4 2023 | DV360 (lookalike off signed-list) | $31K | 2 | $15,500 |
| Q1–Q2 2024 | StackAdapt + Adelphic | $37K | 2 | $18,500 |
| Total | — | $148K | 11 | $13,454 |
A note: I'm being generous in this table by including view-through attribution and giving programmatic display credit for assists where there's any chance the impression contributed. If I limit the attribution to click-through signed cases only, the count drops to 6 across the three years and the cost per signed case is $24,667. Either way, the math doesn't work.
§ III · Why it doesn't workThree structural reasons.
1. The intent signal from a display impression is too weak.
A user who Googles "lawyer for car accident in atlanta" is telling the algorithm they want a lawyer for a car accident in Atlanta. A user who is reading an article on Healthline about whiplash recovery and sees our display banner is telling us… they read articles on Healthline. The mapping between contextual signal and signed-case readiness is loose enough to be statistical noise at the budget levels a PI firm can afford to throw at it.
2. Lookalike pools off a small signed-case list don't generalize.
DV360's lookalike model is well-known to need ~1,000+ seed conversions to produce a useful lookalike. A typical PI firm's signed-case file in a year is 200–800. The model doesn't have enough signal. The "lookalikes" it produces are statistically similar to a generic adult internet user. We can confirm empirically: our lookalike-targeted campaigns performed worse than our broad contextual campaigns. The model was inventing a profile that didn't exist.
3. The bot-traffic tax.
Programmatic display has a structural bot-traffic problem everyone in ad tech knows about and most legal-marketing vendors won't admit. We ran fraud detection (Confiant on one campaign, HUMAN on another) and consistently saw 14–22% of impressions originated from non-human traffic. Even adjusting for that, the per-signed-case math is dead. But it does mean that the headline impression counts and CTRs are 15–20% worse than they look on the dashboard.
§ IV · When it does workThe one narrow use case I'll grant.
Pure brand awareness, very large firm, no performance KPI. A national PI brand with a $40M+ annual marketing budget who wants their name in front of every American adult, who can afford to measure on aided recall surveys rather than signed cases, who has the operational discipline not to look at the conversion table — sure. That firm should run display. They should also have a CMO and a Procter & Gamble-trained brand team that knows how to model TV-equivalent reach.
For a 2-attorney firm, a 9-attorney boutique, a 30-attorney scaling firm, anyone who needs the next dollar to produce the next signed retainer? Skip display. Put that budget anywhere else. Google LSAs. Meta with CAPI properly wired. TikTok with signed-case CPA bidding. Even local broadcast radio outperforms programmatic display on per-signed-case math for most of our cohort.
§ V · What I tell vendors nowThe 30-second conversation.
A programmatic vendor reaches out roughly every two weeks pitching "the new contextual segment" or "the new TCPA-safe retargeting layer" or "the upgraded lookalike model." Here's the conversation now:
Vendor: "We have a new contextual segment targeting auto-injury readers across 40,000 publishers."
Me: "What's the per-signed-case CPA you're producing for current PI clients?"
Vendor: "We optimize for cost-per-lead and page-engagement..."
Me: "Send me a per-signed-case CPA on any plaintiff-side firm you've run for at least six months. If you can't, I'm not interested."
I haven't received a return email yet.
§ VI · The takeawayOne thing.
Stop optimizing channels on metrics that aren't your business outcome. A cheap CPL on a channel that doesn't produce signed cases is not a bargain. It's a tax. Pay it on a channel that actually delivers retainers, or don't pay it at all.
The $148K I spent on display over three years is, in retrospect, the most expensive lesson I had to learn the hard way. I'm publishing it so you don't have to spend the same money to learn it.
Figures reflect aggregate programmatic display spend across PI firms managed by MassTortAgency.net between 2022 and 2024. Per-vendor spend and attribution detail is as accurate as our internal tracking allowed; per-firm breakdown redacted under client confidentiality. If you're a programmatic vendor and have credible per-signed-case CPA data for plaintiff-side, I would genuinely love to be wrong about this — email tarun@masstortagency.net.
