Case Study · VA Disability Claims · Meta Lead Gen
How We Cut Cost Per Lead 37% While Nearly Doubling Monthly Lead Volume
In February 2026 we took over Meta advertising for a VA disability claims consulting firm. Strong product, real demand, and an account with no way to tell what was working. Four months later cost per lead was down 37%, monthly lead volume had nearly doubled, and spend had scaled past $100K a month without losing efficiency.
-37%
Cost Per Lead
$17.83 down to $11.23 blended since takeover
+89%
Monthly Lead Volume
From ~3,410 to ~6,460 leads per month
27,371
Veteran Leads
Generated since takeover on $307K+ in spend
The Challenge: A Fragmented Account With No Signal
Most underperforming lead-gen accounts don't have a targeting problem. They have a creative and structure problem, and no system for figuring out which variable is actually moving the number.
This client is a performance-based practice with an in-house medical team that helps veterans file and maximize the disability ratings they have earned. A genuinely differentiated, no-win-no-fee offer in a category most veterans find slow and confusing. The demand was real. The account it was being advertised through couldn't get out of its own way.
Spend was fragmented across more than half a dozen overlapping lead-form campaigns, cost per lead had settled around $17.83 and was drifting upward, and there was no controlled way to tell which creative, script, or form was responsible for a good week or a bad one.
Blended cost per lead under the prior structure, compared with the blend since takeover and the most recent full month.
The Core Problems
- Fragmented structure. More than six overlapping lead-form campaigns split the budget and competed for the same veterans, starving Meta's algorithm of the clean signal it needs to optimize. Cost per lead on the highest-spending campaigns had climbed past $18.
- No creative testing system. Legacy static and video assets ran indefinitely with no controlled read on what actually drove a lead. Volume without direction.
- An untested conversion surface. The account ran on Meta instant forms, and the form itself, one of the highest-leverage and most overlooked levers in lead gen, had never been tested.
- No isolation of variables. Because creative, audience, and forms all changed at once, there was no way to attribute a result to any one of them. Wins couldn't be repeated and losses couldn't be diagnosed.
The overlap was measurable. The largest national campaign delivered most heavily in Texas and Florida, the two states that also had their own dedicated geo campaigns running at the same time. Two of the account's campaigns were buying the same veterans simultaneously in both states. The three biggest national campaigns also delivered against the same age bands, 45 to 54, 55 to 64, and especially 65 plus, across the same national footprint. Same objective, same age profile, same geography.
The mandate was straightforward: consolidate the account, build a repeatable testing system, drive cost per lead down without buying junk volume, and prove it quickly.
The Strategy
Isolate One Variable, Then Compound the Winners.
We rebuilt the account around one principle: never test more than one variable at a time, then stack only what has been proven. The fragmented campaigns were collapsed into a single prospecting campaign built as a controlled testing rig, with separate ad sets isolating avatars, scripts, forms, and formats, running against both a defined veterans audience and an open, broad-targeted audience. The open audience ultimately produced the lowest cost per lead, and budget concentrated on the ad sets combining the winning avatar and form. Inside that structure, we ran a deliberate sequence.
Isolate the Avatar
We held the brand's single best-performing script completely constant and ran it across a slate of presenters. Because the script never changed, the only variable being measured was the presenter, which surfaced the winning avatar cleanly with no creative noise muddying the read. The two top performers were an older Caucasian man and a younger African American woman: demographically opposite profiles that intuition would never have paired. That is the entire case for isolating the variable. The winners weren't guessable, and an account running on assumptions would have missed both.
Expand the Scripts and Angles
With the avatar locked, we took the proven script as a template and modeled new variations on its structure: news hook, supporting stat, urgency, objection flip, team introduction, low-friction CTA. At the same time we tested entirely new angles against it. We were deepening a proven format and hunting for the next breakthrough at once: volume with direction.
Optimize the Conversion Surface
Because the brand ran Meta lead forms, the form was its own lever, so we treated it like one. We tested form architectures head to head: background-image, hero-image, review-carousel, bullet-point, and paragraph layouts. Most accounts stop at the ad. The form turned out to be one of the highest-leverage variables in the build.
Compound the Winners
Finally we combined the winning avatar, the winning script and angle, and the winning form into a single configuration. Because each element had already been validated in isolation, stacking them drove efficiency rather than reintroducing guesswork.
Close the Loop With Backend Data
We integrated the client's Zoho CRM directly with Meta so down-funnel events flowed back into the platform: not just form fills, but qualified leads, discovery calls scheduled, and signed contracts. That changed what the account could optimize toward. Instead of chasing the cheapest form submission, every avatar, script, and form could be judged on whether it produced leads that actually advanced. Cost per lead only matters if the leads convert, and the integration lets us steer budget toward the creative and audiences generating real cases rather than the lowest cost per click.
The Breakthrough: The Compounding Effect
Through February to April, as we scaled spend roughly fourfold and ran aggressive tests, cost per lead moved from about $9 in the ramp to a peak of $12.61 in April. Still well below the $17.83 baseline, but rising as we pushed volume and pressure-tested new variables.
The proof came in May. With the winning variables combined into one configuration, we pushed the campaign to its highest spend of the entire period, $101,350 in a single month, and cost per lead dropped back to $10.46, roughly 41% below the legacy baseline, while generating the most leads of any month. Highest spend, lowest cost at scale, most volume, all at once. That is the signature of a system that compounds instead of fatiguing, and it is the difference between buying a good month and building one you can repeat.
Spend scaled roughly 4x from February to May 2026 while cost per lead held in a tight band and fell to its lowest at-scale point in May.
Leads generated per month, February to May 2026.
Why It Worked
A Testing Engine, Not a Lucky Creative.
One Variable at a Time
Every test produced a clean, attributable read. We always knew whether the avatar, the script, the angle, or the form was responsible, so winners could be repeated and losers diagnosed.
The Form Is a Lever, Not a Formality
Treating the lead form as a dedicated CRO surface unlocked efficiency most accounts leave on the table by defaulting to whatever form they started with.
Consolidation Over Fragmentation
Collapsing six-plus competing campaigns into one gave Meta the clean signal it needs and gave us the budget control to scale decisively.
The System Is Repeatable
This wasn't a lucky creative or a one-month spike. It was a controlled testing engine, isolate then validate then compound, that can be rebuilt on any lead-gen account facing the same problem.
Figures cover February through May 2026. Cumulative lead and spend totals run from the February takeover date.
Ready to Build Results Like These for Your Brand?
We turn down brands we don't think we can move the numbers for. Tell us about yours and we'll be honest about whether we're a fit.