What did this campaign set out to do?

A benefits fund in California wanted its eligible members statewide to sign up for coverage, and it had to connect with English and Spanish speakers from a single budget. Success meant a registration at the lowest sustainable price, steady enough that one fixed spend could cover the state for the full flight.

What is connected TV and why buy it through MNTN?

Connected TV is advertising on streaming television, bought the way digital media is bought: by audience, with response measured rather than assumed. MNTN is one of the platforms that makes that inventory buyable and reportable alongside search and social, so a streaming placement can be judged on the same cost per registration as everything else in the account.

Why did the Spanish-language ads get so much cheaper?

Early Spanish spend leaned on video and streaming, where the cost of a registration stayed high. The Spanish-speaking audience the fund needed was reachable far more cheaply on Meta. Once Spanish creative ran there, the price dropped sharply and the conversion rate climbed. The fix was the channel the language was bought on, and the creative was largely already in hand.

Did the falling cost come from better creative or from budget changes?

Mostly budget changes. Each month, spend shifted out of the expensive segments and into the cheap ones, whether the difference was a channel, a market, or a language. Meta's arrival and a couple of strong videos helped, but the repeatable engine was reallocation against an honest per cut cost on a monthly cadence.

How do you decide which geographies to fund on a fixed budget?

You balance two jobs. Big markets carry the population needed to hit a volume target, even at a high cost per registration. Small markets often convert much cheaper but cannot reach the goal alone. The split keeps enough spend in the large markets to make volume and hands the rest to smaller markets while their cost stays low.

CCPU Case Study: A Bilingual Campaign That Cut CPA 93%

In June, a registration on this campaign cost $35.12. By October it cost $2.43, a 93% drop over five months. The offer, the landing page, and most of the creative stayed exactly where they were. All I moved was the budget, every month, toward whatever was converting.

This was a 2024 paid media engagement for a health care benefits fund in California. The job was to get eligible members across the state to register for their benefits, in both English and Spanish, out of one fixed budget.

Working families were entitled to coverage many of them had never claimed, and the campaign had to reach them cheaply enough that a single flight could cover the whole of California.

Over six months, May through October, the account spent $121,455, served 4.94 million impressions, reached more than 350,000 households, and produced 7,175 registrations at a blended $16.93 each. The blended number is the least interesting figure here. The story lives inside the months.

The media ran across four channels. Connected TV, which is streaming television advertising bought the way digital is bought, audience by audience with response measured, placed through a platform called MNTN. YouTube. Search. And later Meta.

Each channel reaches a person in a different moment, and each one prices that moment differently, which turned out to be the whole game.

The efficiency curve

Here is the whole walk, month by month: $35.12 in June, $8.50 in July, $7.21 in August, $4.41 in September, $2.43 in October. Every month I read the account by channel, by market, and by language, pulled budget off whatever was registering members expensively, and pushed it toward whatever was registering them cheaply.

The blended number fell every single time.

The steepest drop is June to July, before Meta arrived, as spend left costly connected TV prospecting. Meta's August entry held the curve down rather than starting it.

Look at where the steepest drop falls. June to July, the number went from $35.12 to $8.50, the single largest move of the whole flight.

Meta had not launched yet. Nothing new had been added to the account at all. The spend simply came off expensive connected TV prospecting and concentrated where registrations were already landing.

The biggest efficiency gain came from spending the same budget in better places, before a single new channel went live.

Meta arrived late and changed the math

Meta launched in August, the last channel to come online. From its first full month it registered members between $6 and $9 each. Google search and YouTube were running $13 to $19 over the same stretch, and connected TV higher and far more variable.

So the closing months of the flight became an exercise in feeding the cheapest channel as fast as it could absorb budget without its cost creeping up.

Meta entered late and undercut every other channel from its first full month, which is why the closing weeks of budget moved toward it as fast as it could absorb them.

Eleven markets did most of the work

California splits into fifty three media markets, and the money was never spread evenly across them. Eleven markets carried about 80% of the spend and returned about 82% of the registrations.

Los Angeles was the engine: $48,449 in spend for 1,691 registrations at $28.66 each, more volume than any other market by a wide margin. San Joaquin County, far smaller, registered members at $12.74, under half the Los Angeles rate.

The market that makes your volume and the market that makes your efficiency are rarely the same one. A fixed budget has to buy both.

That gap is the argument for watching geography every week. A market the size of Los Angeles has to stay in the mix because it is where the people are, but a fund with a fixed budget cannot let expensive volume markets eat the whole allocation.

The cheap markets lack the population to hit the registration target on their own, so the work is a balance: enough of the costly big markets to make the numbers, and as much of the cheap small ones as they can absorb.

Spanish was the whole story

The campaign ran dedicated English and Spanish tracks, roughly 78% of budget to English and 22% to Spanish.

Early on, Spanish looked like a tax. In June it registered members at $55.36, well above the English rate and well above the blended average. The easy read would have been to cut it.

The Spanish audience sat on Meta the whole time, priced at a fraction of what streaming charged to reach the same people.

What Spanish actually needed was a different channel. The costly early Spanish spend was flowing into connected TV and video, where the price of a registration stayed stubborn.

Once Spanish creative reached Meta, the cost fell off a cliff. By October a Spanish registration cost $3.69, and the Spanish conversion rate that month hit 27.92%, meaning nearly three in ten people who clicked went on to register.

The collapse tracks Spanish creative reaching Meta. By October nearly three in ten Spanish clicks turned into a registration, a rate the video channels never came close to.

The lesson travels well past this fund. The Census Bureau put Hispanic residents at 19.5% of the country's population in 2023, while Nielsen reported that less than 1% of digital ad spend from U.S. online retailers went to Spanish-language websites in Q1 2025.

Bilingual audiences often get treated as a translation problem, where the English plan is rebuilt word for word in Spanish and bought on the same channels. But a language is also an audience, and audiences sit in different places and respond to different formats.

The Spanish speakers this fund needed were on Meta, and cheap, well before the plan went looking for them on streaming television. Choosing the channel per language, rather than translating the whole plan wholesale, is what turned Spanish into the efficient track.

Same faces, different price per channel

Four testimonial spots carried the campaign, each built around a single presenter. Call them Alicia, Carmen, Martha, and Natasha. The interesting part is how differently the same person performed depending on where she ran.

Martha scaled beautifully on YouTube: 2.4 million impressions at an $18.50 registration cost, the top video on the platform. The same faces on Meta ran between $6 and $10. A spot that was a solid YouTube performer became a cheap Meta performer, and a spot that struggled in one place could carry another.

That is why I never rank creative on one leaderboard. A video's cost per registration is a fact about the video and the channel together, so the same asset earns a different verdict in each place it runs. The scoring has to happen per channel or it talks you into cutting winners.

The same reasoning drives the creative scoring system I use to keep good assets alive in the channels where they actually pay.

What actually produced the 93%

The engagement was scoped for three months and extended to six once the early results came in. Across the flight the blended registration cost fell from $35.12 in June to $2.43 in October, about 93%.

That number came from moving budget every month toward the channels, markets, and language cuts that were converting, and away from those that were falling behind.

Better creative helped. Meta helped most of all. But the mechanism underneath it was monthly reallocation on an honest per cut cost number.

This is the same discipline at the core of my growth operating system: pick one honest efficiency number, watch it per channel and per segment, and move money toward what pays. It is also what a growth audit goes looking for, the distance between where the budget currently sits and where the results are coming from.

On this account that distance was wide in June and nearly closed by October.

If you are spending across channels and languages and suspect the budget is not sitting where the results are, that gap is exactly what I get brought in to close. Let's talk.