Marketing built for how people find you now.

At $23.74 per acquisition, the model didn’t work at any volume, and no amount of budget was going to fix a number that far off.

I started by separating the two problems hiding inside one metric. Part of the cost was paid inefficiency: wrong audiences, wrong creative, wrong landing pages. Part of it was structural. Every single customer was being bought, because nothing organic, referred, or earned was doing any of the work. Fixing only the first would have moved the number a little. Fixing both changed the shape of the curve.

On the paid side, the work was disciplined iteration: tighter audience definitions, creative tested against a single conversion event rather than engagement vanity, and landing pages rebuilt around one next step. On the owned side, I built the email and SMS lifecycle that turned a signup into an active user, then wired the whole handoff together in Zapier so a new lead was routed, tagged, and contacted without anyone touching it. Speed of first contact was doing quiet work on conversion rate the entire time.

The two moves that changed the trajectory were less conventional. The first was the Campus Director program: a nationwide team of student marketers running local events and grassroots recruiting on their own campuses. It solved both sides of a two-sided marketplace at once and produced acquisition costs that paid channels can’t touch. The second was a PR effort that landed coverage in AP News, Morningstar, and Yahoo Finance, which gave the SEO program something real to build on. Rankings tripled in two months and traffic went from 7,300 to 55,000 a month.

The result was a blended cost per acquisition of $4.36 and a product that became the company’s flagship offering.

An 80% bounce rate on a Contact page is unusual, because by the time someone arrives there they have already decided to talk to you. Losing four out of five of them means the page is undoing a decision the visitor already made.

Reading the page, two things stood out. The form asked for more than it needed, and every optional field is a small invitation to reconsider. It also sat below a long block of copy, so on most screens the visitor’s first impression was text rather than an action. Meanwhile the page did nothing to reassure a buyer who was about to hand security concerns to a stranger: no certifications, no client logos, no testimonials.

The complication wasn’t diagnostic. The owner liked the page, had built it himself, and wasn’t looking for a critique. So I stopped arguing and proposed a test instead. That is nearly always the faster route, because a split test moves the decision from taste to evidence and gives the person on the other side a way to change their mind without losing anything.

I built the alternative page and ran a Google Search campaign that split clicks evenly between the two. Sending paid traffic straight to a contact page is not something I’d normally do. It worked here because the campaign targeted hyper-longtail queries, the phrasing of someone looking for one specific service in one specific city. Intent at that stage is high enough that an intermediate landing page adds a step rather than removing one. Ad copy was written to match that intent exactly.

The variant cut bounce rate from 78% to 53% and lifted form completion from 14% to 27%. Same traffic, same spend, roughly double the inquiries. The report made the argument on its own, and the new page stayed.

A partnership with two professional teams is the kind of asset that gets announced and then quietly does nothing. The question was whether the credibility could be converted into direct sales of equipment at a price point where buyers deliberate for weeks.

I built the test around three KPIs and ignored everything else: cost per add-to-cart, cost per purchase, and return on ad spend. At a $750 target cost per purchase, engagement metrics are noise. An ad set can look excellent all the way to the checkout page and still lose money. Structuring the account around purchases from day one meant the optimization was pointed at the only number that mattered.

The build was 8 campaigns, 16 ad sets, and 32 creatives across the two markets, run for about six weeks. Two combinations separated from the rest at roughly $500 per purchase, a third below target. Both were in Kansas City, and both were reaching the same person: middle-aged, interested in both football and Olympic weightlifting, and not a site visitor in the previous twelve months. That last exclusion mattered more than it looks. Retargeting warm traffic would have produced a flattering cost per purchase by taking credit for demand that already existed. Excluding it meant every purchase in the test was genuinely new.

Google Display and Discovery ran alongside at about 10% of budget, enough to hold upper-funnel presence without diluting the read on Meta.

I scaled spend into the two winners and left the rest off. That configuration ran for a full year and was retired only when the team partnership expired, never for performance. Across the year, blended ROAS came in at 7.5 against 4.0 the year prior.