Snapshot
| Website development kickoff | April 1, 2026 |
|---|---|
| First paid channel live | May 26, 2026 (Google Ads), 8 weeks after kickoff |
| Channels live today | Google Ads, Reddit, Meta |
| Total paid spend to date | $120,305 |
| Cost per purchase reduction (first order) | 77% (week of 6/1 to week of 7/6) |
| Weekly spend scaled | ~6x over the same window |
The situation
The client, a LegitScript-certified telehealth compounding pharmacy, needed a paid acquisition engine built from nothing: no site, no tracking, no ad accounts, no attribution pipeline. The category (prescription ED treatments, peptides, NAD+) is one of the more heavily policed verticals in paid media, so platform approvals and compliance had to be solved before a single dollar of spend went out the door.
The approach: experiment-led, not launch-and-hope
Rather than flipping on every channel at once, we treated the first 90 days as a sequence of staged tests, each informing the next:
- Google Ads first (5/26). Search and Shopping went live first to capture existing demand and validate the funnel (landing page, questionnaire flow, GA4 and BigQuery attribution) against real traffic before scaling spend anywhere.
- Reddit second (6/1). Once the funnel was proven, we added a colder, community-driven audience tested against the same conversion benchmarks.
- Meta third (6/22). With two channels of conversion data in hand, Meta launched last on a structured CBO/ABO testing framework, so budget went straight to the audience and creative combinations the earlier data pointed to.
Every channel launch was gated by data from the one before it, not a calendar. That sequencing is the core of the methodology: each new dollar of spend is informed by the last dollar's results, tracked through a single BigQuery pipeline (GA4, ad platforms, and order data via webhook) so decisions were never made on platform-reported numbers alone.
The results
Spend scaled fast, and efficiency improved as it did.
Anonymized engagement, first 90 days. Figures from the client's BigQuery pipeline.
| Month | Spend | Visitors | Purchases |
|---|---|---|---|
| May (partial) | $969 | 1,291 | 1 |
| June | $24,037 | 8,651 | 32 |
| July (through 7/21) | $95,300 | 31,843 | 258 |
Spend grew roughly 100x from May to July. If acquisition costs had scaled with spend, this would be unremarkable. Instead, cost per purchase moved the other way:
| Week of | Weekly spend | Cost per purchase (first order) |
|---|---|---|
| 6/1 | $4,940 | $1,235 |
| 6/8 | $5,523 | $1,841 |
| 6/15 | $5,527 | $790 |
| 6/22 | $6,286 | $629 |
| 6/29 | $16,042 | $349 |
| 7/6 | $30,337 | $289 |
From the first full week of multi-channel spend to the most recent full week on record, cost per purchase dropped 77% while weekly spend scaled ~6x, the signature of a system getting more efficient under load as budget increased.
Full-funnel volume, first 90 days: 4.56M impressions, 48,098 clicks, 40,623 site visitors, 3,604 questionnaires started, 447 patients approved, 291 completed purchases.
(First-order figures captured via GA4/BigQuery attribution. They exclude recurring subscription revenue, which compounds separately over the patient lifecycle.)
Why it worked
- Sequenced channel launches instead of a big-bang rollout, so every new platform inherited a validated funnel and real conversion benchmarks.
- One attribution pipeline across every channel (GA4, Google Ads, Meta, Reddit, and order data unified in BigQuery), so spend decisions were made on actual patient outcomes, not platform-reported conversions.
- Budget followed evidence. Spend scaled fastest into the channels and campaigns already proving efficient, which is why cost per purchase fell as volume rose.
Figures from the client's BigQuery marketing performance pipeline, April 1 to July 22, 2026.