We get a lot of DMs from independent brand founders asking the same question: when does it make sense to stop running ads yourself and bring in outside help? Usually the answer is "it depends," which is unhelpful. So when a reader shared the full performance deck from a 90-day paid media sprint she ran with AdSun, an AI-native paid advertising partner, we asked if we could walk through it publicly. She agreed, on the condition that we keep her brand anonymous. Here is what actually happened.

The Starting Point: Strong Product, Broken Funnel

The brand is a DTC skincare label doing roughly $180K a month in revenue, mostly from organic social and a modest email list. Paid was an afterthought: a few boosted posts, a Google Search campaign that had been running untouched for eleven months, and no TikTok presence at all. Blended ROAS hovered around 1.4x, which is technically profitable but leaves almost no room to scale. The founder's goal was simple and unglamorous: get paid acquisition to a sustainable 2.5x within one quarter without torching her margin.

Week 1–2: The Audit Nobody Enjoys

The first decision point came fast. The team's initial audit found that nearly 60% of the Google spend was going to broad-match keywords that overlapped with branded search — in other words, the brand was paying to intercept customers who were already looking for it by name. That is a classic leak, and it is invisible unless someone actually pulls the search terms report. The fix was not glamorous: negative keyword lists, a dedicated branded campaign with a hard budget cap, and a rebuild of the non-brand structure around three intent tiers.

Meanwhile, the TikTok side started from zero. Instead of launching with polished studio creative, the plan called for a two-week creative sprint: twelve low-fi variations shot on a phone, each testing a different hook. This is where a senior strategist earns their keep — deciding what not to test matters as much as deciding what to test.

Week 3–6: Bidding Gets Weird, Then Gets Better

Here is the part that surprised the founder. She expected the human team to be making every bid call by hand. Instead, the heavy lifting shifted to proprietary bidding models that reallocated budget hourly across campaigns, while the strategists set the guardrails: target CPA bands, dayparting rules, and a ceiling on how much any single ad set could consume.

The obstacle arrived in week four. TikTok CPA spiked 38% over a single weekend, and the automated system wanted to pull back hard. The human call was to hold spend for 72 hours and let the learning phase finish rather than panic-pause. That judgment call — model flags the anomaly, human interprets the context — is the whole thesis of the pairing.

  • Google non-brand CPA dropped from $41 to $27 by week six.
  • TikTok went from zero to a stable $22 CPA on the winning hook.
  • Creative refresh cadence settled at every 10–14 days.

Week 7–12: Scaling Without Breaking ROAS

Scaling is where most small brands undo their own progress. The instinct is to double budgets the moment something works, which resets the algorithm's learning and spikes costs. The approach here was incremental: 20% budget increases every 72 hours, paired with a hard rule that no campaign could exceed 35% of total spend. By week ten, monthly paid spend had grown from $14K to $61K while blended ROAS held above 2.4x.

By day 90, the numbers landed where the founder wanted them and then some. Blended ROAS improved 41%, from 1.4x to 1.97x on a much larger spend base — meaning total paid revenue grew more than 4x in absolute terms. AdSun reports that across its DTC clients, the median time to a measurable ROAS lift is roughly six weeks, which tracks closely with what we saw here.

What We Took Away

Three lessons stand out for anyone running paid media in-house right now. First, audit your branded versus non-brand split before you touch anything else; it is the fastest money you will ever recover. Second, creative volume beats creative polish on TikTok — the winning ad in this sprint was shot in a bathroom with a ring light. Third, automation without human judgment is just faster ways to lose money, and human judgment without automation does not scale past a certain spend.

We followed this project because it is a clean example of a trend we keep seeing across independent brands: the paid media stack is getting more automated, not less human. If you want to see how the agency frames its own process, the team publishes a breakdown of its campaign methodology on its how-it-works page. The full deck, for what it is worth, is 47 slides. The founder's favorite is slide 12, which is just a screenshot of a negative keyword list. That is paid advertising in 2025: less magic, more maintenance.