Client: North American DTC women's apparel brand, $80K/mo ad spend, working with 1 agency + 1 in-house PPC marketer before adopting DEXUN AdWhiz. ROAS was stuck at 2.1x for 6 months. Target was 3.5x+.
Starting point: diagnosing waste
First-week AI audit across three platforms surfaced three main problems:
- Google Ads: 35% of budget on "low-intent informational" search terms (e.g. "how to dress for spring") — high clicks but 0.4% CVR
- Meta Ads: "Lookalike 1%" was 60% of Meta budget but only 1.6x ROAS, mainly because the seed audience was too old (based on purchasers from 2 years ago — no longer behaviorally relevant)
- TikTok Ads: nearly all budget on 2 million-follower creators, but their real engagement rate was under 3% and the latest Spark Ad was visibly fatigued
Month 1: Stop the bleeding
Of 12 AI recommendations, the client applied the 8 highest-confidence ones first:
- Google: paused 47 keywords with CPA > $40, added 23 negative keywords
- Meta: cut Lookalike 1% budget by 60%, moved it to Lookalike 0.5% (tighter but more precise)
- TikTok: paused the 2 million-follower creators, launched a matrix of 8 mid-tier creators (10k-100k followers)
End of month: cross-platform ROAS 2.1x → 2.9x (+38%). Same total spend, just the waste cut out.
Month 2: Reallocate
After stopping the bleeding, budget tilted toward higher-efficiency channels. AI kept proposing:
- Google: +25% budget to Brand campaigns (ROAS 12x, the most stable profit source)
- Meta: paused 3 chronically losing ad sets, redistributed budget evenly across TikTok Spark matrix
- TikTok: started using Custom Audiences for retargeting — CPM dropped from $11 to $6
End of month: ROAS 3.7x (+27%). This was when TikTok crossed from "unprofitable" to "primary growth driver".
Month 3: Fine-tune
With foundational fixes done, AI moved into "micro-optimization":
- Google: Performance Max to catch search-term long tail, asset groups split by category
- Meta: Advantage+ Shopping test — only 15% of budget but 4.2x ROAS
- TikTok: 3 of the 8 creators stood out; budget shifted from even distribution to "top-weighted"
End of month: ROAS 4.6x (+24%). 1.1 points above the 3.5x target.
Quantified return
$80K/month ad spend, ROAS 2.1x → 4.6x:
- Old monthly revenue: $80K × 2.1 = $168K
- New monthly revenue: $80K × 4.6 = $368K
- Monthly uplift: $200K (same ad spend)
- Annual uplift: $2.4M
- Business subscription: $279/mo × 12 = $3,348/year
- ROI: ~873x
Key insight
The client's in-house PPC marketer wasn't replaced — efficiency went up. They went from "5 hours/day pulling reports + manually adjusting bids" to "1 hour/day reviewing AI recommendations + setting brand direction". AI replaces mechanical execution, not strategic thinking.