Your TOF campaign shows a healthy 3.2 ROAS. But if only 25% of those orders are from new customers, your actual New Customer ROAS is only 0.8. You're losing money on the one job a prospecting campaign is supposed to do.
Performance Marketing
THE FORMULA
New Customer ROAS = Revenue from first-time buyers ÷ Total Ad spend
If 25% of your prospecting orders are new customers and blended ROAS is 3.2, your NC ROAS is approximately 0.8 and is unprofitable on its stated job.
Why This Happens?
Three Mechanisms of Audience Leakage
1. Missing customer exclusions
If you haven't uploaded your full customer list as a suppression audience, your prospecting campaign is freely serving ads to people who've already bought from you. Those people convert easily, the algorithm detects the signal, and it doubles down on them. Over time, the campaign self-selects toward your existing base - even if you started with a genuinely cold audience.
A critical detail many brands miss: pixel-based purchaser lists are incomplete. They miss app orders, offline orders, draft orders, POS orders, and anyone who deleted their cookies. Only a fully integrated server-side setup gives you a complete purchase list. Use it.
2. Lookalike audience bleed
A lookalike seeded from all purchasers is, by construction, a list of people who statistically resemble your buyers. The tighter the lookalike (1–2%), the more it skews toward people already familiar with your brand, because brand familiarity predicts purchase, and the model picks this up as a feature.
Seed your lookalikes from your top-LTV customers but cast a wider net. The goal is to bring in a new audience, not drawing water from the same well
3. New Customers vs New Visitors
Even if you exclude your customer list, you may not be excluding recent site visitors. A prospecting campaign reaching 180-day website visitors who haven't purchased yet will convert like a retargeting audience because that's exactly what they are.
They're warm, they've seen your brand, they're new customers, but they are not new prospects. They are already being targeted by WhatsApp and BOF campaigns.
But they don't appear on your exclusion list, so they slip through undetected.
4. Choice of Attribution
How you measure the performance of your TOF campaigns can easily compound the problem. If you are measuring Last-Touch you are heavily skewed towards shorter conversion cycles, where Retargeting works best. If you measure TOF conversions from the same lens, eventually your TOF targeting will start looking like Retargeting.
A TOF campaign may be great at discovery and terrible at conversions. When you use last touch attribution, you keep selecting the campaigns that are great at conversions.
To get the right measure of a TOF campaign, you need to measure a clean attribution setup which tracks identity stitched First Touch.
THE ANALOGY
Last Touch is a terrible way to measure TOF
If you keep selecting your football team based on the number of goals they scored, you will not have any defenders or creators. Only forwards standing at one end of the pitch waiting for a ball that never arrives.
