Performance Max for service businesses: the Da Life playbook

Service businesses get told Performance Max isn’t for them. "It’s a product-feed thing," "you need ecommerce data," "the algorithm needs hundreds of conversions." All three are wrong, or at least incomplete. I ran PMax for a Kaua’i tour operator for 9.5 months and pulled 750% ROAS — $13,200 in spend returning $98,800 in tour bookings. This is the playbook.

The setup, in plain English

Da Life Outdoors, an adventure tour operator on Kaua’i. Two flagship products — a waterfall rappel ($399) and a hike-and-swim ($371). Both seasonal, weather-dependent, booked through a third-party platform. No physical product, no shipping, no inventory feed. Google’s PMax docs assume a Merchant Center feed. We didn’t have one — we had two services, two booking pages, and a calendar.

The structure that worked

I split into two separate PMax campaigns — one per service. Not one campaign with two asset groups. Two campaigns. This is the most important decision in the build. PMax optimizes budget across asset groups by its own logic; if your services have different margins or seasonality, one starves the other. Splitting lets you control budget and bid strategy per service.

The breakdown above is the first 6.5 months. Over the full window (Oct 2025 – Jul 2026) the account now runs 750% ROAS, $13.2K → $98.8K, 18.1K clicks.

The five decisions that drove the numbers

1. Conversion event = booking, not lead form. Most set the conversion to "form submission" or a calendar page view — soft signals PMax over-optimizes toward. The actual booking-confirmation page (order ID in the URL) became the event. Smaller volume, much cleaner signal.

2. Conversion value = average booking value, dynamic. Hard-coding $399 lies to the algorithm when you have group bookings, add-ons, or upsells. We sent the real transaction value via the booking platform’s webhook into Google Ads. ROAS tracking became real instead of fictional.

3. Audience signals were aggressive. Custom segments from "things to do Kaua’i," "waterfall hike Kaua’i," competitor names, and tour-operator categories, plus a customer-match list of past bookers. Without this, PMax defaults to broad demographic guesses.

4. Asset-group creative was specific to the service. Generic "Kaua’i adventure" pulled bad CTR. Product-specific copy ("Rappel a 60-foot Kaua’i waterfall") cut CPL from $89 to $54 in the first 14 days. The algorithm rewards specificity because the audience does.

5. Final URL expansion = OFF for the first 60 days. Left on, it scatters traffic to your blog, About, and contact pages — none of which convert. We turned it off until we had 100+ conversions, then re-enabled cautiously.

What didn’t work

Single campaign with multiple asset groups. The waterfall asset group consumed 78% of spend in week one because its conversion rate was higher; the hike group never got enough budget to find its audience. Splitting solved it.

Demand Gen as a complement. Tested alongside PMax to seed the upper funnel. Spent $1,200, got 4 bookings. Killed it. For low-frequency, high-consideration purchases like a $400 tour, search and PMax do the work; visual-discovery channels don’t.

What it means for service businesses

• One campaign per service. Not one campaign with multiple asset groups.
• Conversion = transaction, not form fill or page view.
• Conversion value = dynamic, not fixed.
• Audience signals = aggressive and specific.
• Final URL expansion off until you’ve got conversion volume.
• Specificity in creative beats polish. "Rappel a 60-foot waterfall" beats "Adventure of a lifetime."

The full Da Life case study lives here. If you want me on your account, The Roadmap is $497 and we’ll find the leak in 30 minutes.