Google Ads AI Max Pilot — Testing on Underperforming Accounts

Overview

Asymmetric is piloting Google's AI Max optimization tool on a controlled subset of client accounts before rolling it out broadly. The strategy is to start with low-revenue or underperforming accounts — where the risk of disruption is low — validate results, and then expand to high-value accounts.

This approach was agreed upon during the ^[1], where ^[2] was identified as the first candidate.

Rationale

AI Max is a Google Ads feature that uses machine learning to expand targeting, optimize bidding, and adjust creative elements automatically. Because it changes campaign behavior in ways that can be difficult to predict, deploying it on high-revenue accounts (e.g., Doodla, Firestone) carries meaningful risk.

The pilot-first approach mitigates this by:

Limiting downside exposure — underperforming accounts have less to lose if results are mixed
Generating real data before committing high-budget accounts to the tool
Building team familiarity with AI Max behavior and reporting quirks

"I don't want to implement first the AI Max on those, for example, Doodla or Firestone. I would like to test first this AI Max on these companies that have low revenue per month."
— Gilbert, 2025-11-12

Pilot Account: Adavacare

^[2] was selected as the first AI Max test account because:

Current Google Ads results are underperforming relative to spend
Monthly revenue from this account is low, limiting risk
It represents an opportunity to demonstrate improvement with a fresh approach

Owner: Gilbert
Status: Pilot to be initiated following the 2025-11-12 call

Candidate Account Criteria

When selecting accounts for AI Max testing, prioritize accounts that meet one or more of the following:

Criterion	Notes
Low monthly ad revenue	Limits financial risk during testing
Poor conversion performance	AI Max may surface new converting traffic
Stalled or declining results	Status quo isn't working; experimentation is warranted
Not a flagship/anchor client	Avoids reputational risk with top accounts

Rollout Sequence

Phase 1 — Pilot: Deploy AI Max on Adavacare. Monitor conversion volume, cost-per-conversion, and overall account performance over 4–6 weeks.
Phase 2 — Evaluate: Compare pre/post metrics. Assess whether AI Max improved results or introduced noise.
Phase 3 — Expand (conditional): If results are positive, identify the next tier of accounts for rollout. High-revenue accounts (Doodla, Firestone) remain excluded until confidence is established.

AI Max optimization is only as good as the conversion signals it receives. A known issue across several accounts is discrepancies between Google Ads and Google Analytics conversion counts (e.g., 6 conversions reported in Ads vs. 10 in Analytics over 30 days).

Before or alongside the AI Max pilot, Gilbert is pressing Anup to resolve these tracking gaps. Accurate conversion data is a prerequisite for AI Max to optimize effectively.

See: ^[3]

Action Items

[ ] Gilbert — Launch AI Max pilot on Adavacare Google Ads account
[ ] Gilbert — Monitor Adavacare performance bi-weekly alongside standard optimization cadence
[ ] Gilbert / Anup — Resolve conversion tracking discrepancies before or concurrent with AI Max activation
[ ] Gilbert — Report results back to Mark before expanding to additional accounts

^[4]
^[3]
^[5]