Insights · Growth · Sep 11, 2026 · 7 min read
Performance Max, Advantage+ and automation: tips for keeping control of your ad spend
Google and Meta now push nearly every advertiser toward automated campaign types, and they often work — until they quietly spend your budget on the wrong things. Seven guardrails we put on every automated campaign so the machine optimises toward your goal, not its own.
Illustrative product interface — not client data.
The platforms have made their preference clear: Performance Max on Google, Advantage+ on Meta, and automated bidding everywhere. Used well, automation finds conversions a manual campaign would miss. Used carelessly, it optimises toward whatever is easiest to count — your own brand searches, low-value form fills, existing customers — and reports it as success. These are the guardrails we apply to every automated campaign we run for clients. They do not fight the automation; they point it at the right target.
Key takeaways
- Automation optimises toward the conversion you define. Define it precisely, with values, or it will optimise toward the wrong one.
- Separate brand from non-brand so automated campaigns cannot claim demand you already had.
- Feed the machine good creative and clean first-party data; it cannot invent either.
- Set exclusions, pacing and reporting before launch, and review weekly for the first month.
1. Define the conversion the way finance would
An automated campaign will find more of whatever you count. If you count every form submission, it will find cheap, low-quality submissions. Count qualified leads (via offline conversion import or a CRM stage), purchases with value, or booked appointments. Where you cannot measure quality directly, weight conversions by value so the algorithm learns the difference between a newsletter signup and a quote request. Groundwork in honest measurement.
2. Keep brand out of automated campaigns
Performance Max will happily serve on your brand name and take credit for people who were already coming. Add brand exclusions and run brand search in its own tightly controlled campaign. On Meta, exclude existing customers from prospecting unless retention is explicitly the goal.
3. Give it creative worth optimising
Automation combines and tests what you give it. Supply a real library: multiple headlines and descriptions with distinct angles, images and video in every required ratio, and refresh it monthly. Weak inputs produce a confidently optimised weak campaign. Our creative and media team builds these libraries as systems, not one-offs.
| Guardrail | Google (Performance Max) | Meta (Advantage+) |
|---|---|---|
| Conversion definition | Import qualified leads or values; primary vs secondary actions | Value optimisation; Conversions API for reliable events |
| Brand separation | Brand exclusions; separate brand search campaign | Exclude existing customers from prospecting |
| Placement control | Placement exclusions; review video and display placements | Review Audience Network and placement reports |
| Creative inputs | Full asset groups, refreshed monthly | Multiple formats and hooks per ad set |
| Budget pacing | Daily caps plus a weekly pacing check | Campaign budget with spend limits |
| Learning period | Avoid large changes for two to three weeks | Avoid resetting the ad set repeatedly |
4. Set exclusions before launch
Placements, content categories, geographies you do not serve, and audiences that never convert. Automation explores; exclusions define the edges of the map. Review the placement and search-term reports weekly for the first month, then fortnightly.
5. Fix tracking before touching budgets
Server-side tagging on Google and the Conversions API on Meta make conversion signals more complete under consent and browser restrictions. Without them, the algorithm is learning from a partial picture and your reports are guessing. This is unglamorous and it is the single highest-return task on most accounts.
6. Pace budgets and respect the learning period
Automated campaigns need a stable few weeks to learn. Constant budget swings and structural changes reset the clock. Set daily caps, check weekly pacing against the monthly number, and make changes in batches rather than daily.
7. Report on what matters, not what the platform shows
Platform dashboards report platform-attributed conversions, which overlap and flatter. Reconcile against your CRM or order system monthly, report cost per qualified lead or return on ad spend by your definition, and keep a note of what changed each week so results can be explained. Our paid search and paid social teams report this way by default.
Automation is a fast driver. Your job is the destination, the guardrails and the fuel.
Applied together, these guardrails let automation do what it is good at — finding conversions across inventory a human could not manage — while keeping the goal, the budget and the story yours. If an automated campaign is spending well and nobody can explain where, that is the sign to bring them in.
FAQ
Questions on this topic.
For most advertisers with reliable conversion tracking, yes, alongside a separately controlled brand search campaign. Without qualified-lead or value tracking, it will optimise toward cheap actions; fix measurement first.
Usually one of three things: it is claiming brand demand you already had, it is counting low-quality conversions, or the platform’s attribution overlaps with other channels. Separate brand, define conversions by quality, and reconcile against your CRM.
Rarely and in batches. Give it two to three stable weeks to learn, then review weekly and change monthly unless something is clearly wrong. Constant tweaks reset the learning and hide what worked.
Yes. The first two weeks are audit: tracking, conversion definitions, exclusions, brand separation and creative. Most accounts improve from those fixes before any bid or budget changes.
Keep reading
More from the studio.
Content tips for the AI-answer era: pages that get cited and still convert
Writing for AI answers and writing for people who buy are not in conflict, but they are not the same job either. Twelve tips for pages…
How to evaluate an AI vendor before you buy: a twelve-question checklist
Every software vendor now says “AI-powered.” Twelve questions separate a product you can rely on from a demo with a subscription attached: what the model does,…
Security tips for small teams in 2026: passkeys, MFA, backups and AI-era phishing
Attackers now write flawless emails, clone voices and automate reconnaissance, which means the old advice to “look for spelling mistakes” is dead. The good news is…
Let’s connect
Want this applied to your business?
Tell us what you’re building. We’ll reply within one business day with next steps and a tailored quote.