Performance Max Campaigns

Performance Max is a single goal-based campaign that serves ads across all of Google’s surfaces — Search, Shopping, YouTube, Display, Gmail, Discover and Maps — using machine learning to decide placements, bids and audiences automatically. Where a search campaign gives you keyword-level control and a Shopping campaign gives product-group control, PMax asks for assets, signals and a target, then handles the rest itself. That trade — control surrendered for reach and automation — is exactly why the quality of its inputs and guardrails decides everything.

It is opaque by design, but not beyond supervision — Google has progressively opened channel distribution, search category, and asset-level reporting, and disciplined management extracts real visibility from them. More importantly, you control the campaign through its inputs: conversion values, assets, audience signals, brand exclusions and negatives. Our job is applying scrutiny the campaign type does not volunteer and adjusting those levers when evidence demands it. Blind trust and blanket refusal are both mistakes; supervised adoption is the profitable middle.

Left unconfigured, yes — brand traffic is the classic way PMax flatters itself, absorbing customers who typed your name and would have converted through a cheap brand search ad or a free organic click. We apply brand exclusions as standard, keeping branded queries in your dedicated brand campaign where their economics stay visible, so PMax must prove itself on genuinely incremental demand. Any PMax report that has not addressed brand traffic should be read with deep suspicion.

More is better, and thin data is the most common reason PMax disappoints — the algorithm learns from conversion patterns, so an account producing a healthy monthly volume of accurately tracked conversions gives it real material, while a handful of monthly leads leaves it guessing expensively. As a working rule we want stable tracking and enough monthly conversions for patterns to emerge before recommending it; below that threshold, search and Shopping build the data foundation first. We assess your volumes during the audit and give you a straight verdict.

Asset groups are PMax’s creative containers — each holds the headlines, descriptions, images, logos and video the system assembles into ads for a theme, plus its own audience signals and final URL. We organise them around coherent product lines or service themes so messaging stays relevant to what each promotes, much as ad groups theme a search campaign. Complete, high-quality assets in every group matter enormously: missing formats shrink your eligible inventory, and weak assets get assembled into weak ads at machine scale.

Audience signals are starting suggestions, not restrictions: you supply customer lists, converter profiles and high-intent segments, and the algorithm begins its exploration there before expanding to whoever it predicts will convert. Good signals shorten the expensive wandering phase of learning; bad or lazy signals leave the machine to find your market by trial and error, billed to you. We build signals from your strongest first-party data, which is one of several reasons the customer-data groundwork matters before launch.

Because the campaign is optimising toward the conversion you told it to count — form submissions — and it has found the cheapest population willing to submit forms, which is not the same as buyers. The fix is redefining success at the signal level: qualified-lead conversions fed back from your CRM, value differentiation between enquiry types, and offline conversion imports that teach the algorithm what a sale-producing lead looks like. PMax does not know your leads are rubbish until your data tells it; we build the loop that does.

For most Indian ecommerce accounts we run both deliberately: standard Shopping holds the products where margin control and query sculpting matter, while PMax hunts incremental volume across surfaces standard Shopping cannot reach. Feed-only PMax offers a middle path for testing the reach without full asset exposure. The right split depends on catalogue economics, data volume and how much averaging your margins can tolerate — which is a scoping question we answer with your numbers, not a doctrine we apply to everyone.

Plan for several weeks of genuine learning whenever a campaign launches or its targets change materially — during which spend distribution shifts around, early results run noisy, and the temptation to intervene daily must be resisted. Constant target changes reset the learning and burn budget re-mapping the same territory. Our supervision during this phase is watchful but restrained: we intervene on clear drift like brand leakage or irrelevant categories, and otherwise let the system converge before judging it.

Google’s marketing sometimes implies so; our account data says no. Brand campaigns preserve the visibility and economics of your own name, tightly managed search keeps control where intent is proven and margins are sensitive, and remarketing runs sharper as a dedicated layer. PMax earns its place as the reach and automation engine within that portfolio — usually a substantial place — but consolidating everything into one opaque campaign surrenders exactly the strategic control that keeps automation honest.