What the first 90 days of a Google Ads account should actually look like

Most new Google Ads accounts are judged far too early and restructured far too often. Ninety days is roughly the minimum honest window to know whether a search programme can work for a business — and only if the first three weeks are spent on measurement rather than on volume.

Before anything is launched

Two numbers decide the entire structure, and neither of them lives in an ad platform.

The first is what a customer is worth. For e-commerce that means contribution margin after returns, not average order value. For B2B it means the value of a closed deal multiplied by the rate at which qualified leads close. The second is what you can afford to pay to acquire one. If those two numbers are not agreed in writing with whoever signs off the budget, every later conversation about performance becomes a matter of opinion.

We also insist on three technical prerequisites before spending a dollar: conversion tracking that fires once and only once, a consent implementation so tags respect the visitor's choice, and a landing page that loads in under 2.5 seconds on a mid-range Android phone on a mobile connection. Google's own guidance on destination requirements is blunt about landing pages that are slow, broken or misleading — those problems get ads disapproved, not merely underperforming (Google Ads destination requirements).

Weeks 1–3: prove the data

The first three weeks exist to generate clean, trustworthy data — not results. Structure is deliberately narrow:

• Branded search in its own campaign with a capped budget, so it never flatters the rest of the account.
• Two to four exact and phrase match campaigns on the queries you already know convert, split by intent rather than by product hierarchy.
• Manual CPC or maximise clicks with a bid cap. Smart bidding needs conversion history; giving it none and then blaming the algorithm is a popular waste of a month.
• A negative keyword list built daily, not weekly. Early accounts attract job-seekers, students, competitors and free-tool hunters, and every one of them is spending your budget.

Success in this phase is measured by data quality: is the conversion count in the platform within 10% of the count in your CRM or order system? Are search terms recognisably commercial? If the answer to either is no, do not proceed to phase two.

Weeks 4–8: widen carefully

With four to six weeks of clean conversions, automated bidding becomes viable. Google needs a meaningful conversion volume for a target-based strategy to behave sensibly — as a rule of thumb, around 30 conversions in 30 days per campaign, though we have seen target CPA behave acceptably at half that when the conversion is well defined.

This is also the phase for the campaign types that need inventory or creative: Shopping and Performance Max for retail, Demand Gen where there is video worth showing. Two rules we apply without exception:

• Never launch Performance Max over the whole catalogue at once. Split by margin band using custom labels in the feed. Otherwise the algorithm optimises toward your cheapest, least profitable products, because those convert most easily.
• Add brand exclusions. Without them, Performance Max quietly absorbs branded traffic you were already getting for free and reports it as new performance.

Run one structured test per fortnight — audience, bid strategy, landing page or offer. One at a time, with a stated hypothesis, a primary metric and an end date. Two simultaneous changes produce a result nobody can attribute.

Weeks 9–13: scale what holds

Scaling means moving budget toward the segments that maintain their cost per acquisition as spend rises, and away from those that do not. It does not mean raising every budget by 20%.

We increase a campaign's budget by no more than 20–25% at a time and then wait a full conversion cycle before judging the change. In a category with a two-week consideration window, a budget increase on Monday tells you almost nothing by Friday. The single most common self-inflicted wound we see is a budget doubled on the basis of four good days.

By the end of week thirteen you should be able to state, per campaign, the cost per acquisition, the incremental cost of the last 20% of spend, and whether the channel is above or below the affordable acquisition cost agreed in week zero.

Three reasons accounts stall in month two

1. The conversion is too shallow. Optimising to a form fill when a third of submissions are unqualified teaches the algorithm to find more unqualified people. Push a later, harder stage back into the platform — an offline conversion import from your CRM is usually a two-day job and the highest-return change available in B2B.
2. The landing page never changed. Six weeks of bid optimisation on a page written in 2021 has a ceiling. If the page does not answer the two questions your sales team hears most, ad optimisation is polishing the wrong surface.
3. Structure is being rebuilt every fortnight. Each restructure resets learning. A campaign that has been rebuilt three times in eight weeks has never had a chance to work, and no report will ever explain why.

What a realistic 90-day outcome looks like

For a new account in a competitive Singapore category with S$8,000–15,000 a month in media, a fair expectation after 90 days is a stable cost per acquisition within reach of the target, a documented understanding of which segments are profitable, and one or two campaigns clearly ready for more budget. Not a transformed business.

What we will not promise, and no agency honestly can, is a specific ROAS or lead volume by a specific date. Results depend on your offer, margin, competition, seasonality and how quickly your team follows up. Anyone quoting you a guaranteed number before seeing your account is selling something other than media management.

---

Sources and further reading: Google Ads destination requirements policy, Google Ads policies overview.

This article reflects our own account experience as at June 2026. Platform features and requirements change frequently; check Google's current documentation before acting on any specific detail.