What you'll find in this article
- Why Google Ads account management requires a structured process, and what happens when you don't have one
- How to prepare before opening the interface: the right mindset for weekly Google Ads optimization
- The Google Ads optimization checklist in priority order, with the logic behind each step
- How to distinguish high-impact interventions from routine maintenance activities
- The most common process errors and how to build a Google Ads optimization workflow that doesn't break
- How to document changes and build an optimization log that becomes a real asset over time
The difference between a well-managed and a poorly managed Google Ads account rarely comes down to a single brilliant decision. It comes down to the quality and consistency of the optimization process over time. I've seen accounts with impeccable structure and generous budgets underperform for months because no one had a clear system for what to look at, in what order, and how often. I've seen accounts with less-than-perfect structures produce excellent results because the manager had built a disciplined optimization rhythm that caught problems before they became costly.
This article documents the weekly Google Ads optimization process I've developed and refined on real accounts since 2021. It's not a list of generic best practices, it's the concrete operational sequence I use every week, with the explanation of why each step is in that order and what experience has taught me to watch out for.
The process is divided into three phases: the orientation phase (before touching any setting), the intervention phase (the actual changes, in priority order), and the documentation phase (recording what was done and why). Each is necessary. Skipping the orientation phase produces interventions based on intuition rather than data. Skipping documentation means starting from scratch every week instead of building on what's already been learned.
Before you open the interface: the right mindset for optimization
The first mistake in the optimization process happens before you even open Google Ads. You open the account directly, look at what "seems" different from last week, and act on the first anomalies you notice. This reactive approach leads to over-optimizing campaigns that had a bad week for external reasons, seasonality, market events, tracking issues, while ignoring less visible but more costly structural problems.
Before opening the interface, I ask myself three questions:
1. What happened this period at the business context level? Were there seasonal events, holidays, client promotions, site or product changes that may have influenced performance? A conversion rate drop this week has completely different meanings if the site underwent a technical update versus if nothing changed. A manager who doesn't know the business context is optimizing in the dark.
2. What is the primary success metric for this account this week? Not every week has the same objective. In a scaling phase, the priority is expanding coverage while maintaining the CPA target. In an efficiency phase, it's reducing wasted spend. In a testing phase, it's collecting statistically significant data. Opening the account without a clear session objective leads to optimizing for the metric that appears most prominently in the interface, which isn't necessarily the most relevant one.
3. What did I plan to do last week that I didn't complete? The backlog of incomplete optimizations is the starting point, not an afterthought. This question forces me to reopen the previous session's notes before looking at new data, which means every session builds on the previous one instead of starting from scratch.
This five-minute orientation phase before opening the interface has significantly changed the quality of my optimization sessions. When I skipped it, I invariably ended up acting on short-term anomalies and missing structural problems. The most common pattern was: the account had a terrible Monday and Tuesday (for seasonal reasons I hadn't considered), I'd open the weekly report seeing ugly numbers, and I'd start lowering bids or pausing campaigns that were actually working well on a monthly scale.
The correct date range for weekly optimization
Before looking at any metric, setting the date range correctly isn't a technical detail, it's an analytical decision. For a standard weekly session I use three ranges in parallel: the last 7 days (to identify recent anomalies and trends), the same period from the previous year (to contextualize seasonality), and the last 30 days (for bid and budget decisions, which require statistically significant volumes). Making bid decisions by looking at only 7 days of data on campaigns with few weekly conversions is one of the most common mistakes I see. Seven days can be a statistical anomaly; thirty days is a trend.
I learned this lesson the most expensive way possible. In early 2022 I was managing an e-commerce account with Shopping campaigns that had 8-12 conversions per week. I lowered the Target ROAS bids on a campaign after a weak week, without looking at the 30-day picture. The campaign was in a learning phase from the previous change, and that "weak" week was statistical noise. The result was that the campaign came out of the learning phase in the wrong state and it took three weeks to restabilize. Since then I never make decisions on automated bidding strategies looking at less than 21 days of data, except for glaring anomalies.
The weekly Google Ads optimization checklist
The Google Ads optimization checklist I follow is ordered by expected impact and diagnostic logic: the initial steps produce information that's necessary to correctly execute the subsequent ones. Skipping the preliminary checks to get to the "real optimizations" faster is like making a medical diagnosis by skipping the patient history, you risk treating the wrong symptom.
1. Actual spend vs. allocated budget check
The first check, always, is spend. Not conversions, not CPA, not ROAS, spend. An account that spent 70% of its planned budget this week is signaling a problem that needs diagnosing before any other optimization. The most common causes of underspend: budget limitations on high-demand campaigns (meaning lost impressions and potential conversions), ad approval issues, impression share drop due to increased competition, or campaigns exiting the learning phase conservatively.
Overspend matters just as much: campaigns that spent 130% of their average daily budget on some days signal either a demand spike (an opportunity to capture by increasing budget) or a targeting problem (unqualified traffic entering at volume).
The spend check has saved me from completely misguided optimization sessions more times than I can count. The most emblematic case: an account had apparently excellent CPA that week. Before congratulating myself, I looked at spend, it had spent 45% of budget. The low CPA wasn't improved efficiency, it was few conversions on low spend because the campaigns had nearly stopped serving. The problem was an ad approval change that had gone unnoticed. If I'd optimized based on CPA without seeing the spend, I'd have raised bids on campaigns that weren't serving for completely different reasons.
2. Campaign learning status check
The second check is the learning status of campaigns using automated bidding. Campaigns in "Learning" or "Limited learning" status behave differently from stable campaigns, they produce greater performance variance and shouldn't be optimized with the same criteria. Before analyzing any performance metric, I identify which campaigns are in a learning state and why.
The learning phase triggers I monitor weekly: significant budget changes (more than 20-30% in a single step), bidding strategy changes, ad group structure modifications, addition or removal of conversions from the optimization goal. Every time a campaign enters learning due to a change I made, I log it in the notes with the date, so I know exactly when the campaign should exit the learning phase and can judge that week's performance in the correct context.
The most frequent error pattern I've seen, in accounts managed by others and in my own accounts in the early phases, is modifying a campaign in learning because "performance has dropped." Performance dropped because the campaign is in learning. Modifying it further resets the learning counter and indefinitely extends the instability period. This cycle can go on for months if you don't have a system for tracking learning status and respecting it.
3. Search query review and negative keyword update
The search query review is the weekly Google Ads optimization step with the most consistently high return over time. For accounts with active Search or Shopping campaigns, I dedicate as much time as needed to this step, not a fixed amount. Some weeks it takes half an hour, others ten minutes. The goal is to identify: irrelevant search terms generating unconverted clicks, high-quality terms not captured by exact match that should be added, and terms signaling an audience that doesn't match the client's target.
The process I follow for query review: filter terms by cost (not impression volume, I want to see where money is going, not where traffic is), identify all terms with significant spend and no conversions in the last 4 weeks, evaluate each term in the context of the keyword that triggered it, and decide whether to add as a negative, modify the original keyword's match type, or accept the term as normal statistical noise.
I learned not to build negative keyword lists mechanically. For every term I'm considering adding as a negative, I ask: has this term ever converted in the last 12 weeks? If so, what was the context? The term "cheap" in a luxury account is clearly a negative. But "reviews" in a B2C e-commerce account? It might be someone evaluating a purchase, a legitimate pre-conversion behavior. Adding "reviews" as a negative without checking conversion data can be a costly mistake. The rule I've adopted: no term gets added as a negative without being checked against the conversions column for the last 90 days.
4. Campaign and ad group performance analysis
With spend and learning context established and queries cleaned up, we move to performance analysis. The framework I use is based on deviations from the mean: I identify campaigns and ad groups that deviate significantly from the historical average on CPA, ROAS, or conversion rate, and focus on those, not on the entire account uniformly.
For each outlier (positive or negative), the question is: is this deviation explainable, transient, or structural? A campaign with CPA 40% above average this week because the site was slow for two out of seven days has a transient problem. A campaign with steadily increasing CPA for three weeks likely has a structural problem, increased competition, ad quality deterioration, or audience fatigue.
Analyzing by deviation from the mean rather than absolute value is one of the most impactful process optimizations I've made. When I looked at absolute values, I'd end up optimizing campaigns that were doing fine because the absolute numbers seemed low, and missing campaigns that were deteriorating because the absolute numbers were still "good." Comparison against the 90-day historical average reveals trends before they become serious problems.
5. Bidding strategy and target review
With the performance analysis complete, we address bidding strategy decisions. The rules that govern my Google Ads optimization methodology:
For Target CPA campaigns: I only modify the target if I have at least 21 days of post-last-modification data and actual performance consistently deviates from the target for more than two consecutive weeks. A single week of above-target CPA isn't sufficient signal. When I do modify, I do so in increments of 10-15% maximum, larger variations unnecessarily reactivate learning.
For Target ROAS campaigns: I apply the same time criteria, with additional attention to conversion volume. A Target ROAS makes sense only if the campaign has enough weekly conversions to allow the algorithm to optimize (a commonly cited reference is around 50 conversions in the prior 30 days as a practical threshold for reliable Smart Bidding, though it isn't a hard requirement). Below that threshold, I evaluate whether Target CPA is more appropriate, or whether it makes sense to expand the conversion window.
One of the most common mistakes I see in accounts managed by others is target chasing. The account had a $20 CPA target, the campaign delivers $28 CPA for two weeks, the manager lowers the target to $16 hoping the algorithm will "compensate." The algorithm instead reduces bids, loses impression share, and the campaign delivers 30% fewer conversions at $22 CPA, technically closer to target, but less efficient in absolute terms. The bid target isn't a magic lever, it's an optimization parameter that makes sense only when supported by volume and a bid strategy with sufficient data.
6. Ad and asset review
Ad review in the weekly optimization doesn't mean creating new ads every week, it means systematically monitoring existing ad performance and identifying when a test has reached statistical significance. For campaigns with Responsive Search Ads, I look at the "Ad strength" column and conversion rate per ad, not just CTR. An ad with high CTR and low conversion rate is attracting the wrong clicks, a relevance mismatch between message and landing page that CTR alone doesn't reveal.
For structured ad tests (variant A vs. variant B), I check every week whether the test has reached statistical significance using Google's built-in test tool or an external significance calculator. Concluding a test before statistical significance, even when "it seems clear who's winning", is one of the most common data waste patterns in ad optimization.
7. Extensions, assets, and Google's recommendations check
The last step of the weekly session is checking ad extensions (or assets, in current terminology), expiring promotions, and a quick review of Google's automated recommendations. On this last point: Google Ads recommendations should be treated as suggestions to evaluate in the context of the specific account, not as instructions to execute. Some are genuinely useful (adding missing assets, fixing keyword conflicts). Many are optimized to increase the account's spend, not to improve performance relative to the client's specific objective. Never accept recommendations in batch without individual review, the "Apply all" option is designed for the inattentive manager's convenience, not for account performance.
How to prioritize interventions: impact vs. urgency
Not every week has time to complete each checklist step with equal depth. When time is limited, intervention prioritization follows a precise logic: first address problems that cost money every day they go unresolved, then make improvements that increase performance, and finally handle maintenance and monitoring activities.
The prioritization framework I use:
The distinction between High and Medium priority is often underestimated. I've seen managers spend hours every week optimizing bids with millimeter precision while critical campaigns were suspended for an unnoticed approval issue. The rule I've adopted: the first 15 minutes of every session are dedicated exclusively to High Priority checks, campaign status, conversion anomalies, overspend. Only after those verifications do we proceed with standard optimizations.
The over-optimization problem
An aspect of the weekly optimization process that's rarely discussed is the risk of over-optimizing. Google Ads campaigns with smart bidding need stability to perform well. Every significant change, to bids, budgets, targeting, structure, introduces variance into the system. A manager who makes ten significant changes per week to an average account creates a constantly unstable system that never produces clean data to optimize against.
The metric I monitor to evaluate my tendency to over-optimize: how many campaigns have been in a learning state in the last 30 days? If the answer is "most of them," I'm modifying too frequently. The goal of the weekly optimization process isn't to make the maximum possible number of changes, it's to make the right changes, at the right time, with the expected impact documented.
What I've learned the hard way: lessons from years of optimization on real accounts
Documenting changes is process optimization, not bureaucracy
The change annotation system is the part of the weekly optimization process most often skipped because it "takes time." It's also the one that produces the greatest return over the long term. When I change a bid target, add significant negative keywords, or modify a campaign structure, I log: the date, what was changed, why it was changed (referencing the data that motivated the decision), and what the expected result is over the next 2-4 weeks.
This log has three concrete utilities. First: it lets me distinguish between performance that changes due to my modifications and performance that changes due to external factors. Second: after six months, it becomes a database of my decisions that reveals patterns, which types of interventions have worked consistently on this account and which haven't. Third: if the client or another manager needs to understand the account's state, the change log is the most useful document in existence, more useful than the account structure itself.
The real value of the change log became clear to me when I had to explain to a client why February performance was better than November's. Without the log, I could only say "we optimized." With the log, I could identify exactly the four changes that produced the improvement and when they became effective. That conversation transformed the client from someone who evaluated monthly results into someone who understood the optimization process, and trusted it was working even in weeks when the numbers weren't perfect.
The right optimization frequency depends on volume, not the calendar
A weekly session is the right cadence for most accounts with monthly spend above $3,000. But "weekly" doesn't mean "every seven days regardless of data volume." The correct optimization frequency depends on how quickly the account accumulates statistically significant data. An account spending $500/day accumulates data much faster than one spending $50/day, and can support more frequent optimizations without running into the insufficient data problem.
For high-budget accounts (above $10,000/month), I run bi-weekly sessions with a quicker weekly monitoring session. For accounts with budgets below $2,000/month, I can limit to bi-weekly sessions without losing significance, provided automated alerting is active for critical anomalies.
The "do something" bias in every session
One of the hardest cognitive biases to fight in Google Ads account management is the implicit pressure to do something in every optimization session. If you open the account every week and close it without making changes, you feel like you haven't worked. This feeling leads to unnecessary changes, marginal bid adjustments, premature ad tests, targeting modifications based on insufficient data, that introduce variance without value.
The realization that changed my approach: some of the most effective optimization weeks I've had were those where I opened the account, verified everything was working as expected, and closed it without touching anything. Letting campaigns accumulate data without interference is often the most positive intervention you can make. Not every session needs to produce changes, every session needs to produce an informed assessment of whether changes are necessary.
Building calibrated intuition for an account
After months of structured weekly optimization on the same account, you develop something I'd call calibrated intuition: the ability to open the account and immediately notice when something is off, even before looking at the data in detail. This intuition isn't magic, it's the result of having seen the same account in enough different conditions to recognize when it deviates from its normal behavior.
Calibrated intuition is one of the greatest advantages of continuity in account management versus frequent manager turnover. A new manager taking over an account has to build this calibration from scratch, which typically requires 4-8 weeks of data before they can optimize with confidence. During that period, the frequency and depth of optimizations should be more conservative, precisely because the historical context needed to distinguish normal from abnormal is missing.
When the weekly optimization process isn't enough
The weekly optimization process is necessary but not sufficient for long-term Google Ads account management. There are two types of work that can't be compressed into a weekly session and need to be planned separately: structural optimization (reorganizing campaign structure, reviewing keyword match type strategy, redesigning audience segmentation) and strategic optimization (reviewing budget allocation across campaigns and channels, evaluating new campaign types, planning quarterly tests).
The risk of doing only weekly tactical optimization without ever doing structural optimization is progressively improving a system that has reached the limits of its current architecture. At some point, there aren't enough bid adjustments or negative keywords that can compensate for an account structure fundamentally unsuited to the business's current growth phase.