Google Ads optimization: the order of work

Most Google Ads accounts get optimized in the wrong sequence. Someone rewrites ad copy every week while conversion tracking is broken. Someone else tests bidding strategies while the search terms report is full of queries containing “free,” “jobs,” and “salary.” The result is activity without progress: spend goes up, the dashboard looks busy, and cost per lead stays flat.

Every decision in an account is graded on one number. Cost per lead for lead generation, cost per acquisition for ecommerce. Clicks and click-through rate are diagnostics, not goals. If you can’t measure the real number accurately, every change you make is a guess.

Here is the sequence that moves it:

1. Measurement and conversion tracking
2. Business targets and CPA thresholds
3. Account and campaign structure
4. Ad group design around search intent
5. Keywords and match types
6. Negatives and search terms cleanup
7. Bidding strategy
8. Ads and assets
9. Landing pages and the post-click experience
10. Audiences and targeting
11. Campaign types, including Performance Max
12. Operating cadence

The order matters more than any individual tactic, because each step depends on the ones above it being right. What follows is a short overview of each area, with links to the deeper guides where they exist.

Measurement and conversion tracking come first

You can’t optimize a number you can’t measure. This is always step one, and skipping it is why most accounts plateau.

Configure primary conversion actions under Goals > Summary. A primary action feeds Smart Bidding and appears in the Conversions column. A secondary action appears in All conversions but doesn’t influence bids. Mark exactly one primary action per business outcome: the lead form, the qualified call, or the purchase.

Link Google Analytics 4 and import conversions, or fire tags through Google Tag Manager. Note that GA4 conversions imported into Google Ads arrive as secondary; you have to promote them yourself if you want them steering bids.

The pitfalls worth knowing:

• Double counting. A tag on the form submission and another on the thank-you page, both firing for one lead.
• Micro-conversions marked primary. Page views and scroll events dilute the signal Smart Bidding runs on.
• Reading “All conversions” instead of “Conversions” when judging cost per result, which quietly inflates your apparent volume with secondary actions.

Before scaling spend, trigger each conversion manually and confirm it fires in Tag Assistant. If browser privacy restrictions are costing you data, enhanced conversions and server-side tagging recover some of it.

This is the area where a wrong setting does the most damage, because everything downstream inherits it. The full treatment is in Google Ads conversion tracking.

Define what a lead is allowed to cost

Ad metrics mean nothing until they connect to your P&L.

A worked example, not a benchmark: average order value $500, gross margin 40% (so $200 of margin), and you’re willing to spend 20% of that margin on acquisition. Target CPA is $40. Change any input and the answer changes, which is exactly why published averages can’t set this for you.

Those targets drive your bidding strategy and your budget caps, so they need to come from your own numbers rather than Google’s suggestions.

• Document target CPL, CPA, and ROAS per campaign so later decisions on bids and keywords stay consistent instead of drifting.
• Treat branded and non-branded separately. Brand converts cheaply and will flatter any average it’s mixed into. Non-brand deserves its own, higher threshold.
• Allow higher CPAs for cold, non-brand traffic where downstream revenue justifies it, and make that allowance explicit rather than accidental.

Working backward from margin is covered in what is a good cost per lead.

Get structure under control

Structure is second because messy campaigns make it impossible to see what’s driving profit. When one campaign holds brand, competitor, and generic terms on a shared budget, the cheapest clicks eat the budget and the high-intent searches starve.

A clean structure separates by intent and by product:

• Brand and non-brand in separate campaigns
• Search, Performance Max, and Shopping kept apart
• High-value service lines or product categories in their own campaigns
• Lead gen separated from ecommerce where you run both

Inside each campaign, build ad groups around one theme rather than dumping fifty unrelated keywords into one group. If one ad can’t speak to every keyword in the group, split it. How to lower CPC in Google Ads has the restructuring moves.

Common restructuring moves are splitting a generic campaign by service, pulling competitor terms into their own ad group, and separating lead gen from ecommerce.

One caution against overcorrecting: over-fragmentation starves campaigns of the data Smart Bidding needs. A campaign producing very few conversions a month usually gets worse when you split it further.

Design ad groups around one intent

“Emergency plumbing repair” and “plumbing maintenance plans” are different searches with different urgency, different pages, and different copy. They belong in separate ad groups.

Single-theme ad groups remain the right model. Legacy single-keyword ad groups no longer behave predictably, because close variant matching expands even an exact match keyword into queries you didn’t pick. Group keywords tightly enough that one ad can speak to all of them, and write that ad to the specific need.

• Separate competitor terms where the page and offer differ, such as a comparison page rather than your standard product page.
• Keep informational long-tail queries apart from commercial ones. “How to fix a leaking faucet” and “emergency faucet repair” want different pages entirely.
• Check the ads actually reference the theme. One generic ad running across five groups throws away the specificity you just built.

Keywords and match types

Keyword strategy controls which auctions you enter, which controls what you pay and what converts.

Match types have loosened. Exact now includes close variants such as plurals, reordering, and implied words. Phrase matches queries sharing the meaning rather than the word order. Broad reaches widest, including synonyms and adjacent topics. Because close variants have expanded, watching the search terms report matters more than it used to.

A workable mix:

• Exact for proven, high-intent queries where you know the conversion rate
• Phrase for controlled expansion into related searches
• Broad only where tracking is solid, negatives are in place, and the budget can absorb the learning

Long-tail terms, usually three to five words, carry more specific intent at lower cost per click. That gap matters most in B2B and high-ticket work, where head terms are expensive and each conversion is worth a lot. For scale, the LocaliQ 2026 benchmarks put the all-industry average search CPC at $5.42, and competitive head terms run well above that.

Mine the search terms report under Campaigns > Insights and reports. When a query converts repeatedly, promote it into its own keyword in the right ad group. When it wastes money, make it a negative.

Cut wasted spend with negatives

Negative keywords stop your ads showing on searches that were never going to convert. It’s the fastest available improvement on most accounts, and it needs no new creative or budget.

Three levels of control:

• Account level for themes that never belong anywhere: profanity, job seekers, pure research
• Campaign level for categories irrelevant to that campaign’s strategy
• Ad group level for sculpting traffic between overlapping groups

A first cleanup usually adds “free,” “jobs,” “DIY,” “definition,” “salary,” and whatever industry mismatches show up in your own report.

The detail most guides get wrong: negative match types do not match close variants. An exact negative for [plumber] will not block “plumbers.” Plurals and misspellings have to be added explicitly, and the failure is silent, so the negative looks like it’s working while the plural keeps spending.

Review weekly, starting with high-spend, zero-conversion queries. Be careful with broad negatives on generic words, which can quietly block searches you wanted.

Full treatment in Google Ads negative keywords.

Pick a bidding strategy your data supports

Bidding comes after tracking and structure because automated strategies can only optimize toward the signals you give them. Feed them corrupted conversion data and they’ll chase the wrong outcome efficiently.

Google’s own Target CPA guidance points at roughly 30 conversions in the preceding 30 days. Below that, automation isn’t optimizing, it’s adding variance you can’t diagnose.

Set your starting target from your actual recent cost per lead. Google’s suggestion is a guess at your economics. If you’ve been running at $80 and you set a $30 target, you’ll throttle delivery and stall learning, and a campaign that stops serving doesn’t produce cheaper leads. It produces fewer.

Bid adjustments stack on top of each other, so device, location, and schedule modifiers compound in ways that are easy to miss. Review them together rather than one at a time.

Expect a learning period of roughly one to two weeks after any significant change, and resist editing targets inside it. Every edit restarts the clock.

Write ads that qualify the click

Ad copy’s job isn’t only to win clicks. It’s to attract the right searchers and repel the wrong ones, because a click you didn’t want costs exactly as much as one you did.

• Put the intent phrase in a headline. The searcher should recognize their own query.
• State the value plainly, then a specific action: “Schedule a free estimate,” not “Learn more.”
• Use qualifiers to filter. “From $5,000,” “commercial only,” “certified” all deter people who were never going to buy. Fewer clicks at a higher conversion rate is a lower cost per lead.

Responsive search ads accept up to 15 headlines and 4 descriptions. Supply real variety rather than 15 rewordings of one idea, and pin sparingly, since heavy pinning removes the algorithm’s ability to find the combination that works.

Test two or three genuinely distinct angles per ad group: features against benefits, urgency against value, pricing against proof. One variable per test.

Assets, formerly ad extensions

Assets expand the clickable area and give Google more formats to work with, which lifts click-through rate and Ad Rank without raising your bid. That makes them one of the cheapest levers in the account.

• Sitelinks to pages matching the ad group’s theme
• Callouts for differentiators: “No contracts,” “24/7 support”
• Structured snippets naming real service or product categories
• Call assets with call hours set, so phone leads reach a person and not voicemail
• Lead form assets where you want the form in the ad itself
• Location assets where geography is part of the decision

Audit them periodically. Remove what nobody engages with, and refresh offers and seasonal language rather than leaving last year’s promotion running.

Assets, ad relevance, and Quality Score all feed the same mechanism, which is what you actually pay per click. That’s covered in how to lower CPC in Google Ads.

Landing pages decide half the equation

If the page is slow, confusing, or doesn’t match the ad, cost per acquisition stays high no matter how good the ads are.

The mechanism is message match. The keyword triggers an ad, the ad promises something specific, and the page has to deliver that promise within seconds. Someone searching “B2B SaaS demo request” should land on a demo page, not a homepage that makes them hunt.

What the page needs: fast load on mobile, a form sized to the ask, one clear call to action above the fold, real proof, and none of the site navigation that gives visitors somewhere else to go.

Worth remembering that landing page experience is a Quality Score input, so the page affects what you pay per click as well as what converts. Both levers, same work.

Detail in Google Ads landing pages that lower cost per lead.

Close the loop on lead handling

Cost per lead is half the picture. What sales does next decides cost per acquisition. A $30 lead nobody calls back is more expensive than a $50 lead that closes.

• Connect Google Ads and your CRM so form fills, calls, and purchases are tracked through to revenue.
• Enforce fast follow-up on leads from high-intent search. Speed to first contact is one of the strongest predictors of close rate.
• Import offline conversions so closed-won data flows back and Smart Bidding optimizes toward customers rather than form completions.
• Add qualification fields to improve downstream close rates, accepting a small hit to raw conversion rate in exchange for leads worth calling.

Without that loop, the algorithm optimizes for the cheapest lead, which is usually the worst one.

Audiences

Audiences matter even in keyword campaigns, because they let you adjust bids and messaging based on who is searching rather than only what they typed.

Add them in observation mode first, so you can see how each performs before you change any bids.

• In-market segments identify people actively considering a purchase in your category.
• Detailed demographics give another cut: income, education, parental status.
• Remarketing lists for search ads let you bid harder on broad or generic terms for people who already visited. Google requires a list of at least 1,000 visitors before RLSA will serve.
• Customer Match uses your own email lists, either to exclude existing customers from acquisition campaigns or to target them with something relevant.

Compare cost per acquisition by audience as well as by keyword. The two views often disagree, and the audience view is usually the one that explains why.

Locations, devices, and schedule

Bad geographic or schedule settings can quietly double your cost per lead, and they’re easy to leave on defaults for years.

Set location targeting to “Presence: People in or regularly in your targeted locations.” The default also includes people merely showing interest in the area, which buys clicks from people who will never walk through your door.

Review performance by region, city, device, and hour, then adjust bids or split out the segments that perform. Raise bids in business hours for lead gen; cut back on a device type that converts far worse than the rest.

Don’t over-segment early. Start broad, gather four to eight weeks of data, then tighten on real patterns. Premature segmentation starves campaigns of the volume automation needs.

Campaign types and Performance Max

Search is the backbone for explicit intent. Someone typing a query is the most direct path to a conversion there is, and for B2B, higher-value, and lead-gen accounts it’s what you build first.

Performance Max reaches all of Google’s inventory: Search, Display, YouTube, Maps, Discover, Gmail. You supply assets, goals, and audience signals, and Google handles the rest. It can scale, but it’s opaque, with far less control over placement and targeting. With loose conversion goals or a thin budget it tends to find cheap, low-quality conversions. Google recommends a daily budget of at least three times your target CPA for PMax, and it has added campaign-level negative keyword lists and channel reporting, so it’s less of a black box than it was.

A sensible order: focused Search first, then PMax or Shopping for catalogs, then Display or Video for remarketing. Give each type its own budget, goals, and measurement, or you’ll blend intent levels and lose the ability to read cost per acquisition at all.

Also check your search partner settings. That network extends reach beyond Google Search and can add volume, often at a lower conversion rate. Report on it separately and turn it off if it isn’t carrying its weight.

Keyword research doesn’t stop

The queries people type shift with seasons, trends, and competitor behavior.

Use Keyword Planner, search suggestions, and your own search terms report to find new profitable queries and prune the dead ones. The Google Ads Transparency Center lets you see which ads competitors are running, which is a reasonable read on what they’re prioritizing.

Alternate between expansion phases, where you test new terms, and consolidation phases, where you concentrate budget on proven winners. Running only one of those modes is how accounts either stall or bleed.

Analytics and cross-channel

Linking GA4 to Google Ads shows post-click behavior, assisted conversions, and paths that mix paid and organic.

Build GA4 segments for high-value behavior, such as people who viewed several product pages or spent real time on pricing, and bring them back into Google Ads as audiences.

Then compare cost per acquisition across channels and move budget accordingly. If GA4 shows a page with a high bounce rate specifically from paid traffic, that page goes to the top of the redesign list, because every bounced paid click is spend with nothing to show for it.

Experiments, not random changes

Change bids, ads, and pages at once and you’ll never know which one moved the number.

Google Ads experiments split traffic inside a campaign so you can test one variable while the control keeps running. Worth testing: headline angles, pricing shown against pricing withheld, different calls to action, or Target CPA against Maximize Conversions on a mature campaign.

Set the conversion threshold before the test starts, while you still can’t see which way it’s leaning. Keep a log of hypothesis, dates, and result, so what you learn compounds instead of evaporating.

The operating cadence

A cadence prevents both failure modes: neglecting the account until it degrades, and changing so much that Smart Bidding never leaves its learning period.

Daily. Spend anomalies, disapproved ads, broken landing pages, and any sign that tracking has stopped firing.

Weekly. Search terms and negatives. Pause or reduce non-converting keywords. Check campaigns creeping above their CPA threshold. Review by audience, device, and time.

Monthly or quarterly. Structural changes, whether PMax should expand or pull back, revisiting targets against updated business results, deeper cuts by geography and demographics, creative refreshes, and landing page work where the post-click data shows drop-off.

The point is to watch the trend line in cost per lead rather than reacting to daily noise. Steady inputs are what automation needs.

Optimization score and Recommendations

Google’s optimization score and Recommendations tab are useful signals and a poor autopilot.

The conflict of interest is worth stating plainly: many recommendations push higher budgets and broader matching, which serves Google’s revenue whether or not it serves your cost per acquisition. Judge each one against your own target. The score is not the target.

Usually helpful: fixing disapproved ads, adding missing sitelink or callout assets, updating tracking for new features.

Usually risky: aggressive broad match expansion before the account is efficient, automatic budget increases before CPA is stable, switching bid strategies mid-learning.

Turn off auto-apply for anything touching budget, match type, or bidding, and review manually. Measure before and after on any significant change, to confirm it lowered cost per lead rather than simply raising spend.

The checklist

1. Verify conversion tracking: primary actions correct, tags firing, no double counting
2. Set business targets from your own margins
3. Clean the structure: separate by intent, product, brand and non-brand
4. Build ad groups around tight themes, one intent per group
5. Choose match types deliberately: exact for proven, phrase for expansion, broad with controls
6. Add negatives and review the search terms report weekly
7. Pick a bidding strategy your conversion volume supports
8. Write ads that qualify the click, and add the assets
9. Align landing pages: message match, fast on mobile, one call to action
10. Close the loop with the CRM and offline conversion imports
11. Layer audiences in observation, then adjust
12. Set location, device, and schedule from real data
13. Sequence campaign types: Search, then PMax or Shopping, then Display and Video
14. Hold a daily, weekly, and monthly cadence

Cost per lead and cost per acquisition are the scorecard when deciding what to work on in any given week. Each area here can be deepened into its own playbook, and several already are.

Optimizing Google Ads is ongoing work. Respect the order, and each improvement compounds on the last instead of undoing it.