Google Ads Optimization: Which Lever To Pull

Last updated: October 18, 2026

1. Executive Summary: Five Key Principles for a Live Account

This article is for accounts already spending money. Every lever described below has a number attached to it: what it moves, how much, and how long before the result can be read. The five principles below are the frame for everything that follows.

• Principle 1 — Diagnose the constraint before touching the budget or bids. Search lost impression share (budget) and Search lost impression share (rank) are different problems with different fixes. An account where lost IS (rank) is 38% and lost IS (budget) is 3% will not improve if the daily budget is raised; the constraint is Ad Rank, and the fix is bid strategy, quality, or both. Raising the budget in that account wastes money without moving impression share. Read those two columns first, every time.
• Principle 2 — Budget is the fastest lever but the most expensive one. Removing a budget constraint (Search lost IS [budget] above 10%) can recover eligible impressions within one to three days. There is no published universal percentage lift, because it depends on the size of the constraint and the quality of the incremental traffic; a reasonable working range for a genuinely budget-limited campaign with stable Ad Rank is 5–25 additional impression-share points. The cost is linear: every added impression costs the same CPC as the impressions the account is already buying. Before increasing budget, confirm that the marginal CPA is within target.
• Principle 3 — Smart Bidding needs at least 15–30 conversions in 30 days to stabilize; below that threshold, the algorithm oscillates and the result is unreadable. Switching a campaign from Maximize Clicks to Target CPA when it has 8 conversions in the last 30 days will trigger a learning period of unpredictable length and is likely to collapse impression share before it recovers. The right fix below that threshold is to improve conversion tracking, consolidate campaigns, or use Maximize Conversions without a target until volume is sufficient. The direction of change is not in dispute; the timeline is.
• Principle 4 — Quality Score components move slowly, and the ceiling is usually the offer, not the copy. Ad relevance can improve within two to four weeks after restructuring ad groups and rewriting headlines. Expected CTR and landing page experience require more traffic and more time: four to eight weeks for a directional signal on a medium-volume keyword. An account where every keyword reads Average or Above average on all three components and still carries a $180 cost per conversion has hit a structural ceiling. No quality optimization will move that number materially; the constraint is the offer, the price, or the geography, not the ad.
• Principle 5 — Some campaigns are close to their economic ceiling and no lever will rescue them without a change outside the platform. A non-brand Search campaign with 78% Search impression share, Above average quality on all components, and a conversion rate of 1.8% at a $6 CPC is paying $333 per conversion. Lifting impression share to 90% adds volume at the same unit economics. Improving Ad Strength from Poor to Excellent is associated with roughly 15% more clicks on average^[1], which at 1.8% conversion rate produces more conversions at the same $333 cost. The real lever is the landing page conversion rate: moving from 1.8% to 3.6% cuts cost per conversion to $167 without touching the bid. Advertisers who have exhausted platform levers must fix the offer, the page, or the price.

2. Benchmarks and Numbers at a Glance

All figures in the table above are vendor claims unless labeled as STUDY. Where country is not stated, do not apply these as US-specific benchmarks; use them as directional references and validate against account data. The CPC figure reflecting a 12.88% year-over-year increase is from October 2026^[2] and is the most recent available.

3. The Share and Performance Metrics That Tell You Which Lever to Pull

Every lever in this article moves one or more of the metrics below. Before pulling any lever, read the account’s numbers in these columns. The diagnostic table states what each metric actually measures, where to find it, which campaign types report it, and what a bad number tells you to do next.

Performance Max and Demand Gen do not report Search impression share, Search lost IS (budget), Search lost IS (rank), or Search top impression share. Performance Max reports a combined Search + Shopping impression share that is not comparable to a Search campaign’s metric.^[4] For Performance Max, use cost, conversions, conversion value, asset-group performance, and search-term insights as proxies. For Demand Gen, use impressions, reach, frequency, CTR, and conversions; traditional share metrics do not apply.^[4]

Failure mode: An advertiser who optimizes Search impression share in a Performance Max account by raising the Target ROAS is reading the wrong metric for the wrong campaign type and may starve Performance Max of budget while the Search campaign’s share metric remains unaffected by the change.

4. The Auction: What Actually Decides Whether You Show and Where

Every Google Search auction calculates an Ad Rank for each eligible ad. Ad Rank determines both whether an ad is eligible to show and what position it receives among eligible ads. Google does not publish a numeric equation, but its current public description identifies six inputs.^[3]

• Bid: the auction-specific amount associated with the active bid strategy. Under Smart Bidding, Google sets this automatically within strategy constraints. The advertiser controls the strategy and targets, not the individual auction bid.
• Auction-time ad quality: Google’s real-time assessment of expected click-through rate, ad relevance, and landing page experience for this specific impression opportunity. This differs from the reported Quality Score, which is a 1–10 diagnostic updated over a 90-day comparison window.^[5] The auction-time signal accounts for the exact query, device, location, time, user context, and landing page.
• Ad Rank thresholds: minimum Ad Rank requirements an ad must clear to be eligible. Thresholds vary by position sought, query topic, location, device type, and other user signals.^[6] An advertiser cannot read or set thresholds directly.
• Search context: the user’s location, device, time of search, nature of the query, other ads and organic results on the page, and other user signals.^[3] The advertiser cannot control the user’s context but can adjust targeting to change which contexts the campaign enters.
• Competitiveness: the bids, quality scores, thresholds, and context of every other eligible advertiser in that auction.^[3] Competitors’ choices are outside advertiser control.
• Expected impact of assets and ad formats: Google estimates whether sitelinks, callouts, images, call assets, and other formats will improve the ad’s usefulness and the user’s experience.^[3] Advertisers create and maintain assets; Google decides whether and how to show them.

The older shorthand “Ad Rank = bid × Quality Score” is not the current model. Quality Score is a keyword-level diagnostic, not the live auction quality signal. Google explicitly states that Quality Score is not an input in the live auction.^[1]

What the advertiser directly controls, influences, or cannot touch:

Why a higher bid does not guarantee a higher position. A bid increase raises the bid component of Ad Rank but leaves every other component unchanged. If a competitor’s auction-time quality is materially stronger, their Ad Rank can exceed the higher bid’s Ad Rank. A bid can also increase cost without clearing the threshold for a more prominent format. Conversely, an advertiser with stronger quality can outrank a higher bidder while paying less per click, because actual CPC is determined by the Ad Rank of the competitor ranked below, divided by the winner’s quality, plus one cent.^[3]

Failure mode: An advertiser who raises bids to recover impression share when Ad Rank thresholds are the binding constraint spends more per click without gaining position, because the threshold requires a quality improvement that bid alone cannot supply.

5. Diagnosing a Campaign in Thirty Minutes: From Symptom to Lever

The order of operations for the first thirty minutes in an underperforming account is: (1) check campaign status and conversion tracking, (2) read Search lost IS (budget) and Search lost IS (rank) at campaign level, (3) check Quality Score components for keywords above $100 monthly spend, (4) review the Search terms report for obvious irrelevant traffic, (5) check the bid strategy and whether it has sufficient conversion volume to be stable, (6) open Auction Insights to assess competitive pressure. Do not change anything in the first fifteen minutes. Diagnosis before action.

Failure mode: An advertiser who pulls the budget lever first—because it is the most visible recommendation in the interface—when the account’s binding constraint is Ad Rank or a broken conversion tag will spend more money on the same underperforming traffic, making the CPA worse and the account harder to diagnose.

6. The Complete Lever Map: What Each Lever Moves, How Fast and How Much

The two tables below are a scanning reference for every major lever available in a live Google Ads account. Table 6a answers what each lever moves and how quickly you can read the result. Table 6b, covering the same levers in the same order, answers why each lever works and the specific account conditions that make it the right or wrong move. Rows are ordered fastest-to-read first, so the lever with the shortest feedback cycle appears at the top. Every row in Table 6a carries a number or range in the Realistic size of effect column; every row in Table 6b carries a testable condition in the Use it when and Do not use it when columns.

Before pulling any lever, confirm that conversion tracking is firing correctly and recording actions that represent real business value. Every other optimization in this article is downstream of that single precondition. An account with broken or miscounted conversions will make Smart Bidding learn the wrong objective, make Quality Score diagnostics misleading, and make every before-and-after comparison unreliable. Fix measurement first; then pull levers.

What each lever moves, and how fast

Table 6b lists the same levers in the same order as Table 6a. Use both tables together: confirm that the time horizon in 6a fits your reporting cycle, then use 6b to verify that your account meets the conditions that make the lever active.

When to pull each lever, and when not to

Three levers most often pulled first that should not be. The most common wrong first move is raising the daily budget when Search lost IS (budget) reads near zero—the constraint is rank, not money, and the extra budget is absorbed at higher CPA without recovering share. The second is switching to Target CPA before conversion tracking is verified—Smart Bidding then optimizes toward a corrupted signal and the learning period burns budget for nothing. The third is adding negative keywords before fixing ad relevance when Ad relevance reads Below average—negatives shrink an already under-ranking campaign and worsen data density. In all three cases, the correct first move is in Table 6b’s Use it when column for levers 3, 10, and 2 respectively: check the IS (budget) reading, verify tracking, and address Ad relevance before anything else.

Failure mode for this section: pulling levers in the wrong order—particularly changing bid strategy before fixing conversion tracking—trains Smart Bidding on corrupted data, resets the learning period when tracking is later corrected, and can waste several weeks of budget at a CPA the account cannot carry.

7. Budget Levers: Lost Impression Share (Budget) and What More Money Actually Buys

A budget increase answers exactly one diagnostic question: was the campaign prevented from competing in auctions it was otherwise eligible to win? If Search lost IS (budget) is near zero, adding money does not recover impressions, does not lower CPA, and does not expand click volume in any meaningful way. The budget lever is only active when the campaign is spending to its daily cap before the auction day is over. Confirm this with two data points: Search lost IS (budget) greater than 10%, and the campaign’s daily spend curve showing it hitting the cap before 6 p.m. local time consistently over the last seven days. When both conditions are true, the campaign is genuinely budget-constrained and more money will produce more impressions and conversions—at a marginal cost that rises as share climbs.

The two questions a budget increase actually answers

Question one: is the campaign losing impressions purely because the budget runs out? This is a yes-or-no diagnostic answered by Search lost IS (budget). Non-brand Search campaigns typically lose 32% of eligible impressions to budget on average^[2]; brand campaigns lose around 15%^[2]. Those benchmarks tell you whether your campaign is average, better, or worse—not whether buying back those impressions is economically rational.

Question two: is the marginal impression worth buying? This requires arithmetic on your own account numbers. The first dollar of budget recovery captures the most accessible auctions—typically cheaper, higher-intent queries the campaign was missing in its peak hours. As Search impression share rises toward 70%, 80%, and 90%, the remaining eligible impressions are increasingly contested, lower-converting, or appearing at times with weaker purchase intent. The marginal CPA rises. There is no published universal figure for how fast it rises, because it depends on auction prices, query mix, competitors, and conversion rate in those marginal auctions. The arithmetic below illustrates the shape of the curve from account-level numbers.

The diminishing return on budget: what the arithmetic shows

The shape of the diminishing return is not linear. The first 10–20 impression-share points recovered from a budget constraint are typically the cheapest, because the campaign was missing high-frequency, high-intent queries it had already won in the auction but could not afford to serve consistently. As IS rises, the incremental queries are lower-frequency, more competitive, or served in lower-converting time windows. This is why Google’s own guidance recommends evaluating marginal CPA and marginal ROAS after a budget change rather than simply celebrating higher impression share.^[3]

A practical rule: budget increases of 10–20% at a time allow the campaign to spend up while giving you 7–14 days of conversion data to check whether marginal CPA is staying inside the target before committing a larger increase. A single doubling of budget can push the campaign into marginal auctions it cannot convert economically, and because Smart Bidding also needs time to learn the new volume, the first two weeks after a large budget change will look worse than steady state regardless of direction.

The campaign is near its budget ceiling—and no further budget increase is warranted—when: Search lost IS (budget) remains above 5% after multiple increases, actual CPA is already at or above target, and the marginal-CPA arithmetic above shows the next IS point costs more than the target CPA. At that point, the binding constraint is offer economics, Ad Rank, or both. More money does not solve either.

Shared budgets and portfolio allocation

When multiple campaigns share the same conversion goal, a shared budget can improve allocation efficiency by letting higher-opportunity campaigns draw more spend on high-demand days while lower-demand campaigns underspend. Google recommends shared budgets alongside portfolio bid strategies for campaigns with the same objective.^[3] The improvement from shared budgets is typically allocation efficiency—5–15% more conversions from the same total spend in accounts where individual campaign budgets were previously misaligned. This is not a demand-creation lever; it is a redistribution lever. If the total budget across all campaigns is already too low to generate the needed conversion volume, a shared budget redistributes the shortfall rather than solving it.

When to pull it, and when not to

Pull the budget lever when: Search lost IS (budget) is above 10% and the campaign is spending to its daily cap before peak conversion hours, marginal CPA arithmetic confirms the next 10 IS points can be recovered at a CPA inside the account target, and conversion tracking is verified so that the marginal conversion count is reliable. Raise budget in steps of 10–20% and recheck IS (budget) and marginal CPA after 7–14 days before the next increase.

Do not pull the budget lever when: Search lost IS (budget) is near zero—the constraint is rank, not money, and raising budget delivers the same impressions at the same rank and the same CPA with more total spend. Also do not raise budget when actual CPA is already at or above target; buying more volume at an unprofitable CPA compounds the loss. Do not raise budget as a substitute for fixing Ad Rank: if Search lost IS (rank) exceeds 20%, address bid strategy, target CPA, ad relevance, and landing page experience first. Those levers can recover impressions without increasing spend. Finally, do not raise budget during the learning period after a bid-strategy change; the system needs stable conditions to calibrate, and a simultaneous budget change extends the learning period and makes the result harder to interpret.^[3]

Failure mode for this section: raising the daily budget when the binding constraint is Search lost IS (rank) rather than IS (budget) results in higher total spend at the same weak Ad Rank, the same or higher CPA, and no recovery in impression share—because the denominator of eligible impressions does not change and the campaign is still losing auctions on quality, not on budget.

8. Bidding Levers: Strategy, Targets and Lost Impression Share (Rank)

Bid strategy is the single lever with the widest reach across all four dimensions that matter: Search impression share, click-through rate, cost per click, and cost per conversion. It also carries the highest risk of a learning-period reset when changed incorrectly. This section explains which strategy to choose, how large a target change is safe, what triggers learning, and what breaks when the lever is pulled badly.

Campaign types: All bid strategies below apply to Search. Maximize Conversions, Maximize Conversion Value, Target CPA, and Target ROAS are also available for Performance Max, Shopping, and (in limited form) Display. Target Impression Share is Search-only. Manual CPC is available for Search and Standard Shopping but is increasingly removed from new campaign creation flows. Demand Gen, Video, and App have their own strategy sets—see sections 13 and 14.

Failure mode: The most common bidding failure is changing strategy or target while conversion volume is below the practical threshold, which puts the campaign in a prolonged learning period, triggers bid volatility, and can burn 30–50% of a monthly budget at CPAs 40–80% above target before the account owner notices the instability in reporting.

Safe target change size, learning periods, and what resets them

The conservative rule is to move a Target CPA or Target ROAS by no more than 10–20% in a single change, then hold for at least one full conversion cycle—typically 14–21 days for accounts with 30-day conversion windows—before moving again.^[3] A change from a $120 Target CPA to a $95 Target CPA in one move is a 21% reduction: it exceeds the safe band and commonly triggers a sharp drop in impression share of 15–30 points within 48 hours, accompanied by bid volatility as the system searches for auctions it can win at the new constraint.

Learning is triggered by: switching bid strategy; changing Target CPA or Target ROAS by more than roughly 15–20%; changing the daily budget by more than 20% in one move; adding or removing campaigns from a portfolio strategy; changing the primary conversion action; or altering campaign structure materially (new ad groups, pausing large keyword sets, changing geography).^[3] During learning, expect CPA and ROAS to be 20–50% worse than steady-state and do not make a second major change because it resets the clock. Google does not publish a fixed day count for learning; in practice it runs 7–14 days for campaigns averaging 5–10 conversions a week, and longer for lower-volume campaigns.

Share arithmetic note: raising a Target CPA increases the numerator of Search impression share (more impressions received) while the denominator—eligible impressions—stays roughly constant, so measured share rises. Lowering a Target CPA has the opposite effect. Target Impression Share bidding directly targets the numerator but can simultaneously expand the eligible pool by entering more auctions, so the measured share does not always reach the stated target immediately.

When to pull it, and when not to

Pull the bidding lever when: Search lost impression share (rank) is above 20% and the primary cause is low bid rather than weak ad quality (check that Expected CTR and Ad relevance read Average or above for the keywords carrying the majority of spend); or when the current strategy is Maximize Clicks but conversion tracking has now been live for 30+ days with 30+ conversions, making Target CPA viable. Also pull it when Target CPA or Target ROAS is set more than 20% below actual account performance, causing unnecessary volume restriction. Read Search lost IS (rank) under Campaigns → Columns → Competitive metrics.

Do not pull the bidding lever when: Search lost IS (budget) is above 15% and IS (rank) is below 10%—the constraint is budget, not bid, and changing strategy wastes a learning period on the wrong problem. Do not change bid strategy when conversion tracking was recently modified, added, or repaired; the new data stream will distort learning. Do not tighten a Target CPA or ROAS when the campaign is already in a learning period from a prior change. If the account has fewer than 15 conversions in the last 30 days, do not move to Target CPA or Target ROAS; stay on Maximize Conversions or Manual CPC until volume is sufficient.

9. Quality Levers: Expected CTR, Ad Relevance and Landing Page Experience

Quality Score is a 1–10 diagnostic reported at the keyword level for Search campaigns. It is not a direct auction input, a campaign-level metric, or a metric available in Performance Max, Demand Gen, Video, Display, or App campaigns.^[5] In those campaign types there is no Quality Score column; use auction-time signals (placement exclusions, creative quality diagnostics, asset-group feedback) as proxies. Shopping has its own relevance signals in Merchant Center rather than a Quality Score.

The three components—Expected CTR, Ad relevance, and Landing page experience—are each rated Above average, Average, or Below average relative to advertisers whose ads showed for the same searches in the preceding 90 days.^[5] The 90-day window describes the comparison baseline, not a guaranteed refresh interval. Low-volume keywords can show stale ratings for weeks after a change.

What Quality Score is diagnostic of: structural weaknesses in keyword-to-ad alignment, ad copy relevance, and landing page usefulness. What it is not diagnostic of: conversion rate, CPA, ROAS, or profit. A keyword can carry a Quality Score of 9 and convert at 1% because the offer is wrong. Chasing Quality Score in that situation is the wrong use of the next four weeks.

Failure mode: An advertiser who rewrites ads solely to push Quality Score from 5 to 7 on low-volume keywords—where the rating is statistically noisy—spends 3–4 weeks on a metric that moves slowly, cannot be read reliably below 200 impressions per keyword, and does not directly move CPA or ROAS.

Which component is cheapest to fix, and which is most expensive

Usually cheapest: Ad relevance. Rewriting headlines to include the keyword theme, splitting a bloated ad group into two tighter ones, and adding a description that matches intent can move a Below average rating to Average within one to two weeks on keywords receiving 50+ impressions a day. No developer time is required. The risk is that a relevance-driven rewrite reduces CTR if the new copy is less compelling, so test before replacing the control.

Usually intermediate: Expected CTR. Improving expected CTR requires either better copy (fast) or a better offer (slower). A creative refresh can move expected CTR directionally in 2–4 weeks on high-volume keywords, but low-volume keywords may take 6–8 weeks to accumulate enough impressions for a meaningful diagnostic update. Adding assets (sitelinks, callouts, images) does not improve the reported Expected CTR component because Google removes asset influence from it.^[6]

Usually most expensive: Landing page experience. Moving from Below average to Above average typically requires design, development, mobile QA, and sometimes legal review. Timeline from decision to live page: 4–12 weeks depending on the organization. The business payoff is often the largest of the three components because landing page improvements compound into conversion rate gains that no bid or copy change can replicate.

How long each component takes to move

There is no published fixed day count for any component update.^[5] The practical estimates below are based on auction mechanics, not documented Google timelines. A keyword receiving 200+ impressions a day can show directional changes in Ad relevance within 7–14 days of a restructure. Expected CTR on the same keyword may take 14–28 days. Landing page experience may take 3–6 weeks after a page change because Google must recrawl, re-evaluate, and accumulate new traffic before the diagnostic updates. Keywords receiving fewer than 20 impressions a day may show no meaningful change for 8–12 weeks regardless of the improvement made.

Account states where chasing Quality Score is the wrong use of the next four weeks

Do not prioritize Quality Score work when: (1) Search lost IS (budget) exceeds 20%—the campaign is not entering enough auctions for quality to matter at the margin and budget is the binding constraint; (2) fewer than 15 conversions in the last 30 days mean the account is too thin to read any lever reliably; (3) the conversion action is broken or miscounted—fixing measurement returns more value than any copy or relevance change; (4) the offer itself is uncompetitive (price, availability, or guarantee) because no Quality Score improvement will overcome a structurally weak offer at the landing-page stage.

When to pull it, and when not to

Pull quality levers when: Search lost IS (rank) is above 20% AND at least one keyword carrying more than $100 of monthly spend shows Ad relevance Below average or Expected CTR Below average in the keyword diagnostics column (Campaigns → Audiences, keywords and content → Search keywords → Modify columns → Quality Score). This confirms rank loss is partly quality-driven and not purely a bid problem. Also pull landing page levers when Landing page experience reads Below average on keywords accounting for more than 30% of conversion volume—because fixing it improves both Quality Score and conversion rate simultaneously, the highest ROI quality action available.

Do not pull quality levers when: Search lost IS (rank) is below 10% and all three components read Average or Above average—quality is not the constraint and further copy polishing will not move the metrics. Do not invest in landing page development when the campaign has fewer than 100 monthly clicks to that page; the traffic volume is too small to read conversion rate changes reliably and the budget is better spent on acquiring traffic first. If the primary constraint is budget (IS budget loss above 20%), fix that before addressing quality, because quality improvements on a budget-constrained campaign do not recover the lost impressions.

10. Keyword, Match Type and Search Term Levers

Match type determines which pool of searches the campaign enters. Bid strategy then prices each auction within that pool. The two levers interact: broadening match without a reliable conversion signal means Smart Bidding has no basis for filtering poor-quality auctions within the expanded pool, and the campaign buys traffic it cannot evaluate. Narrowing match without checking Search lost IS (budget) first can raise impression share mechanically while reducing absolute impressions and conversions.

Campaign types: Match types apply to Search campaigns. Performance Max uses search themes as signals rather than traditional keywords; Shopping uses product feed attributes rather than keywords. Match type is not a lever in Demand Gen, Video, or App.

Failure mode: Switching high-spend keywords from exact to broad match while the campaign is on Maximize Clicks—without conversion tracking—causes the system to buy cheap, high-volume, low-intent clicks across an enormous query pool, exhausting budget at a CPA the account cannot measure or control.

Share arithmetic: Broadening match expands the eligible impression pool (the denominator of Search impression share). Even if absolute impressions rise, measured impression share often falls because the denominator grows faster than the numerator. This is not a performance decline—evaluate incremental clicks, conversions, and CPA, not share alone. Conversely, narrowing match shrinks the denominator and can mechanically raise measured impression share while the campaign reaches fewer people.^[4]

Search term review cadence

Review the search terms report (Campaigns → Insights and reports → Search terms) at least weekly for campaigns spending more than $3,000 a month. For campaigns spending $500–$3,000 a month, review every two weeks. For campaigns below $500 a month, monthly is acceptable. Do not add a negative keyword from a single click unless the query is clearly prohibited or commercially impossible. For conversion decisions, require at least 20–30 clicks on a query before treating non-conversion as evidence of poor intent; lower-volume queries need even more patience given conversion lag.^[1] Google withholds queries below privacy thresholds, so the report is not a complete log of every search—treat the visible portion as a sample.

When adding negatives from the search terms report, the interface defaults to negative exact unless another type is specified.^[1] Apply negatives at the campaign level for query patterns specific to one campaign; use shared negative keyword lists (Tools → Shared library → Exclusion lists) for patterns that apply across multiple campaigns.

Broad match and the eligible pool warning

Google’s documented recommendation is to use broad match with Smart Bidding when the account has clear goals and reliable conversion measurement.^[1] The critical caveat: broad match uses the landing page, assets, and user context as signals alongside the keyword. If the landing page is generic or the conversion action is low-quality (form fills that are rarely qualified), broad match will expand into queries that are semantically adjacent but commercially poor. The conversion signal drives broad match quality—not the match type itself.

When to pull it, and when not to

Pull match type and keyword levers when: the Search terms report shows that 20% or more of clicked queries over the last 30 days are irrelevant or off-intent (pull it toward tighter match or add negatives); or when exact and phrase keywords together are generating fewer than 60% of the impressions the account could be eligible for (Search impression share below 50% with IS rank loss above 25%), and conversion tracking is reliable enough to support broad match with Smart Bidding. Pull the negative keyword lever whenever a query with 30+ clicks shows zero conversions and does not fit any plausible commercial intent.

Do not pull the match type lever when: conversion tracking is broken, under-counting, or recently changed—broadening match in that state sends the system into an unmeasurable query pool. Do not broaden match if the campaign is already spending its full budget; adding broader eligibility without more budget means the existing budget is simply redistributed toward lower-intent queries, not additional reach. Do not add negatives to a campaign that is already missing more than 30% of eligible impressions due to rank loss—the constraint is bid quality, not query irrelevance, and adding negatives shrinks an already-restricted pool.

11. Creative Levers: What Actually Lifts Click-Through Rate

Click-through rate is determined by three factors in declining order of an individual advertiser’s control: (1) ad position—ads in the top two positions receive structurally higher CTR regardless of copy; (2) query mix—brand, high-intent, and navigational queries carry baseline CTRs 3–5× higher than generic informational queries; (3) the ad itself—copy, offers, assets, and relevance. A creative refresh that improves CTR on paper may simply reflect a position improvement or a shift toward more branded traffic. Always segment CTR by brand versus non-brand and by position before attributing gains to the creative.

Campaign types: RSA creative levers apply to Search. Performance Max asset groups have overlapping principles but different mechanics (see section 13). Demand Gen, Display, and Video have separate creative formats. Shopping has no text ad copy—impression share and CTR there are driven by product title, image, and price (see section 15).

Failure mode: Declaring a creative test a winner after fewer than 100 clicks per variant produces a false positive roughly 40% of the time at the 5% significance level; the account then pauses a better ad and locks in an inferior one, degrading CTR by 5–15% for months before the error is detected.

RSA asset counts and what the evidence says

A responsive search ad supports up to 15 headlines (30 characters each) and up to 4 descriptions (90 characters each). Google shows up to 3 headlines and 2 descriptions per impression. Google reports that moving from one RSA to two in an ad group produced an average 6.6% more conversions at similar CPA, and adding a third RSA added a further 3.7% on average.^[1] These are conversion benchmarks, not CTR figures, and they are observational—the ads that benefited may also have had better offers or better coverage.

There is no published headline-by-headline CTR curve. The reliable operational conclusion is: supply all 15 headlines where they are genuinely distinct, because more varied inputs give the system more combinations to test across query types. Near-duplicate headlines waste slots. Asset diversity matters more than slot count.

What pinning does to Ad Strength and performance

Pinning fixes a headline or description to a specific position (Headline 1, Headline 2, or Description 1). It is appropriate for legally required disclosures, regulated product language, or a validated offer that must always appear. Its cost is reduced combination testing. An Optmyzr study of more than 93,000 RSAs found that fully pinned ads sometimes showed strong CTR and conversion metrics, but the authors caution that selection bias makes causality uncertain—those ads were likely already well-optimized before pinning.^[9] Google recommends pinning 2–3 unique headlines in each pinned position when pinning is necessary to preserve some combination flexibility.^[1] Excessive pinning lowers Ad Strength, but Ad Strength is not a Quality Score input and does not directly set bids.

Is Ad Strength worth chasing?

Google states that improving an RSA from Poor to Excellent Ad Strength is associated with 15% more clicks and conversions on average.^[1] An Optmyzr analysis of more than 1 million ads found no clear correlation between Ad Strength labels and CTR or overall performance.^[9] These findings conflict. The conservative interpretation: use Ad Strength as a checklist to catch missing or repetitive assets, but do not sacrifice a proven offer headline to reach Excellent. Measure performance by impression-weighted CTR, conversion rate, and CPA—not by the Ad Strength label itself.

Realistic CTR movements from a creative refresh

There is no published universal figure for the CTR lift from a creative refresh. Based on auction mechanics and the Optmyzr asset study,^[9] realistic planning ranges are:

• 0–5% relative CTR improvement when existing ads are mature, assets are mostly complete, and the offer is already competitive.
• 5–15% relative improvement when asset slots are half-empty, headlines are near-duplicates, or the offer is not stated explicitly in any headline.
• 15–30% or more relative improvement when the prior RSA has fewer than 6 headlines, no offer language, and no call to action—this is a recovery from a severely underdeveloped starting state, not a normal expectation.

A 5% relative CTR increase on a campaign with a 6.66% baseline CTR^[2] means moving to 6.99%—a difference of 0.33 percentage points. To read that difference with 95% confidence requires at least 500 clicks per variant, preferably 1,000. Below 200 clicks per variant, a 5% relative movement is statistically indistinguishable from noise. Do not declare a creative winner until each ad has received a minimum of 200 clicks, and use 500+ for conversion-rate comparisons given conversion lag.

What the position and query mix contribution means for creative decisions

If Search absolute top impression share rises (because Target CPA was raised and the campaign is winning more top positions), aggregate CTR will rise even with identical copy. Conversely, if the campaign expands to broader match types and enters lower-intent queries, CTR will fall even if the ads improve. Before attributing a CTR change to creative, check whether Search abs. top IS and the brand/non-brand split moved in the same period. If they did, isolate the creative effect by segmenting by match type and by brand versus non-brand in the campaign data.

When to pull it, and when not to

Pull creative levers when: Expected CTR reads Below average on keywords carrying more than $100 of monthly spend (Campaigns → Search keywords → Modify columns → Quality Score → Expected CTR); or when the campaign’s RSAs have fewer than 8 distinct headlines or have not been updated in more than 6 months; or when ad group-level CTR is more than 20% below the account’s brand average after controlling for position. These are testable conditions, readable in the keywords and ads reports.

Do not pull creative levers when: the campaign has fewer than 200 clicks a month per ad group—there is not enough volume to read a creative test within a reasonable timeframe and the effort is better directed at impression share or bid levers that move volume first. Do not change creative mid-test if a bid strategy change or match type change was made in the same 14-day window—the confounded signals make the creative result unreadable. Do not chase Ad Strength from Good to Excellent by replacing a headline with a proven high-CTR offer message just to satisfy the diagnostic; conversion economics take priority over the label.

12. Audience, First-Party Data and Signal Levers

Audience levers split into two fundamentally different mechanisms. An audience used for targeting restricts who sees your ad—if someone is not on the list, they are ineligible. An audience used as a signal or observation layer does not restrict eligibility; it tells the bidding system that users matching the audience are higher or lower value so it can shade bids accordingly. Confusing the two is the most common audience mistake in a live account, because a targeting decision that was meant to be a signal can collapse volume overnight, while a signal layer that was intended to restrict targeting produces no exclusion at all.

What each audience mechanism is allowed to do by campaign type

In Search, set an audience to Observation unless you have a specific reason to restrict — for example, a members-only promotion. Targeting mode in Search eliminates all non-audience queries, which typically cuts impression share by 60–90% in a non-brand campaign and requires 2–4 weeks of data collection before the cost per conversion stabilizes. There is no published universal figure for this drop; the arithmetic below assumes the audience covers roughly 10–15% of eligible traffic, which is common for remarketing lists on non-brand campaigns.

Minimum list sizes and time to usability

Google does not publish an exact matched-user percentage for Customer Match; typical industry experience is 30–60% match rate on hashed email uploads, so a list of 3,000 emails may yield 900–1,800 matched users — below the 1,000-user activation threshold. Upload at least 5,000 email addresses to reliably clear activation, and at least 15,000 to generate a usable Smart Bidding signal. Failure mode: uploading a small Customer Match list, setting it to Targeting mode in Search, and watching impression volume collapse while conversions from the tiny eligible pool are attributed to the “audience lever” rather than to the pre-existing intent of users who were already searching.

What each lever realistically moves

For a Search campaign running on observation with a well-populated remarketing list (10,000+ users), adding a bid adjustment of +20% to in-market audience members typically shifts 5–15% of total spend toward higher-intent users, with a 10–25% improvement in conversion rate for that segment compared with the campaign average — there is no published Google figure for this; the range is reasoned from auction mechanics and the typical conversion-rate differential between site visitors and cold traffic. The result can be read after 3–4 weeks. At lower traffic volumes the segment conversion data is too sparse to read reliably inside 6 weeks.

Customer Match used as an observation layer in a Target CPA campaign does not change the CPA target — it tells Smart Bidding that matched users are worth more, allowing it to bid higher in those auctions without the advertiser manually setting a value. This is most powerful when the matched list contains high-lifetime-value customers and conversion-value rules assign a higher value to their conversions, enabling Target ROAS bidding to self-optimize toward them.

When first-party data work is not yet worth doing

Do not prioritize audience signal work when: (a) the account has fewer than 30 conversions in the last 30 days — Smart Bidding cannot use the signal reliably and the measurement window is too short; (b) conversion tracking is broken or counting micro-conversions that do not reflect revenue — first-party signals will amplify a bad measurement signal; (c) the remarketing list has fewer than 1,000 matched users — the list is inactive and contributes nothing; or (d) the campaign is losing more than 20% of eligible impressions to budget — fix the budget constraint first, because adding audience signals to a budget-starved campaign does not expand eligible impressions.

When to pull it, and when not to

Pull it when: the campaign has at least 30 conversions per month and clean conversion tracking; your Customer Match list has 5,000+ matched users or your remarketing list has 10,000+ active users; you can identify a segment — past converters, high-LTV customers, 30-day site visitors — that demonstrably converts at a different rate than cold traffic; and the campaign is using Target CPA or Target ROAS so Smart Bidding can act on the signal automatically.

Do not pull it when: the primary constraint is Search lost IS (budget) above 15% — the audience signal cannot expand the eligible pool, only the budget can; conversion volume is below 15 per month, making segmented analysis statistically meaningless; or the list covers fewer than 5% of the campaign’s eligible traffic, in which case the bidding impact is negligible and the section 8 bid-strategy lever (adjusting Target CPA) will move results faster.

13. Performance Max: Which Levers Exist and Which Are Illusions

Performance Max gives advertisers fewer direct controls than Search. Most inputs are signals — information fed to Google’s automation — rather than constraints. The critical skill is distinguishing levers that actually change auction eligibility or spend from inputs that guide the system without guaranteeing an outcome. Where a lever is genuinely absent, the reader needs to know what proxy exists and what it cannot do. Failure mode: treating audience signals as targeting restrictions and concluding that Performance Max is “not reaching the right people” when the system is operating as designed — the correct response is to improve asset quality and feed quality, not to add more signals.

What the platform still decides — and the reporting you do not get

Performance Max allocates budget across Search, Shopping, YouTube, Display, Discover, Gmail and Maps internally. The advertiser cannot see a reliable, real-time channel-by-channel spend split in a fully controllable form. Channel performance reporting was in rollout as of October 2026 but provides directional delivery evidence, not a controllable budget line.^[5] Search-term reporting exists but remains subject to privacy thresholds — not every query that generated an impression appears. There is no auction-level quality score, no keyword-level Ad Rank diagnostic, and no placement-level conversion report equivalent to what a standalone Display campaign provides.

The interaction with Search campaigns matters most for branded traffic. A Search keyword that exactly matches a query receives priority over Performance Max serving for that query in applicable situations, but Performance Max can still serve when the Search campaign is budget-limited, ineligible or has insufficient Ad Rank.^[5] Use brand exclusions in Performance Max plus a dedicated brand Search campaign to own branded queries cleanly — do not rely on keyword priority alone.

When to pull it, and when not to

Pull Performance Max levers when: the account has at least 50 conversions in the last 30 days with clean tracking; the campaign has been running at least 6 weeks since the last major structural change; feed disapproval rate is above 3% (fix listing groups and feed quality first); search-term insights show a high-volume irrelevant query category (add campaign-level negative); or branded queries are appearing in PMax when a Search campaign should own them (add brand exclusion).

Do not pull PMax levers when: the campaign entered learning fewer than 2 weeks ago — any structural change resets the learning clock and reading results is not yet meaningful; the conversion tracking action being optimized is a micro-conversion with low revenue correlation — improve measurement first (section 16 wins here); or the expectation is to replicate Search-campaign keyword control — Performance Max is the wrong campaign type if granular keyword management is the requirement, and the right move is to run a Search campaign alongside it with appropriate brand exclusions in PMax.

14. Demand Gen, Video and Display: Levers on the Demand Side

Demand Gen, Video and Display campaigns serve inventory where users are not actively searching for your product. Their role is to create or accelerate demand that Search and Shopping then capture. They are not interchangeable with intent-based campaigns and cannot be held to the same cost-per-conversion standard without distorting both measurement and optimization. The practical question for any account is not whether these channels work in isolation but whether the account is in a state where the next dollar belongs here rather than in Search or feed optimization. In most underperforming accounts, it does not.^[10]

Demand Gen

Demand Gen runs on YouTube (in-feed and in-stream), Discover and Gmail. It supports hard audience targeting — not just signals — making it the closest Google equivalent to paid social prospecting. Supported audience mechanisms include Customer Match, website remarketing, custom segments built from URLs or search terms, affinity, in-market and lookalike segments. Lookalike segments require a seed list meeting Google’s current minimum; verify the threshold in the current UI before planning.

The creative levers are image carousels, single-image ads, and video. Video creative length affects delivery: shorter formats (under 15 seconds) tend to reach broader inventory; longer formats enable more narrative but face higher skip rates on YouTube. There is no published Google benchmark for Demand Gen cost per conversion by industry in the available research; treat any vendor claim as directional and measure against a controlled holdout before scaling.

What Demand Gen cannot be held to: a cost per conversion comparable to non-brand Search. Users who see a Demand Gen ad and later convert via a Search click will typically be attributed to Search under last-click, or split under data-driven attribution. The real effect of Demand Gen is on the rate at which new users enter the consideration funnel — measurable via Search impression share trends over 60–90 days, branded query volume trends, and holdout geo-experiments. Without a holdout, there is no reliable way to isolate its incremental contribution. Failure mode: holding a Demand Gen campaign to a Target CPA equal to the Search campaign target, cutting it when it reports a high CPA, and incorrectly concluding the channel does not work when the actual effect was upstream of the attributed conversion.

Video

Video campaigns — primarily YouTube — run in skippable in-stream, non-skippable in-stream, in-feed video and bumper formats. The primary performance levers are audience targeting, creative, bid strategy and budget. For performance-oriented video, Maximize Conversions or Target CPA bidding is available where the account has sufficient conversion volume (approximately 30 conversions in the last 30 days is a practical minimum). For awareness and reach, CPM-based bidding with reach and frequency controls is appropriate.

Video CTR is structurally incomparable to Search CTR. A 0.5% video CTR can represent strong performance for a cold-audience skippable in-stream format; a 6.6% benchmark is a Search figure and applying it to Video is a category error.^[2] The relevant Video metrics are: view rate (views or engagements divided by impressions), earned actions, completed-view rate, and — for performance video — cost per conversion and ROAS against a holdout baseline. There is no published industry benchmark for Demand Gen or Video CPA in the available research; do not use a Display CPA of $75.51^[2] as a Video proxy because the inventory, audience and intent are materially different.

Creative is the primary lever in Video. A video ad that fails to communicate the offer in the first 5 seconds before the skip button activates wastes the impression. The first-5-seconds rule is not a Google-published guarantee of lift but an inference from skip-rate data: users who skip before 5 seconds never receive the message. No published figure quantifies the conversion impact of first-5-second optimization universally; measure it by comparing view-through rates and post-view conversion rates across creative variants.

Display

Display reaches users on third-party sites and apps in the Google Display Network. Average Display CTR is approximately 0.46%^[2] — roughly 14× lower than Search CTR — reflecting the intent gap between a user browsing content and a user searching for a product. Average Display CPA is approximately $75.51^[2] versus approximately $48.96 for Search,^[2] though both figures are cross-industry averages and individual accounts vary widely.

The creative levers in Display are responsive display ads (RDAs), which accept headlines, descriptions, images and logos and allow Google to assemble combinations, and uploaded ads for brand-controlled creative. Placement exclusions are a real lever: specific placements confirmed as low-quality in the placement report can be excluded to reduce wasted impression spend. Display lost IS (rank) averaging 45% across benchmarks^[2] indicates that quality and bid are the binding constraints more often than budget for Display campaigns. Raising bids or improving creative relevance addresses rank-driven loss; raising budget alone does not.

The honest account condition test

Upper-funnel spend is the wrong next dollar when: Search lost IS (budget) is above 10% — that budget belongs in Search first; Search lost IS (rank) is above 20% — quality and bid must be fixed before demand is amplified; conversion tracking is unreliable — upper-funnel measurement is impossible without a clean attribution base; or the average order value or lifetime value is too low to absorb a $75+ Display CPA on cold traffic. Upper-funnel spend is appropriate when Search impression share is above 70%, the account is conversion-tracking-clean, and the marginal cost of an incremental Search conversion is rising — indicating the intent-capture market is approaching saturation and new demand must be created.

When to pull it, and when not to

Pull Demand Gen, Video or Display levers when: Search impression share (non-brand) is above 70% and marginal CPA is rising; the account has 60+ days of clean conversion data and a confirmed holdout measurement plan; the business has a meaningful consideration phase where brand exposure demonstrably accelerates purchase intent; or a specific remarketing use case exists — for example, Display remarketing to 30-day cart abandoners at a cost per conversion already measured below the account CPA target.

Do not pull it when: any Search or Shopping campaign has Search lost IS (budget) above 10% — that money closes a proven gap; the conversion tracking setup is incomplete or double-counting — upper-funnel attribution will be meaningless; the account spends less than $5,000 per month total — the budget does not support parallel prospecting and intent-capture without starving both; or the client or stakeholder will measure Demand Gen against the Search CPA target — align measurement expectations before spending, or the channel will be cut before a valid read is possible.

15. Shopping and Merchant Center: Feed Levers That Move Impression Share

For a retail account running Shopping or Performance Max with a product feed, the Merchant Center feed is the single biggest lever available. Unlike Search, Shopping has no keywords the advertiser writes. Google matches products to queries by reading the feed—title, category, GTIN, price, availability, and images—alongside bid and budget signals. A feed with weak titles, missing identifiers, or high disapproval rates limits eligible impressions at the source, and no bidding change can recover impressions that the feed never made eligible. Fix the feed first; then adjust bids and budget.

Failure mode: an advertiser who raises bids or increases budget while the feed contains disapproved items, mismatched prices, or generic titles buys more auctions for the products already eligible—often the wrong ones—while the highest-potential products remain ineligible or irrelevant to the queries that would convert them.

Reading the right Merchant Center report for each action

Each feed lever has a specific report. For price competitiveness: Merchant Center → Products → Pricing, which shows the click-weighted benchmark price and your deviation from it, matched by GTIN.^[6] For disapprovals and attribute errors: Merchant Center → Products → Needs attention. For category and identifier coverage: Merchant Center → Products → All products, filtered by the relevant attribute. For best-seller demand signals to inform custom-label priority: Merchant Center → Analytics → Best sellers. For feed processing errors after a supplemental feed upload: Merchant Center → Data sources → select the feed → Diagnostics.

Shopping impression share denominators are product-eligibility denominators. When you fix disapprovals or add GTINs, the denominator may expand (more auctions become eligible) before the numerator rises, causing measured share to drop briefly while the campaign is actually doing more. Monitor absolute impressions and clicks alongside the share ratio for the first two weeks after a feed change.

When to pull it, and when not to

Pull the feed lever when: Merchant Center → Products → Needs attention shows disapprovals for any of your top-50 revenue-contributing products; or when the pricing report shows your price is more than 10% above the click-weighted benchmark on products with Shopping impression share below 50%; or when title audits reveal that fewer than 60% of titles front-load brand and product type; or when GTIN warnings appear for products that have manufacturer-assigned identifiers. These are testable states visible in Merchant Center before any Google Ads change is made.

Do not pull the feed lever first when: Merchant Center shows zero disapprovals and all major items are approved, titles are already optimized, and the pricing report shows you are at or below benchmark—in that state the feed is not the constraint, and further feed work is inert. Instead, diagnose whether the constraint is budget (Search lost IS budget above 15%) or rank (Search lost IS rank above 25%), and pull the budget or bidding lever first. Also do not make simultaneous title, category, and image changes across the full catalog in one feed upload; the resulting impression and CTR volatility cannot be attributed to any single change, and a feed-processing error that affects 100% of titles is harder to roll back than a targeted supplemental change covering one attribute.

16. Landing Page and Conversion Levers: The Cheapest Wins in the Account

Conversion rate improvement is almost always cheaper than impression share improvement. The arithmetic is straightforward: if a campaign spends $10,000 a month and converts at 3%, it produces 300 conversions at a $33 CPA (assuming a $1.10 average CPC and 9,090 clicks). Lifting the conversion rate to 4.5% on the same traffic produces 409 conversions—an extra 109—at zero additional media spend and a CPA of $24. To achieve the same 109 extra conversions through impression share expansion, the campaign would need to buy roughly 109 ÷ 0.03 = 3,633 more clicks at $1.10 each, costing $4,000 per month in added spend. The landing page fix costs development time; the impression share expansion costs $4,000 every month thereafter. This arithmetic holds at any CPC level and should be the first comparison any underperforming account makes before touching bids or budgets. There is no published Google figure for this; the arithmetic above assumes the account’s existing conversion rate and CPC hold constant on incremental traffic, which is conservative because incremental clicks tend to be more expensive as impression share rises.

Failure mode: an account optimizing to a badly configured conversion action—one that fires on page load rather than form submission, double-counts, or tracks a proxy event worth far less than an actual sale—cannot be fixed with any bidding lever. Smart Bidding will optimize faithfully to whatever it is told to optimize, and if that action is wrong, raising or lowering the target CPA, switching strategies, or expanding the keyword set will all make performance worse in the metric that actually matters to the business.

Page speed

Page speed affects both conversion rate and Google’s landing page experience rating, which feeds into Quality Score and auction-time ad quality. One published comparison found a median conversion rate of 5.1% for pages with LCP under 2.5 seconds versus 3.4% for pages exceeding 4 seconds—a relative gap of approximately 33%.^[6] That figure is study-specific and should not be applied as a universal multiplier, but the direction is consistent with every major web performance study: pages that load in under 2.5 seconds on mobile convert materially better than those that take over 4 seconds.

The levers to pull are: reduce server response time (Time to First Byte below 600 ms on mobile), eliminate render-blocking JavaScript and CSS above the fold, compress and lazy-load images below the fold, remove redirect chains (each hop adds 150–300 ms on mobile networks), and audit third-party tag payloads. Measure with Google PageSpeed Insights and Chrome UX Report real-user data, not just lab scores. A lab score of 90 on a fiber connection is irrelevant if the majority of converters are on a 4G connection in a rural area. Read the effect on conversion rate after 14–21 days and at least 200 post-change sessions; fewer sessions cannot distinguish a speed effect from normal weekly variance.

Message match

Message match is the degree to which the headline, offer, and language of the ad are continued on the landing page. An ad that says “Same-day HVAC repair, licensed technicians, Phoenix AZ” that lands on a generic home-services page breaks the contract the ad made with the user. The effect is a higher bounce rate and a lower form-completion rate. There is no single published conversion-rate multiplier for message match, but the fix is cheap: create a dedicated landing page per ad group or per major campaign theme, carry the ad’s primary headline into the page’s H1, repeat the specific offer, and show proof near the CTA. Read the result after 14 days and at least 150 sessions on the new page.

Form length and friction

Shorter forms reduce friction but can reduce lead quality if qualification questions are removed. The correct test is not “did form completions rise” but “did qualified leads rise.” A form that drops from 8 fields to 4 and doubles raw submissions may deliver leads that close at half the rate, leaving revenue flat. Test using a split: keep the full form as a control, run the shorter form as the variant, and measure downstream qualification rate and revenue per submitted lead, not just form-fill rate. Run the test for at least 21 days and 100 submissions per variant before reading the result.

Conversion tracking quality

This is the most important lever in the account and the one most commonly skipped. A campaign cannot be fixed with any bidding or budget change if it is measuring the wrong thing. Audit conversion actions before touching any other lever. The specific checks are: verify that each conversion action fires on the correct event (purchase confirmation page, not add-to-cart; form thank-you page, not form load); verify that values are passed correctly for value-based bidding; check for duplicate firing using Tag Assistant or the Google Ads conversion action table; confirm that offline conversion imports are running and that lead-to-sale data is being uploaded within the 90-day attribution window; and verify that the primary conversion action—the one Smart Bidding optimizes—is the one that represents actual business value. Path: Google Ads → Goals → Conversions → Summary, then inspect each action’s status, last conversion, and tracking method.

If enhanced conversions for web are not enabled, enable them. Google reports that enhanced conversions can improve conversion measurement accuracy where cookies are blocked or delayed, and the setup requires only a hashed email or phone number from the thank-you page. The incremental conversion data feeds Smart Bidding directly.^[11]

When to pull it, and when not to

Pull the landing page and conversion lever when: Google Ads → Goals → Conversions → Summary shows any conversion action with a status other than “Recording conversions” or shows a tag firing on a non-conversion page event (use Tag Assistant to verify); or when the account’s conversion rate is more than 30% below industry benchmarks for the campaign type and vertical; or when PageSpeed Insights reports LCP above 2.5 seconds on mobile for the primary landing URL; or when the ad’s primary headline promise does not appear anywhere above the fold on the landing page. These are testable states that do not require waiting for auction data.

Do not pull this lever first when: conversion tracking is already verified and the conversion rate is at or above vertical benchmarks—in that state, further CRO work has diminishing returns and the constraint is likely traffic volume or bid competitiveness. Also do not launch a landing page experiment while simultaneously changing bid strategy or target CPA; the two changes interact and neither effect can be cleanly isolated. Fix tracking first, then test the page, then adjust bids after the page test has read out.

17. Sequencing and Realistic Expectations: What Is Reasonably Possible in 30, 60 and 90 Days

The most common sequencing error is changing several levers at once. When bids, budgets, match types, ad copy, and landing pages change in the same week, no individual result can be attributed to any single change, and if performance worsens, the cause cannot be isolated. Change one lever at a time, wait for the result to read, then move to the next. The minimum conversion volume required before any change can be judged is 30 conversions in the measurement window at the campaign level; below that threshold, observed changes in CPA or conversion rate are more likely noise than signal. For Target CPA or Target ROAS strategy changes specifically, wait for at least one full conversion cycle (typically 14–21 days) plus the account’s average conversion lag before reading the result.

Realistic total improvement across 90 days

For an account that is not competently run at the start—meaning tracking is misconfigured, the feed has material disapprovals, landing pages are slow, and bids are manual or poorly targeted—a realistic 90-day sequence can deliver a 30–50% reduction in CPA and a 15–25 percentage-point gain in Search impression share. Those ranges assume the underlying offer and market are sound. They are planning estimates based on the arithmetic of fixing compound inefficiencies, not a published guarantee.

For an account that is already competently run—tracking verified, feed clean, bids on a working Smart Bidding strategy, landing pages at benchmark speed—the realistic 90-day improvement is much smaller: 5–15% CPA improvement and 5–10 percentage points of additional Search impression share before marginal costs make further expansion uneconomic. At that point the campaign may be near its ceiling for the current offer, budget, and geography, and the honest recommendation is to improve the offer or expand the market rather than optimize the campaign further.

Warning about simultaneous changes: every simultaneous lever change delays the read by at least one additional conversion cycle. An account that changes bid strategy, match types, landing page, and ad copy in the same week cannot attribute any result and cannot reverse a bad decision selectively. Impose a one-major-change-per-two-weeks discipline during the first 90 days.

When to stop optimizing

Stop optimizing the campaign—leave it alone—when: Search lost IS (budget) and Search lost IS (rank) are both below 10%, conversion rate is at or above the vertical benchmark, CPA is at or below the account target for at least four consecutive weeks, and the last three creative or bid changes produced less than 5% relative improvement each. In that state, further optimization work is likely to introduce variance rather than improvement, and the account should be left to run while attention moves to offer development, budget scaling, or new campaign types.

Stop advertising on this channel for this offer entirely when: the CPA at any achievable impression share is structurally above the lifetime value of the customer and cannot be closed by landing page or offer improvement; or when the search volume for the core queries is insufficient to generate the minimum 30 conversions per month needed for any Smart Bidding strategy to function; or when the offer itself is undifferentiated and no creative, bidding, or landing page change has moved conversion rate in two full test cycles. In those cases, the correct action is product or offer development, not Google Ads optimization.

18. What Changed Recently (Last 30 Days)

The following changes are drawn from verified dated sources covering 7–17 October 2026.^[3][4][5][9] No other changes in this window were substantiated by the available research and are therefore excluded.

15 October 2026 — Channel prioritization and budget-allocation controls began a progressive rollout for Search, Shopping, and Performance Max campaigns.^[5] Channel prioritization is available for Demand Gen but budget-allocation controls are not. Check campaign settings as the controls appear; availability varies by account during the rollout.

14 October 2026 — Google Ads API v26 changed language-targeting behavior for Search, AI Max for Search, and Performance Max.^[4] Campaign-level language targeting is being removed from Search and AI Max for Search campaigns, and Performance Max will no longer use campaign-level language settings for Google Search—matching will follow ad language instead. Advertisers using automated campaign-management integrations must update to a supported API version, audit language assumptions in affected campaigns, and confirm that ad assets are written in the intended matching language.

7 October 2026 — Google Ads API v22 sunsetted.^[4] Requests using v22 began failing. Any reporting, bidding, feed, or campaign-management tool still using v22 must be migrated to a supported API version immediately to avoid broken automated workflows.

12 October 2026 — YouTube responsive-ad uploads require a business name and logo when default values are not already assigned to the parent advertiser.^[3] This affects structured-data uploads and integrations, not ordinary manual campaign editing. Populate business identity assets before updating ads to avoid blocked uploads.

October 2026 — Merchant Center Shopping ads and free-listings policies consolidated into a single policy set.^[9] The change reorganizes policy documentation rather than altering eligibility rules. Update internal policy references and review any active disapprovals against the consolidated Shopping policy structure to confirm the correct remediation path.

---

References

1. [1] https://support.google.com/google-ads/answer/13738235?hl=en support.google.com
2. [2] https://support.google.com/google-ads/answer/6167118?hl=en-GB&co=ADWORDS.IsAWNCustomer=fal… support.google.com
3. [3] https://support.google.com/google-ads/answer/1752122?hl=en support.google.com
4. [4] https://support.google.com/google-ads/answer/6167118?hl=en support.google.com
5. [5] https://support.google.com/google-ads/answer/10215?hl=en support.google.com
6. [6] https://support.google.com/google-ads/answer/7640032?hl=en support.google.com
7. [7] https://support.google.com/google-ads/answer/1722122?hl=en support.google.com
8. [8] https://support.google.com/google-ads/answer/6167130?hl=en support.google.com
9. [9] https://support.google.com/google-ads/answer/12732144?hl=en support.google.com
10. [10] https://support.google.com/google-ads/answer/6167118?visit_id=638628741156820547-312902798… support.google.com
11. [11] https://support.google.com/google-ads/answer/2497703?hl=en support.google.com