A 100% Optimization Score is not the goal. The score reflects how many of the recommendations Google surfaces you have accepted, not how well the account performs. Which to apply, which to dismiss, and why auto-apply stays off.
Last updated: November 4, 2026
In This Guide
- Executive Summary
- Benchmarks & Numbers at a Glance
- What It Measures
- Is 100% the Goal
- Recommendation Types
- Recommendations to Apply
- Recommendations to Dismiss
- Risks of Auto-Apply
- Score in Audits
- Broad Match & Budget Recs
- Common Mistakes to Avoid
- What Changed Recently
- References
1. Executive Summary
Google Ads Optimization Score is a real-time, 0–100% estimate of how well an account or campaign is configured relative to Google’s recommendation framework. As of November 2026, five principles should govern every practitioner’s approach to the score and its recommendations.
- The score is a prioritization tool, not a KPI. It measures recommendation adoption, not profitability, conversion quality, or return on ad spend. A 100% score can coexist with poor economics; a 70% score can coexist with excellent results.[1]
- Apply recommendations selectively against business goals. Every recommendation must be evaluated against the account’s real objective—qualified leads, revenue, profit margin—before it is accepted or dismissed. Applying a recommendation solely to raise the score number is an error of process.[2]
- Dismissal is a valid and documented control action. Google explicitly accounts for dismissed recommendations in score history. A deliberate, logged dismissal is strategically superior to a blind acceptance that damages lead quality, efficiency, or brand control.[1][2]
- Auto-apply must be governed, not assumed safe. As of January 26, 2026, Google removed “Add responsive search ads” from auto-apply eligibility. This change confirms that the auto-apply menu evolves over time and must be audited regularly, not configured once and forgotten.[2]
- Measure outcomes in business terms after every material change. Score uplift is an estimate, not a guarantee. Post-implementation review must track qualified conversions, CPA, ROAS, search-term quality, and profit—not the score percentage itself.[1][3]
2. Benchmarks and Numbers at a Glance
| Metric | Typical range or threshold | Applies when | Source |
|---|---|---|---|
| Optimization Score range | 0–100% | All eligible active campaigns; vendor-defined scale | [1] |
| Individual recommendation score uplift range | <0.1 to 100 percentage points | Per recommendation; vendor claim, sample size not stated | [1] |
| Median conversion increase per 10-point account score increase | 14% | Vendor claim; sample size not stated; treat as directional, not guaranteed | [4] |
| Additional conversions: exact match → broad match with Target CPA | +35% average | Vendor claim; conversion-based campaigns with stable Target CPA bidding; sample size not stated | [3] |
| Additional conversions: phrase match → broad match with Target CPA | ~+25% average | Vendor claim; Google internal data, 2020; account count not stated | [3] |
| Additional conversion value: phrase match → broad match with Target ROAS | ~+12% average | Vendor claim; Google internal data, 2020; account count not stated | [3] |
| Conversion-value increase: Target CPA → Target ROAS at similar ROAS | +14% average | Vendor claim; sample size not stated; applies to campaigns with transaction-level value data | [3] |
| Broad match share of total spend in one US six-month test ($40,000 total) | 53% of spend | Single US test; not a population average; treat as one data point only | [5] |
| Broad match share of total conversions in the same US test | 81% of conversions | Single US test; same caveat as above; efficiency ratio requires independent verification | [5] |
| Broad match conversion count improvement (days 1–30 vs. days 31–90 in one test) | 94 conversions → 156 conversions (+66%) | Single campaign; market not specified; CPA dropped from €118 to €84 in same period; EUR-denominated | [5] |
| Auto-apply RSA creation: eligible from | No longer eligible as of January 26, 2026 | All accounts globally; Google removed this recommendation type from the auto-apply menu | [2] |
| Weighted portfolio score formula | ∑(score × weight) ÷ ∑(weight) | Multi-account manager (MCC) rollup; use Google’s score weight field, not a simple average | [5] |
Note: Where sources conflict on broad-match conversion lift figures (Google vendor claims of +25–35% versus single-test study data), the most conservative position is to treat all figures as directional and to require an in-account experiment before acting on projected uplift.
3. What Optimization Score Measures
Optimization Score is a real-time estimate of how closely an account or campaign aligns with Google’s current best-practice recommendations for its stated objective. It is recalculated continuously as campaign statistics, settings, available recommendations, and recommendation history change.[1]
The score reflects five inputs simultaneously:
- Current campaign statistics and settings — budget pacing, bid strategy configuration, keyword coverage, asset completeness, and audience setup.
- Available recommendations and their estimated impact — the set of recommendations Google has generated for the account and the score uplift assigned to each.
- Recent recommendation history — whether recommendations have been applied, dismissed, or left unactioned. Both applied and dismissed recommendations factor into the calculation.[1]
- Campaign objective — a Maximize Conversions campaign is evaluated differently from a Maximize Conversion Value or Target ROAS campaign.
- Google’s performance models and ecosystem trends — Google’s internal models, historical account data, and recognized best practices, which evolve over time.[1][3]
The score is available at the customer (account) level and at the campaign level for eligible active campaign types: Search, Display, Discovery, Video Action, App, Performance Max, and Shopping.[2]
What Optimization Score Is Not
Optimization Score is explicitly not Quality Score. Quality Score is a 1–10 keyword-level diagnostic covering expected click-through rate, ad relevance, and landing-page experience. The two metrics are independent: Optimization Score does not influence Quality Score, and Quality Score does not determine Optimization Score.[2] Optimization Score is also not an auction signal, a ranking factor, or a measure of profitability, conversion quality, or ROAS.
The score uplift assigned to each recommendation is an estimate of how much the score could move if that recommendation were applied—it is not a forecast of incremental revenue or guaranteed conversion improvement.[1]
Worked example (illustrative)
Reading What a Score Actually Reflects
- Setup: A Chicago law firm running a Search campaign for personal injury keywords, spending $12,000 per month, with Target CPA bidding set to $280 per qualified call. Current Optimization Score is 61%.
- Numbers: The Recommendations page shows three open items: (1) “Switch to broad match” — estimated uplift of +18 points; (2) “Add responsive search ad assets” — estimated uplift of +12 points; (3) “Fix conversion tag on /thank-you page” — estimated uplift of +9 points. Total open uplift: 18 + 12 + 9 = 39 points. Applied and dismissed history accounts for the remaining score gap to 100%.
- Decision: The practitioner prioritizes item (3) — the conversion tag fix — and applies it immediately. Items (1) and (2) are logged for manual review against lead-quality data before any action is taken.
- Why: Google’s documentation states the score reflects available recommendation impact; the tag fix affects measurement integrity, which underlies every other automated recommendation, so it must be resolved before bid-strategy or match-type changes are evaluated.[1]
4. Is 100% the Goal?
No — not automatically and not in every account context. Google defines 100% as the point at which an account is estimated to perform at its full potential within Google’s recommendations framework.[1] That definition contains an important qualifier: the framework is built on Google’s models and best practices, which may not perfectly align with every advertiser’s business constraints, compliance requirements, or profit objectives.
A score of 100% does not guarantee:
- Maximum profit or return on ad spend
- Minimum cost per acquisition
- Highest possible conversion volume at acceptable quality
- That every accepted recommendation is suitable for the account’s market, offer, or audience
- Sound conversion tracking or accurate attribution
Conversely, a score below 100% is not evidence of a poorly managed account. An account may deliberately maintain a lower score because it has dismissed recommendations that would expand match types beyond profitable thresholds, increase budgets beyond approved media plans, alter bid strategies during stable-performance periods, or publish ad copy that has not received legal or compliance review.
The practical target for most US accounts in 2026 is “high score with justified exceptions”: apply every recommendation that improves business outcomes, dismiss those that do not, and document the reason for every dismissal so that future auditors understand that the gap from 100% is intentional.[1][4]
When a Sub-100% Score Is Strategically Correct
- The account runs in a regulated vertical — healthcare, legal, financial services — where copy, targeting, or audience expansion requires compliance review before any change.
- The campaign uses exact-match keywords deliberately to control query intent in a high-CPA, low-volume lead environment.
- Budget recommendations would exceed an approved media plan or move funds from a higher-margin campaign.
- Bid-strategy recommendations would restart Smart Bidding learning during a critical sales period, such as Q4 e-commerce or fiscal year-end.
- Conversion data is insufficient or unreliable, making automated bidding changes premature.
Worked example (illustrative)
Deliberate Score Gap During Q4 Peak Season
- Setup: A Dallas e-commerce account selling consumer electronics, spending $95,000 per month in November and December 2026, running Target ROAS at 500% across three Search campaigns. Current Optimization Score is 74%.
- Numbers: The largest open recommendation is “Switch to Maximize Conversion Value (remove ROAS target)” — estimated score uplift: +19 points. Accepting this would bring the score to approximately 93%. The account averaged 1,840 conversions in October 2026 at a blended ROAS of 512%. Removing the ROAS floor during peak season carries a risk of spending the $95,000 monthly budget at a ROAS of, for example, 380% — a revenue-per-dollar drop of approximately 26% ((512 − 380) ÷ 512).
- Decision: Dismiss the “Maximize Conversion Value” recommendation with a logged reason: “ROAS floor protects margin during November–December peak; score gap of 19 points is acceptable.” Score remains at 74% by choice.
- Why: Google’s documentation confirms that dismissing a recommendation is a valid control action that changes the score, and that 100% does not mean maximum profit or minimum CPA.[1]
Managed for you
Everything on this page, handled every month
You are reading how to do it properly. Omologist does it, inside a written mandate, for 15% of your budget.
✓ Conversion tracking built and tested before any ad runs
✓ Search campaigns built to what your budget can actually feed
✓ Changes made inside a written mandate, not whenever
✓ A report on the 1st that says what was done and why
15% of budget plus GST · minimum A$330 incl GST · accounts under about A$10,000 a month
5. Recommendation Types Explained
Google groups its recommendations into six broad categories. Each category contains multiple specific recommendation types, and not every type appears in every account. The available set changes as Google updates its best practices, as confirmed by the January 2026 removal of “Add responsive search ads” from auto-apply eligibility.[2]
| Category | Examples of included recommendation types | Primary business impact | Auto-apply risk level |
|---|---|---|---|
| Bidding and budgets | Switch bid strategy, adjust target CPA/ROAS, increase budget, fix budget-limited campaigns, enable enhanced CPC | High — directly changes spend, bid behavior, and campaign delivery | High — keep manual |
| Keywords and targeting | Add broad match, add new keywords, add audience segments, expand location targeting, remove redundant keywords | High — alters who sees ads and at what intent level | High — keep manual |
| Ads and assets | Improve ad strength, add asset types (sitelinks, callouts, images), add new ad variations; note: RSA auto-creation removed January 26, 2026 | Medium — affects ad quality and CTR; copy must be reviewed for accuracy and compliance | Medium — review copy before accepting |
| Measurement | Fix conversion tags, enable auto-tagging, set up conversion value rules, import offline conversions | High — underlies all automated bidding; fix these before any other change | Low risk to apply; high risk to ignore |
| Campaign maintenance | Fix disapproved ads, resolve Merchant Center feed errors, repair broken URLs, resolve policy issues | Medium-high — prevents delivery loss and policy suspension | Low — safe to apply after verifying the fix is factually correct |
| New campaign features and setup | Try Performance Max, enable new campaign subtypes, adopt new bidding features | Variable — requires full strategic assessment before adoption | High — do not auto-apply |
Score uplift per recommendation ranges from less than 0.1 percentage points to a theoretical 100 percentage points, depending on the account’s current configuration and the estimated impact of the change.[1] A single high-uplift recommendation — such as a bid-strategy switch — can represent a larger share of the gap to 100% than ten smaller maintenance items combined. This weighting creates an incentive to accept high-impact recommendations quickly; resist that incentive and evaluate the substance of each change independently of its uplift figure.
Worked example (illustrative)
Categorizing a Full Recommendation Queue for Triage
- Setup: A Seattle B2B SaaS account spending $28,000 per month across four Search campaigns. Optimization Score is 58% at the account level. The Recommendations page shows seven open items with a combined estimated uplift of 42 points.
- Numbers: Measurement: “Fix conversion tag — contact form” (+8 pts). Campaign maintenance: “Fix two disapproved ads” (+4 pts). Ads and assets: “Add 3 sitelink assets to Campaign 3” (+6 pts). Keywords and targeting: “Add broad match to Campaign 1” (+12 pts). Bidding: “Switch Campaign 2 from Manual CPC to Target CPA $420” (+7 pts). Bidding: “Increase Campaign 4 budget from $4,200/month to $6,100/month” (+3 pts). New features: “Create a Performance Max campaign” (+2 pts). Total: 8+4+6+12+7+3+2 = 42 pts.
- Decision: Apply the tag fix and disapproved-ad fixes immediately (12 pts combined). Schedule the sitelink addition for copy review within 5 business days. Place broad match, Target CPA switch, budget increase, and Performance Max in a “review required” queue with a 10-business-day deadline and a defined success metric (CPA ≤ $420 per Marketing Qualified Lead) for each.
- Why: Google’s documentation categorizes measurement and maintenance fixes as the foundation for all other automated recommendations; applying them first ensures that any subsequent bid or targeting change operates on accurate data.[1][2]
6. Recommendations to Apply
The following recommendation types are generally safe to apply, subject to the verification steps listed below. “Safe” means the change is unlikely to materially increase spend, expand targeting beyond intended parameters, or alter customer-facing messaging without review — but it is never a substitute for checking the specific details of each recommendation in the account.
Measurement and Tracking Repairs
These should always be the first priority. Automated bidding, Smart Bidding targets, and the accuracy of every other recommendation depend on reliable conversion data. Apply conversion-tag fixes, auto-tagging enablement, duplicate-conversion resolutions, and offline-conversion import setup before evaluating any bidding or keyword recommendation.[2]
Policy and Delivery Repairs
Disapproved ads, suspended feeds, broken destination URLs, and Merchant Center errors prevent delivery entirely. Apply these after confirming the underlying issue — for example, verify that a disapproved ad’s copy genuinely complies with policy before resubmitting rather than simply clicking apply.
Ad and Asset Completeness (with Copy Review)
Adding sitelinks, callout assets, structured snippets, image assets, and lead form extensions typically improves auction eligibility and CTR without changing bidding or targeting. Apply these after a human reviews the proposed copy, URL, and landing-page alignment for accuracy, brand compliance, pricing accuracy, and legal wording.[2]
Genuine Keyword Housekeeping
Removing confirmed duplicate keywords, resolving overlapping ad groups where the overlap is unintentional, and fixing clear structural issues are low-risk when the change history and search-term report confirm that the removed keyword is genuinely redundant and not part of a deliberate strategy.
Before Applying Any Recommendation: Five Verification Steps
- Confirm the campaign’s stated conversion action is correct and tracking accurately.
- Check whether the change will restart bid-strategy learning — if so, assess the timing against upcoming promotions or seasonal peaks.
- Read the full recommendation detail panel, not just the headline and uplift figure.
- Verify the change does not conflict with an active experiment, budget ceiling, or approved media plan.
- Define the success metric and review window before applying, so post-implementation evaluation is objective.
Worked example (illustrative)
Applying a Conversion Tag Fix Before Evaluating a Bid-Strategy Switch
- Setup: A Miami dental practice running a Search campaign at $5,500 per month with Target CPA set to $95 per appointment booking. Optimization Score is 54%. Two open recommendations: “Fix conversion tag on /booking-confirmation” (+11 pts) and “Switch to Maximize Conversions (remove CPA target)” (+22 pts).
- Numbers: The conversion tag fix is estimated to restore tracking on a page that generates approximately 60% of the campaign’s conversions. Without that fix, the reported CPA is $142 (based on 39 tracked conversions in October 2026 at $5,538 spend), but the true CPA — including untracked conversions — is estimated at approximately $89 ($5,538 ÷ 62 estimated actual conversions). Applying the bid-strategy switch at a $142 reported CPA would cause Smart Bidding to over-bid by approximately 50%.
- Decision: Apply the conversion tag fix immediately. Hold the “Maximize Conversions” recommendation for a minimum 30-day re-evaluation after the tag is confirmed working, then reassess whether Target CPA at $95 or Maximize Conversions is more appropriate based on corrected data.
- Why: Google’s documentation confirms that automated bidding recommendations depend on the reliability of conversion data; a tag error that suppresses 35% of conversions (62 − 39 = 23 untracked, or 37% of true volume) will cause any bid strategy to optimize against incorrect signals.[1][2]
Worked example (illustrative)
Adding Sitelink Assets After Copy Review
- Setup: A Portland residential solar installer spending $18,000 per month on a Search campaign. The Recommendations page suggests adding four sitelink assets — estimated score uplift of +7 points. Google’s auto-generated sitelink suggestions include “Free Consultation,” “Solar Tax Credits,” “Get a Quote,” and “See Our Projects.”
- Numbers: The IRS federal solar tax credit (Section 25D) is 30% through 2032 for residential installations. The “Solar Tax Credits” sitelink is factually accurate as of Q4 2026. All four proposed URLs resolve correctly. The landing page for “Free Consultation” matches the offer — no fee is charged for the initial site visit. All four sitelinks pass a manual compliance review in 15 minutes.
- Decision: Apply all four sitelinks as proposed. No copy edits required. Document the review date (November 2026) and note that the tax credit percentage should be re-verified if legislation changes.
- Why: Asset-completeness recommendations that do not expand targeting or change bidding are classified as low-risk; the only required gate is a human copy review for factual accuracy, compliance, and brand alignment.[2]
Not ready to change anything today?
We post the sharpest bits of these guides — what a change is worth, how long it takes to show up, and when to leave it alone. Short enough to read on your phone.
Follow on LinkedIn Follow on Instagram7. Recommendations to Dismiss
Dismissing a recommendation is a normal, documented, and strategically valid action. Google explicitly states that both applying and dismissing recommendations affect the Optimization Score calculation history.[1] A dismissed recommendation that would have damaged lead quality, exceeded budget, or introduced compliance risk is worth more to the business than the score points that would have been gained by accepting it.
Categories That Most Often Warrant Dismissal
Bid-strategy changes on stable, profitable campaigns. If a campaign is meeting its CPA or ROAS target with the current strategy, a recommendation to switch strategies introduces a learning period — typically 2–4 weeks of volatile performance — with no guaranteed improvement. Dismiss unless there is a specific, data-supported reason to change.[2]
Target loosening recommendations. Recommendations to lower a Target CPA or raise a Target ROAS are designed to increase delivery volume, but they may do so at the cost of efficiency. Dismiss if the current target is correctly calibrated to margin.
Budget increases that exceed the approved media plan or push marginal CPA above the break-even threshold. Being “budget-limited” does not prove that additional spend is profitable. Dismiss budget-increase recommendations when the marginal conversion cost at the proposed budget is above the acceptable CPA, or when the additional spend would come from a lower-priority reallocation.[2]
Audience or location expansion in regulated or quality-sensitive verticals. Healthcare, legal, financial services, and high-ticket B2B accounts often require precise geographic and demographic targeting. Expansion recommendations that broaden eligibility without a quality gate should be dismissed.
Ad copy generated without compliance review. Any recommendation that proposes publishing new copy — particularly in regulated industries — must be reviewed by a human before application. If a workflow does not allow that review before the recommendation is acted on, dismiss and recreate the ad manually.
Recommendations that would undo deliberate strategy. If exact-match keywords exist intentionally to protect query intent, if negative keywords have been built to exclude low-quality segments, or if a geographic restriction reflects legal or operational limits, dismiss any recommendation that would reverse those decisions.
How to Dismiss Properly
- Use Google’s dismiss option for individual recommendations or for all recommendations of a specific type.
- Log the reason in an internal record — for example: “Target CPA $95 meets margin requirement; loosening to $130 would reduce profitability below acceptable threshold.”
- Set a review cadence — for example, revisit dismissed broad-match recommendations quarterly as conversion volume and tracking quality evolve.
- Do not dismiss permanently without a review date; account conditions change and a recommendation that is inappropriate today may be valid in six months.
Worked example (illustrative)
Dismissing a Budget-Increase Recommendation Below Break-Even CPA
- Setup: A Phoenix landscaping company running a Search campaign at $4,000 per month. The campaign is flagged as “limited by budget.” Google recommends increasing the budget to $6,200 per month (+$2,200/month) and projects 34 additional conversions per month — estimated score uplift: +5 points.
- Numbers: Current performance: 48 conversions at $4,000 spend = $83.33 CPA. Business’s acceptable CPA ceiling: $110. Projected incremental conversions at the additional $2,200: 34 conversions = $64.71 incremental CPA ($2,200 ÷ 34). On its face, the marginal CPA looks acceptable. However, the account’s conversion action is “form submission,” and historical close rate on form submissions is 22%. True cost per closed job = $83.33 ÷ 0.22 = $378.78 at current spend. At the increased budget, if incremental form-submission quality is consistent, true cost per closed job on incremental volume = $64.71 ÷ 0.22 = $294.14. Average job value is $310. Incremental margin per job = $310 − $294.14 = $15.86. That is a 5.1% margin on incremental spend — below the company’s 15% minimum threshold.
- Decision: Dismiss the budget-increase recommendation. Log reason: “Incremental margin per closed job at $15.86 (5.1%) is below the 15% minimum. Revisit when close rate improves or average job value increases.” Score remains lower by 5 points; that is acceptable.
- Why: Google’s documentation confirms that being budget-limited does not by itself prove additional spend is profitable; marginal CPA/ROAS must be evaluated against business economics before accepting budget recommendations.[2]
8. The Risks of Auto-Apply
Auto-apply allows Google to implement selected recommendation types on a recurring, automatic basis without per-instance human approval. When properly scoped, it can reduce maintenance overhead. When misconfigured or left unmonitored, it can materially alter spend, targeting, bid behavior, and customer-facing messaging without an explicit human decision.[2][3]
The January 2026 RSA Change as a Governance Lesson
Effective January 26, 2026, Google removed “Add responsive search ads” from the auto-apply eligibility menu. Accounts that had previously enabled this category found that the setting was no longer operative — and that the auto-apply menu had changed without a per-account notification.[2] This confirms a core governance risk: the set of recommendation types eligible for auto-apply changes over time, and an auto-apply configuration that was reviewed and approved at one point in time may become stale, inapplicable, or — in the opposite scenario — may include new types that were not present when the configuration was last reviewed.
Categories to Keep Manual
The following recommendation categories should not be auto-applied in the vast majority of US accounts. The rationale for each is that the change either directly increases spend, expands audience reach, alters bid behavior, or publishes unapproved copy.[2][3]
- All bid-strategy changes (switch to Target CPA, Target ROAS, Maximize Conversions, Maximize Conversion Value)
- Target CPA or Target ROAS adjustments (up or down)
- Campaign budget increases or reallocations
- Broad-match expansion and all keyword match-type changes
- New keyword additions
- Removal of negative keywords
- Audience expansion (similar segments, broader in-market targeting)
- Location expansion
- Search partner and Display Network opt-in
- New campaign creation (Performance Max, new campaign types)
- Ad copy generation (since January 26, 2026, RSAs are no longer auto-applied; any other ad-copy generation should also remain manual)
Categories Where Narrow Auto-Apply May Be Acceptable
Even within these categories, verify the specific recommendation type before enabling. The rule is: if the change could increase spend, broaden reach, or publish copy, it must remain manual.[2]
- Enabling auto-tagging (measurement; no spend or targeting impact)
- Fixing broken ad destination URLs (maintenance; does not change copy or targeting)
- Resolving clearly defined Merchant Center feed errors (maintenance; apply only if the error type is well understood)
Governance Requirements for Any Auto-Apply Configuration
- Review the Auto-apply History tab at least once per month. It shows each type enabled, how recently it was applied, and the last application date.[3]
- Cross-reference the Change History report against performance timelines whenever a metric moves unexpectedly. Auto-applied changes appear in Change History.[3]
- Maintain a written record of which categories are enabled, who approved the configuration, the date of approval, and the rationale.
- Re-audit the auto-apply menu after any account restructure, conversion-action change, bidding-strategy change, or new campaign launch.
- Confirm that auto-apply does not control budgets — Google states that it does not automatically increase the account budget — but verify that the existing budget does not become a performance constraint as a result of other auto-applied changes.[2]
Worked example (illustrative)
Detecting an Unintended Broad-Match Auto-Application via Change History
- Setup: A Minneapolis staffing agency running four Search campaigns at a combined $22,000 per month. Auto-apply was configured 14 months ago and not reviewed since. In the first week of October 2026, cost-per-conversion rises from $74 to $118 (a 59% increase) while conversion volume drops from 297 conversions in September to an on-pace 189 for October.
- Numbers: The practitioner opens Change History filtered to “Automatic” source for the period September 25 – October 7, 2026. Result: on September 27, 2026, auto-apply switched 214 exact-match keywords in Campaign 2 to broad match. Campaign 2 spend increased from $5,200 in September to an on-pace $8,900 for October (+$3,700/month). Search-term report for October 1–10 shows 38% of impressions on queries with no demonstrable staffing intent. Estimated wasted spend: 38% × $8,900 = $3,382 in the first month alone.
- Decision: Immediately disable the broad-match auto-apply setting. Revert Campaign 2 keywords to exact match via Change History rollback. Set a calendar reminder to audit all auto-apply settings on the first Monday of each month.
- Why: Google’s documentation confirms that auto-applied recommendation changes appear in Change History and can be audited; this is the primary mechanism for detecting and reversing unintended automated changes.[3]
“So what do I actually have to do?”
Tell us the business and give us the account. We build the conversion tracking and prove it works with a test conversion before any ad runs, then launch the campaigns your budget can feed.
After that, changes happen inside a written mandate, so you know in advance what gets changed without asking and what never gets touched. A report on the 1st, a budget email around the 23rd, and nothing else unless it needs your yes or no.
How it works9. Using Score in Audits and Across Accounts
Optimization Score is most useful in an audit context as a triage layer — a structured starting point for identifying where recommendations are concentrated — rather than as a headline performance metric. A score of 65% is not evidence that the account is performing badly; it is evidence that 35% of the available recommended improvements have not been applied or dismissed with a documented reason.[3]
Single-Account Audit Protocol
Use this sequence to prevent the score from distorting the audit’s conclusions:
- Step 1 — Record the baseline: Account score, campaign-level scores, recommendation categories, estimated uplift per recommendation, current bid strategies, budgets, conversion actions, and attribution model. Date-stamp the record.
- Step 2 — Validate measurement first: Confirm that all primary conversion actions are firing correctly, that values are populated where the campaign uses value-based bidding, and that offline conversion imports are current. Any measurement gap invalidates the recommendations that depend on it.
- Step 3 — Classify recommendations: Sort every open recommendation into one of four buckets — Apply now, Test before applying, Review later, or Dismiss — based on business objective, conversion data quality, margin requirements, and strategic constraints.
- Step 4 — Separate score from results: Compare the account’s actual CPA, ROAS, qualified conversion volume, and profit against the period before and after major recommendation applications. Score uplift is not equivalent to these outcomes.[3]
- Step 5 — Audit auto-apply and Change History: Review which recommendation types are enabled for auto-apply and cross-reference recent automatic changes against performance timelines.
- Step 6 — Document dismissals: Every dismissed recommendation should have a written reason in the audit record so that future practitioners understand the account’s deliberate exceptions.
Multi-Account (MCC) Score Aggregation
When reporting Optimization Score across a portfolio of accounts, use Google’s weighted average formula rather than a simple arithmetic mean of account scores. The correct calculation is:[5]
Portfolio score = ∑(account score × score weight) ÷ ∑(score weight)
Google’s API provides a score weight field for each customer account. Using a simple average of account percentages gives disproportionate weight to small accounts and understates the influence of large-spend accounts on overall portfolio health.
Reporting Framework for Agency and In-House Teams
| Measure | Reporting use | Frequency |
|---|---|---|
| Account Optimization Score | Overall recommendation coverage; triage starting point | Weekly review; monthly report |
| Campaign score distribution | Identify campaigns requiring priority attention | Weekly |
| Weighted portfolio score | Comparable multi-account rollup using Google’s score weights | Monthly |
| Recommendation acceptance rate by category | Governance discipline; track which categories are acted on vs. dismissed | Monthly |
| Auto-apply coverage | Which categories can change accounts without per-instance approval | Monthly audit |
| Performance KPIs | CPA, ROAS, qualified conversions, revenue, profit; the actual outcomes | Weekly; real-time alerts on material changes |
| Change History exceptions | Detect unintended or unauthorized automated changes | Weekly |
Worked example (illustrative)
Calculating a Weighted Portfolio Score Across Three Accounts
- Setup: An agency managing three Google Ads accounts for US clients. At the end of October 2026, the account scores and score weights (as returned by the Google Ads API) are as follows: Account A (national retailer): score 82%, weight 4.2; Account B (regional services): score 67%, weight 1.1; Account C (startup e-commerce): score 91%, weight 0.4.
- Numbers: Weighted numerator = (82 × 4.2) + (67 × 1.1) + (91 × 0.4) = 344.4 + 73.7 + 36.4 = 454.5. Sum of weights = 4.2 + 1.1 + 0.4 = 5.7. Weighted portfolio score = 454.5 ÷ 5.7 = 79.7%. Simple (unweighted) average = (82 + 67 + 91) ÷ 3 = 80.0%. In this case the difference is 0.3 points, but if Account B had a score of 40% and a weight of 3.9, the difference would be material: weighted = (82×4.2 + 40×3.9 + 91×0.4) ÷ 8.5 = (344.4 + 156.0 + 36.4) ÷ 8.5 = 536.8 ÷ 8.5 = 63.2% vs. simple average of 71.0%.
- Decision: Report the weighted portfolio score of 79.7% in November 2026 monthly reports. Flag Account B (score 67%) for a recommendation review in the first two weeks of November 2026.
- Why: Google’s API documentation specifies that the weighted formula — not a simple average — produces a comparable multi-account rollup; using a simple average over-represents small accounts.[5]
10. Broad Match and Budget Recommendations
Broad match and budget-increase recommendations consistently represent two of the highest score-uplift items in an account and are also two of the highest-risk categories for unreviewed application. Both require a disciplined evaluation against actual business economics before any action is taken.[2]
Broad Match: When to Test and When to Hold
Google’s broad-match recommendation is typically surfaced for campaigns using automated bidding — Target CPA, Target ROAS, Maximize Conversions, or Maximize Conversion Value — where Smart Bidding can theoretically manage query expansion. Google’s vendor claims cite +35% average conversions when switching from exact match to broad match with Target CPA, and +25% average when switching from phrase match, though both figures are vendor claims without published sample sizes.[3]
These averages do not apply uniformly. In practice, broad match performance is highly sensitive to the quality of conversion tracking, the adequacy of negative-keyword management, and the tolerance for query variation. A single US test involving $40,000 in spend over six months showed broad match capturing 53% of spend but 81% of conversions — a favorable efficiency ratio in that specific context — but this represents one data point, not a population benchmark.[5]
Test broad match only when all of the following are true:
- Conversion tracking is confirmed accurate and stable for at least 60 days.
- The campaign has a minimum of 30 conversions per month (preferably 50+) to give Smart Bidding a sufficient signal base.
- Automated bidding is calibrated to the account’s real business outcome — qualified leads or revenue — not a proxy action like page views.
- A negative-keyword list is active and maintained on at least a weekly review cadence.
- Geographic, brand, and audience controls are already in place.
- The test is isolated to one campaign, with a control group or comparison period defined in advance.
Delay or dismiss the broad-match recommendation when:
- Conversion volume is below 30 per month — Smart Bidding cannot learn effectively at lower volumes.
- The campaign optimizes for a low-quality proxy conversion action.
- Search intent is highly specialized, regulated, or brand-sensitive (e.g., pharmaceutical, legal, financial).
- There is no process for reviewing new search terms at least weekly.
- The recommendation would materially increase spend without a pre-defined acceptable CPA or ROAS threshold for the incremental volume.
Budget Recommendations: Five-Point Evaluation Checklist
Before accepting any budget-increase recommendation, confirm all five of the following:[2]
- The campaign is genuinely losing valuable impressions due to budget — not due to low Quality Score, low bids, or policy issues.
- The campaign’s marginal CPA or ROAS at current spend is within the business’s acceptable range.
- There is sufficient search demand and conversion capacity to absorb the additional spend without an efficiency collapse.
- Increasing this campaign’s budget does not redirect funds from a higher-margin campaign in the same account.
- Seasonality, inventory levels, sales team capacity, and cash-flow limits are consistent with an increased spend commitment.
Worked example (illustrative)
Testing Broad Match in an Isolated Campaign Before Portfolio Rollout
- Setup: A San Antonio home security company running three Search campaigns at a combined $14,000 per month. Campaign 1 (branded) spends $2,000/month; Campaign 2 (installation services, exact match) spends $7,000/month at a CPA of $68 per qualified installation appointment; Campaign 3 (monitoring plans, phrase match) spends $5,000/month at a CPA of $91. Google recommends switching Campaign 2 keywords to broad match — estimated score uplift: +14 points, projected +35% conversion volume.[3]
- Numbers: Campaign 2 currently delivers 102 appointments per month ($7,000 ÷ $68.63). A +35% lift would project 138 appointments. At the current $7,000 budget with higher volume, the implied CPA would need to drop to $7,000 ÷ 138 = $50.72 — which requires Smart Bidding to find cheaper conversions via broader queries. The business’s acceptable CPA ceiling is $85. To test safely: isolate Campaign 2, hold the $7,000 budget fixed, enable broad match on November 3, 2026, and evaluate over a 45-day window (November 3 – December 17, 2026) using CPA ≤ $85 and qualified appointment rate ≥ 80% as the two success criteria. Do not change bids, budget, or landing pages during the test.
- Decision: Run the isolated broad-match test on Campaign 2 only. Keep Campaigns 1 and 3 unchanged as a control reference. Review results on December 18, 2026. Roll out to Campaign 3 only if both success criteria are met.
- Why: Google’s documentation recommends testing material changes on a controlled subset before portfolio rollout; changing one major variable at a time (match type only, with budget and bids held fixed) is essential for attributing any performance change to the match-type switch.[2]
Worked example (illustrative)
Evaluating a Budget-Increase Recommendation Against Marginal CPA
- Setup: A Nashville event-rental company running a Search campaign at $3,500 per month, flagged as budget-limited. Google recommends increasing the budget to $5,400 per month (+$1,900/month) and projects 22 additional conversions per month — estimated score uplift: +6 points.
- Numbers: Current performance in October 2026: 41 conversions at $3,500 = $85.37 CPA. Business’s CPA ceiling: $100 per booked event. Projected incremental CPA = $1,900 ÷ 22 = $86.36. That is within the $100 ceiling. Five-point checklist: (1) Budget-limited? Yes — campaign serves 100% of daily budget by 2:00 PM on 73% of days in October. (2) Marginal CPA within range? Yes — $86.36 < $100. (3) Sufficient search demand? Search Impression Share lost to budget = 34%; 34% lost IS × current impression volume = approximately 1,200 additional eligible impressions per month. (4) No higher-priority campaign needs the budget? Confirmed — Campaigns 2 and 3 are not budget-limited. (5) Seasonality and capacity consistent? November–December is peak season for event rentals; the sales team can handle additional bookings. All five criteria pass.
- Decision: Accept the budget recommendation. Increase Campaign 1 budget from $3,500 to $5,400 effective November 1, 2026. Review CPA and conversion quality after 30 days (December 1, 2026). If CPA exceeds $100 or booking quality declines, reduce budget to $4,200 and re-evaluate.
- Why: All five evaluation criteria in Google’s recommended budget-review process are satisfied; the marginal CPA of $86.36 is within the $100 ceiling and demand headroom exists as evidenced by 34% budget-limited impression share lost.[2]
11. Common Mistakes to Avoid
The following errors appear repeatedly in accounts that use Optimization Score as a primary management tool rather than a decision-support instrument. Each is actionable and avoidable.
Optimizing for the Score Instead of the Business
Accepting every recommendation to reach 100% is the most common misuse of the score. The score measures recommendation adoption, not profitability. An account that accepts all recommendations and reaches 100% may simultaneously have a rising CPA, declining lead quality, expanding wasted spend, or misaligned bid targets. The score and business outcomes must be tracked separately.[1]
Applying Bid-Strategy Changes Without Verifying Conversion Data
Smart Bidding recommendations — switch to Target CPA, Target ROAS, Maximize Conversions, or Maximize Conversion Value — are only as reliable as the conversion data they optimize against. Applying a bid-strategy switch when conversion tracking has a tag error, a duplicate-counting problem, or a proxy action instead of a real business outcome will cause the bidding algorithm to optimize for the wrong signal. Fix measurement first, every time.[2]
Enabling Auto-Apply and Not Auditing It Monthly
Auto-apply configuration is not a one-time task. The set of eligible recommendation types changes — as demonstrated by the January 26, 2026 RSA change — and previously acceptable settings can become inappropriate after a campaign restructure, conversion-action change, or seasonal strategy shift. Failing to audit auto-apply monthly is a governance failure, not a minor oversight.[2][3]
Using a Simple Average for Portfolio Score Reporting
Averaging account-level scores without weighting gives disproportionate influence to small accounts and produces a misleading portfolio number. Use Google’s weighted formula — ∑(score × weight) ÷ ∑(weight) — for any multi-account rollup.[5]
Treating Score Uplift as a Revenue Forecast
A recommendation’s estimated score uplift — for example, “+18 points if you switch to broad match” — is an estimate of score movement, not a guarantee of incremental conversions, revenue, or profit. The figures Google cites (e.g., +35% conversions for exact-to-broad-match switch[3]) are vendor claims without published confidence intervals or population-level sample sizes. Treat them as directional hypotheses that require in-account testing, not as projections to quote in a media plan.
Dismissing Recommendations Without Logging the Reason
Dismissal without documentation creates confusion for future auditors, new team members, or account transitions. When a recommendation is dismissed, record why — in a shared document, a campaign note, or an internal tracking system — with enough specificity to reconstruct the reasoning. “Dismissed broad match because conversion volume is 18/month, below the 30-conversion minimum for stable Smart Bidding” is useful. “Dismissed” alone is not.
Making Multiple Major Changes Simultaneously
Changing match type, bid strategy, budget, and landing page at the same time makes it impossible to attribute subsequent performance changes to any single variable. Apply one material change at a time, define a review window (minimum 2 weeks; 4–6 weeks preferred for bid-strategy changes), and hold other variables steady before proceeding to the next change.[2]
Worked example (illustrative)
Isolating Variables After a Recommendation-Driven Change
- Setup: An Atlanta digital marketing account for a B2C subscription service spending $31,000 per month. In September 2026, the team simultaneously accepts three recommendations: (1) switches from Manual CPC to Target CPA at $55; (2) expands three campaigns to broad match; (3) increases the monthly budget from $31,000 to $42,000. By October 15, 2026, CPA has risen to $89 — a 62% increase — and subscription sign-up rate has dropped from 4.2% to 2.9%.
- Numbers: With three simultaneous changes, it is impossible to determine whether the CPA increase was caused by the bid-strategy learning period (typically 2–4 weeks), by broad-match query expansion introducing irrelevant traffic, by the budget increase funding lower-quality impressions, or by a combination of all three. The team has no clean before-and-after comparison for any single variable. Estimated wasted analytical time to untangle the root cause: 6–10 hours. Estimated wasted spend during the 2-week diagnostic period at $89 CPA vs. $55 target: ($89 − $55) × estimated 350 conversions × 2/4 weeks = $34 × 175 = $5,950 in excess CPA cost.
- Decision: Going forward, implement a change sequencing rule: (1) bid-strategy switch first with a 4-week hold; (2) match-type change only after CPA stabilizes within 10% of target; (3) budget increase only after steps 1 and 2 are validated. No more than one material change active at a time.
- Why: Google’s recommendation guidance states that changes should be tested and monitored individually; applying multiple material changes simultaneously prevents attribution of outcomes to any specific recommendation.[2]
12. What Changed Recently (Last 30 Days)
Google’s publicly documented changes through early November 2026 are concentrated in recommendation administration and auto-apply controls rather than a newly announced Optimization Score calculation methodology. The most operationally significant confirmed change is the removal of “Add responsive search ads” from automatic application, effective January 26, 2026. No new scoring formula, new weighting model, or clearly dated new recommendation type is identified in Google’s current public documentation for the 30-day window ending November 4, 2026.[2]
Removal of “Add Responsive Search Ads” from Auto-Apply
Beginning January 26, 2026, Google removed the Add responsive search ads recommendation from the Auto-apply settings page. From that date forward, the recommendation no longer automatically suggests or applies new RSAs.[2] This is a meaningful operational change for any account that had relied on auto-apply to populate ad groups with RSA copy without manual review.
What this means in practice:
- New RSAs must now be created through campaign workflows, the Google Ads API, scripts, or manual review of the Recommendations page.
- Accounts that previously had this recommendation enabled under a broad auto-apply bundle such as “Maintain your ads” should audit their Auto-apply History to confirm whether RSAs were created before the cutoff date and whether those ads align with current brand guidelines, offer accuracy, and legal copy requirements.
- Google explicitly notes that recommendation types are subject to change as best practices evolve, making periodic Auto-apply settings audits a standing operational task rather than a one-time setup activity.[2]
Worked example (illustrative)
Auditing RSA Auto-Apply After the January 26, 2026 Cutoff
- Setup: A Chicago-based legal services account spending $18,500 per month across 6 Search campaigns had enabled the “Add responsive search ads” auto-apply bundle in March 2025. The account manager reviews Auto-apply History in November 2026 and discovers that 14 RSAs were automatically created between March 2025 and January 25, 2026, across 4 ad groups — none of which were manually reviewed for compliance with Illinois attorney advertising rules.
- Numbers: Of the 14 auto-created RSAs, 3 contain the phrase “guaranteed results,” which is prohibited under Illinois Rules of Professional Conduct Rule 7.1. Those 3 ads have collectively accrued 8,400 impressions and 112 clicks at an average CPC of $14.20, representing $1,590 in spend on potentially non-compliant copy. The remaining 11 RSAs contain no prohibited language but have ad strength ratings of “Poor” (4 ads) or “Average” (7 ads) — below the account’s internal threshold of “Good.”
- Decision: Pause all 3 non-compliant RSAs immediately. Submit revised copy for legal review within 5 business days. Evaluate the 11 remaining RSAs against the “Good” ad strength threshold and either improve asset variety or replace with manually written ads. Set a calendar reminder to re-audit Auto-apply History every 60 days.
- Why: Google states that automatic application of recommendations does not remove the advertiser’s responsibility for ad content; the RSA auto-apply removal effective January 26, 2026 means no further auto-created RSAs will appear, but previously created ads remain live and must be audited manually.[2]
Auto-Apply Remains Selective, Not Universal
Google continues to support automatic application for only a select subset of recommendation types. Recommendations are applied only when relevant to a given campaign’s configuration, and advertisers retain the ability to review the queue on the Recommendations page, dismiss individual items, or disable auto-apply entirely for any category.[2] Google’s documentation also reconfirms that automatic application does not increase campaign budgets, though advertisers must separately monitor whether existing budgets become a binding constraint after automated changes shift spend distribution.[2]
For teams managing accounts via the API, Google provides RecommendationSubscriptionService to programmatically subscribe to supported recommendation types. However, the set of recommendation types available through the API and those available in the Google Ads UI are not necessarily identical, so API-based governance rules must be verified against current API documentation independently of UI settings.[5]
Recommendation Types Remain Dynamic
Google’s current “Types of recommendations” documentation confirms that recommendation categories evolve with Google Ads best practices and that advertisers may not see every type in every account.[5] The broad areas currently covered are summarized in the table below, along with their operational risk level for auto-apply purposes.
| Recommendation Category | Examples of Types Included | Auto-Apply Risk Level | Recommended Governance | Source |
|---|---|---|---|---|
| Bidding and budgets | Switch to Target CPA, increase budget, adjust target ROAS | High | Manual approval required; test via experiment first | [2] |
| Keywords and targeting | Add broad match, add new keywords, expand audiences | High | Manual approval required; validate search terms and negatives first | [5] |
| Ads and assets | Improve ad strength, add asset types, add sitelinks | Medium | Manual review of copy, claims, and destination URLs before applying | [5] |
| Measurement | Turn on auto-tagging, fix conversion tag issues, add conversion actions | Low–Medium | Apply after verifying the fix is accurate; never apply if it would duplicate conversions | [5] |
| Campaign maintenance | Repair disapproved ads, fix feed issues, resolve policy violations | Low | Verify the specific fix before applying; safe for narrow auto-apply if monitored | [2] |
| New campaign features and setup | Enable Performance Max, add new campaign types, configure new settings | High | Manual approval required; structural changes affect budget allocation and reporting | [5] |
Worked example (illustrative)
Discovering a Newly Appeared Measurement Recommendation in November 2026
- Setup: A San Francisco e-commerce account spending $22,000 per month runs a weekly Recommendations page review on November 3, 2026. The account manager notices a new measurement recommendation: “Turn on auto-tagging” — which appears because a recent site migration to a new CMS inadvertently disabled auto-tagging on the domain, breaking GCLID capture for 100% of paid traffic since October 20, 2026.
- Numbers: The account recorded 0 Google Ads-attributed conversions between October 20 and November 3, 2026 (14 days), despite Google Analytics 4 showing 318 transactions totaling $41,900 in revenue during the same period. The Optimization Score shows a measurement recommendation with an estimated uplift of 8.4 percentage points. The account’s score dropped from 76% to 61% over the same 14-day window, consistent with the broken tracking degrading the score.
- Decision: Re-enable auto-tagging immediately via Admin → Account settings → Auto-tagging. Verify GCLID capture within 24 hours using Google Tag Assistant. Import the 318 offline transactions from the 14-day gap via enhanced conversions or offline conversion import where the order data is available. Do not apply any bidding recommendations until conversion data is restored and stable for a minimum of 7 days.
- Why: Google’s recommendation guidance states that automated bidding and many recommendations depend on reliable conversion actions; fixing broken measurement is the highest-priority recommendation category before any bidding or targeting change is considered.[1]
No New Scoring Formula Confirmed for the Current 30-Day Window
Google continues to define Optimization Score as an estimate of how well an account or campaign is configured to perform, expressed on a 0–100% scale, with individual recommendation uplifts ranging from less than 0.1 to 100 percentage points.[1] Applying or dismissing recommendations changes the score, and both the score and the available recommendations change as campaign conditions, settings, performance, and Google’s models evolve.[1][2]
No new weighting model, new score metric, or new calculation methodology is identified in Google’s current public documentation for the 30-day window ending November 4, 2026. The defensible interpretation is that Google has updated documentation and recommendation availability — most notably the RSA auto-apply removal — while retaining the existing score framework.[1][5]
Google also continues to surface Optimization Score and recommendations during campaign construction for Search and Performance Max campaigns. During setup, the score is calculated in real time using the campaign’s current settings, available account and campaign statistics, the estimated impact of available recommendations, and recent recommendation history where available.[6] This means a newly created campaign can receive a sub-100% score immediately at launch if Google identifies configuration gaps, and account managers should treat that initial score as a setup checklist rather than a performance verdict.
Worked example (illustrative)
Interpreting a Low Score on a Newly Launched Performance Max Campaign
- Setup: A Portland outdoor retailer spending $9,000 per month launches a new Performance Max campaign on November 1, 2026 with a $3,000 monthly budget, a target ROAS of 400%, and a product feed of 850 SKUs. The campaign receives an Optimization Score of 54% at launch, with 6 recommendations shown, estimated total uplift of 46 percentage points.
- Numbers: The 6 recommendations and their estimated uplifts are: (1) Add audience signals: 18 points; (2) Add more asset group creative (video): 12 points; (3) Set up conversion value rules: 8 points; (4) Add more sitelink assets: 4 points; (5) Enable auto-tagging (already enabled — recommendation appears in error and is dismissed): 2 points; (6) Connect a Google Merchant Center product group: 2 points. Total potential uplift if all applied: 46 points, bringing score to 100%. The erroneous auto-tagging recommendation is dismissed immediately because auto-tagging is already active, confirmed in Admin → Account settings.
- Decision: Apply recommendations 1, 3, 4, and 6 within the first 7 days of launch (audience signals from first-party CRM list of 12,400 customers, conversion value rules assigning 1.5× weight to transactions over $150, 4 additional sitelinks, and Merchant Center product group link). Defer recommendation 2 (video assets) until the internal creative team delivers a 15-second and 30-second product video by November 14, 2026. Dismiss recommendation 5 with the reason “auto-tagging already enabled — confirmed November 1, 2026.”
- Why: Google calculates Optimization Score in real time during campaign setup using available settings and recommendation impact; a 54% launch score on a Performance Max campaign is normal and reflects missing configuration rather than poor performance, since the campaign has 0 conversion data at that point.[1][6]
Recommended Account Actions for November 2026
Based on the confirmed changes and current documentation, the following actions are specifically relevant for US Google Ads accounts in November 2026:
- Audit Auto-apply settings immediately. Navigate to Campaigns → Recommendations → Auto-apply settings and confirm which recommendation types remain enabled. The RSA change effective January 26, 2026 means that any bundle previously including “Add responsive search ads” has already been modified by Google; verify that no unintended categories were left active after that adjustment.[2]
- Review Auto-apply History for RSAs created before January 26, 2026. Any automatically created RSAs from that period are still live and must be audited for brand compliance, legal accuracy, offer validity, and ad strength. Pause or replace any that do not meet current standards.
- Do not treat 100% as a mandatory November target. Google’s vendor claim states a median 14% conversion increase for a 10-point score increase[7], but this is an internal estimate, not a guaranteed outcome. A score below 100% with deliberately dismissed recommendations is a valid and defensible account posture.
- Validate conversion tracking before acting on any bidding or targeting recommendation. The measurement recommendation category — including auto-tagging, conversion tag repair, and conversion action setup — is the highest-priority area and should be resolved before any score-raising activity in other categories.[1]
- For API users, verify supported recommendation types against current API documentation. The UI’s auto-apply options and the types available via
RecommendationSubscriptionServiceare not guaranteed to be identical, and the supported set changes as Google updates its recommendation framework.[5] - Monitor recommendation history on a weekly cadence through Q4 2026. Holiday season campaigns (November–December 2026) are particularly susceptible to recommendation drift as budgets, bids, and traffic patterns shift; a weekly review prevents automated or recommended changes from going undetected during high-spend periods.
- Use experiments before broad application of any Q4 budget or bid recommendation. Google allows eligible bid and budget recommendations to be saved as experiments before full rollout[8]; this is especially important in November and December 2026 when a bidding misstep during peak season has outsized revenue consequences.
Worked example (illustrative)
Managing Optimization Score During a Q4 2026 Peak Season Budget Recommendation
- Setup: A Minneapolis gift retailer spending $28,000 per month in October 2026 receives a budget recommendation on November 7, 2026 suggesting an increase to $52,000 per month — an 85.7% increase — across 3 Shopping campaigns. The estimated score uplift for accepting this recommendation is 22 percentage points, which would bring the account from 71% to 93%. Google’s forecast projects 480 additional conversions at an average CPA of $29.50 during the November 7–December 31, 2026 window.
- Numbers: Current campaign marginal CPA over the prior 30 days (October 7–November 6, 2026): $31.80. Business break-even CPA: $45.00. Google’s forecast CPA of $29.50 is within the break-even threshold. However, the $24,000 incremental monthly spend ($52,000 − $28,000) represents 85.7% more budget with no test data confirming that marginal conversions at the higher spend level maintain a CPA below $45.00. The account has 90 days of conversion history (August–October 2026) showing that CPA rises from $29.00 at $20,000 spend to $33.00 at $28,000 spend — a $4.00 CPA increase per $8,000 of additional spend, suggesting diminishing returns. Projecting that trend: at $52,000, estimated CPA = $33.00 + ($24,000 / $8,000) × $4.00 = $33.00 + $12.00 = $45.00 — exactly at the break-even threshold, with no margin for error.
- Decision: Save the budget recommendation as an experiment rather than applying it outright. Increase budget to $38,000 (a $10,000 or 35.7% increase) for the week of November 10–16, 2026. If CPA remains below $40.00 (a 25.8% buffer below break-even), increase to $45,000 for November 17–30, 2026. Do not proceed to $52,000 unless CPA holds below $40.00 at the $45,000 level.
- Why: Google’s recommendation guidance states that budget recommendations should be evaluated against marginal CPA and business targets, not accepted solely because they project more conversions; the account’s own historical data suggests the $52,000 level lands exactly at break-even with no efficiency buffer.[2]
The overarching principle for November 2026 remains unchanged from earlier in the year: optimize the account for its business objective, not for a 100% score. The RSA auto-apply removal is a useful reminder that Google’s recommendation framework is not static — settings that were acceptable in 2025 may no longer reflect current best practices, and the reverse is equally true. Build a standing monthly audit of Auto-apply settings, Change History, and recommendation categories into every account’s governance calendar, and judge every recommended change by its effect on qualified conversions, revenue, profit, and CPA or ROAS — not by the score points it would add.[1][2][5]
Related reading
- Building a Google Ads Strategy
- Google Ads Optimization: Which Lever To Pull
- Google Ads Auction Insights & Competitor Analysis
- Audience Targeting
Or skip the work and get a one-off Search campaign setup with no monthly management.
References
- [1] https://support.google.com/google-ads/answer/9061546?hl=en support.google.com
- [2] https://support.google.com/google-ads/answer/9061547?hl=en-AU support.google.com
- [3] https://support.google.com/google-ads/answer/12994751?hl=en support.google.com
- [4] https://ads.google.cn/intl/en_sg/home/tools/recommendations/ ads.google.cn
- [5] https://developers.google.com/google-ads/api/docs/recommendations developers.google.com
- [6] https://ads.google.cn/intl/en_ng/home/tools/recommendations/ ads.google.cn
- [7] https://business.google.com/en-all/ad-tools/campaign-recommendations/ business.google.com
- [8] https://support.google.com/google-ads/answer/10169817?hl=en support.google.com

