Google Ads Optimization Score: Review Recommendations With Your Own Data

Understand what changes the score

Google calculates the score using account statistics, settings, campaign status and recommendation history. Applying or dismissing recommendations can change it. A score movement therefore needs context about the actions and suggestions behind it.

The official score explanation describes this calculation and its availability. Check the native Recommendations page for the current score and the focus Google is using.

A hypothetical rise from 78% to 90% is a 12 percentage-point increase in this score. That movement does not specify a 12% improvement in sales, revenue or CPA. Those outcomes need their own reporting.

SourcetoMCP’s recommendation list does not return the headline score or each suggestion’s score uplift. Supply those values from Google Ads if you want them in the review, and keep them separate from the performance estimates returned by the connector.

Separate optimization score, Quality Score and business results

Google states that optimization score is not used by Quality Score in its score-checking guidance. Review them as separate diagnostics.

Use our Quality Score article to inspect ad-quality components. For a broader performance decision, keep the conversion definition and business target beside both scores.

Read the recommendation and its estimated impact

For each suggestion, record its ID, type and affected campaign, plus the exact action described in Google Ads. SourcetoMCP can return available baseline and potential clicks, cost and conversions from Google’s recommendation feed. Some recommendations have no impact estimates.

Potential metrics are Google’s estimates. Baseline metrics in this feed describe the basis Google used for its estimate and should not be silently relabelled as your independently requested reporting period.

The following hypothetical example illustrates how to evaluate an estimate. It is not a recommendation retrieved from a customer account.

The implied incremental cost is AUD 300 divided by five extra conversions, or AUD 60 per additional conversion. That is a different calculation from the potential average CPA of AUD 43.33. Compare both with the business economics before deciding whether the test is acceptable.

Only calculate incremental cost when the estimate has a positive conversion difference. A missing estimate or no additional conversions needs a different explanation, rather than division by zero or a fabricated improvement.

Compare the suggestion with current account evidence

Read campaign performance for a complete period that fits your review. Check spend, conversions, conversion value where meaningful, current budget and bidding strategy. Preserve the actual dates and attribution caveats.

A budget recommendation needs evidence about efficiency and the source of lost visibility. Use the search impression-share workflow to separate budget constraints from rank constraints.

A keyword recommendation needs a check against actual query relevance and the offer. A bidding recommendation needs reliable conversion definitions and a clear business target. A tracking recommendation needs validation of what the site and account currently measure.

Review recommendations for [account]. Read the current feed and campaign performance for [complete period]. For each suggestion, show its ID, type, affected campaign, available baseline and potential metrics, and any missing information. Label Google’s impact figures as estimates. Compare the proposed action with our goal [definition and target]. Classify it as investigate, propose a test, or decline with a specific reason. Make no changes.

If the feed lists more recommendations than the response includes, mark the review as incomplete. Do not turn a partial list into a statement that every account recommendation has been evaluated.

Review the exact action before applying or dismissing it

SourcetoMCP supports applying or dismissing a single recommendation by its ID. Applying uses Google’s suggested values. The operation does not expose custom parameter choices for every recommendation type, so inspect the native recommendation when you need to understand or adjust the proposed values.

Avoid approving an ID based only on a broad type label. Some actions can change budgets, bidding or keywords, so first inspect the exact suggestion in Google Ads and confirm that its current values match your intended action.

For recommendation [ID] in account [ID], show the exact proposed action from the native recommendation details I supplied: [details]. Confirm whether applying it through the connector uses these same suggested defaults. State the scope and budget effect, and wait for approval. After an approved action, report the result and read the relevant account or campaign state where available.

Dismissing a suggestion can also affect the score. Record the business reason so a later reviewer understands why the account did not follow that recommendation. A different score after dismissal does not describe a change in customer outcomes.

Save the decision and evaluate the actual result

For each approved test, keep the recommendation ID, retrieved estimate, observed baseline, action time, owner and review condition. A record of the estimate lets you compare it with the result without rewriting the original expectation.

After comparable data has processed, assess actual spend, conversions and qualified outcomes. Keep concurrent campaign or tracking changes visible. A result that differs from the estimate should prompt a review of the assumptions and conditions.

Add the accepted action and next review date to your weekly Google Ads optimisation routine. Leave any suggestion without enough supporting detail in the investigation queue.