Google Ads recommends a larger budget, broader reach or a different bidding strategy. The account’s Optimization Score promises to improve. That still leaves the owner’s question unanswered: will this change bring useful customers at a cost the business can accept?
Our recommendation is to review each action against the business goal before applying it. Keep a short record of what changes, why it is justified and how the result will be judged. A lower score is acceptable when the proposed action conflicts with a documented commercial constraint.
A higher score can mean that nothing changed in the campaign
Google describes Optimization Score as an estimate of account setup, expressed from 0–100%. Both applying and dismissing recommendations affect it. Reaching 100% therefore does not demonstrate more sales or a better margin.
Do not confuse it with Quality Score, the keyword-level Search diagnostic discussed in our Quality Score article. These tools answer different questions. Neither replaces the order system or CRM when judging customer value.
The useful question is why a recommendation appeared. Check the focus shown in the account and the actual bidding objective. A campaign pursuing clicks can generate sensible click-oriented suggestions while the business needs accepted enquiries. The platform’s inferred objective may not describe your commercial priority well enough.
Read the goal and the conversion before the recommendation
Write one sentence describing a useful outcome. For a shop, that might be a retained order with an acceptable contribution after variable costs. For B2B, define an accepted enquiry using task, service area and sales eligibility. Keep the same definition while evaluating the change.
Then inspect the conversion actions used by the campaign. A contact-button click, a delivered form and a qualified opportunity describe different stages. If bidding treats a weak signal as the final goal, more of that signal can look like progress without creating a useful customer.
Check duplicate actions, value settings and the delay before outcomes become reliable. Do not switch strategy and repair measurement simultaneously if you need to understand which change mattered. When tracking is incomplete, the first decision may be to defer the recommendation while reconciling a sample of enquiries with the CRM.
This is also where business restrictions belong: unavailable products, delivery territory, sales capacity and the maximum acceptable loss during a test. An agency should be able to name the restriction behind a rejection. “We never use automation” is not a sufficient reason.
Choose an action for the specific recommendation
The following table is a review framework, not a claim that every account receives all these suggestions. Read the exact recommendation and its affected campaigns before choosing a response.
| Proposed change | Evidence needed before acting | Suitable response |
|---|---|---|
| Increase budget | Reliable goals, incremental capacity, unit economics and a spending ceiling | Test within the approved exposure, or defer |
| Broaden keywords or targeting | Query intent, exclusions, landing-page fit and useful lead feedback | Test a limited scope rather than every suggestion together |
| Change bidding strategy | Correct conversion actions and values, outcome delay and a comparable baseline | Define the evaluation first; defer if measurement is unreliable |
| Add or revise assets | Accurate claims, usable destination and appropriate creative | Apply a verified correction; evaluate performance separately |
| Remove an apparently redundant element | Its role, history and possible overlap with other changes | Inspect the detail; reject when the proposed removal loses needed coverage |
A recommendation can identify a genuine omission. Dismissing everything to protect an existing account structure is no better than applying everything to obtain a percentage. The owner needs a reasoned decision, including a clear statement when the available evidence is insufficient.
Put the spending consequence in money
Consider an illustrative proposal, not a client case: the approved daily budget would rise from €100 to €150. Over a 14-day planning window, the difference is €700 if those daily planning amounts were spent. This is a scenario for exposure, not a Google billing forecast or a promise that the additional budget will be used.
Ask what must improve to justify that exposure. If the shop cannot fulfil additional orders, or sales cannot process enquiries promptly, extra reach may create costs before value. If capacity exists and current useful outcomes are constrained by reach, a bounded test can be reasonable.
Record actual expenditure afterwards. Do not judge the decision only by a cheaper platform CPA: cancelled orders, unsuitable enquiries or a changed conversion mix can lower that number. Use the same business definition and allow outcomes to mature before evaluating the comparison.
Changing several settings together also makes the result harder to interpret. Choose the smallest coherent intervention. A necessary tracking repair is a repair; it does not need to be presented as a statistically proven campaign uplift.
Review automatic application separately
Google’s auto-apply controls deserve their own review. Inspect enabled categories, who authorised them and the history of applied changes. A considered manual decision about one recommendation does not authorise every future recommendation in that category.
Agree which routine corrections may run automatically and which changes require a person to review the commercial consequences. Keep that rule specific enough to use. The relevant question is the change’s reach and reversibility, rather than a general preference for or against automation.
If performance shifts unexpectedly, compare the timing with change history. That can identify a candidate explanation. It does not prove causation when promotions, competitors or the offer changed at the same time. Restore a previous setting only with a reason and a verification plan.
What a useful agency review should leave behind
Keep one row per recommendation: date, exact proposal, affected scope, current objective, supporting evidence, apply/test/defer/dismiss decision, owner, maximum exposure and review condition. A screenshot of an improved score is not an adequate handover.
For a test, specify what would justify continuing, stopping or retaining an inconclusive result. Use a period that accounts for conversion delay and normal business variation. Do not stop at the first favourable day or invent a universal number of conversions that suits every account.
Canem Errant’s Google Ads service can review one recommendation against campaign settings and confirmed outcomes. Start with the disputed action and the decision it would change. The useful deliverable is an explanation the owner can approve, even when that explanation is to leave the setting alone.
Frequently Asked Questions
Should an agency keep Optimization Score at 100%?
It should review recommendations and explain its decisions. Because dismissal can also increase the score, the percentage alone is not a reliable acceptance criterion for campaign management.
Does dismissing a recommendation mean ignoring a problem?
It can be a justified decision when the suggestion conflicts with the goal or operating limits. Record the reason and when it should be reconsidered; the account’s circumstances can change.
When is it reasonable to apply a recommendation immediately?
When the correction is verified, its consequences are understood and it sits within agreed authority. Changes to spending, reach or bidding deserve an explicit business evaluation before application.
