Optimise Google Ads settings for agencies: review context before applying a change

Define the agency settings-review job

Optimising settings means choosing account configuration that fits a defined client objective. It is not a universal instruction to make every account resemble the same checklist. An agency needs current evidence, commercial context and a precise next decision.

The practical constraint is consistency across clients with different offers and structures. A setting that deserves investigation for a local service provider can be intentional for a travel business. A shared review format should preserve those differences rather than produce identical changes.

Start with the question the agency wants to answer. Is the account targeting the intended geography, using the selected conversion definitions or retaining an obsolete tracking template? Each question needs its own relevant evidence and supported operation.

Keep an audit recommendation separate from an execution request. A finding can direct attention without quantifying commercial loss or authorising a mutation. The agency should decide what action fits the evidence before asking the human to confirm.

A useful review ends with an assessable decision record: current context, client instruction, exact target and proposed effect. It makes the change understandable across specialists and account managers without promising guaranteed performance improvement.

Confirm the client's actual objective

The client supplies commercial facts that the account cannot always reveal. What it sells, where it can serve and which outcomes matter can change the settings decision. A campaign name is not independent proof of those facts.

Choose the measured outcome explicitly when performance informs the review. All primary conversions, primary purchases, primary leads and a named action are different selections. A proposal based on one should not be justified using another without explanation.

Confirm any constraints relevant to the operation. A client may want a campaign paused during a closure or a particular product excluded from one group. The assistant should not infer that the same instruction applies to every related asset.

Keep supplied facts separate from provider evidence. The business can confirm serviceability; the tool can read targeting settings and supported results. Neither contribution should be represented as independently verified by the other.

When the objective changes during the conversation, revisit the recommendation. A settings change that made sense under the original question may no longer fit. The agency should prepare the operation under the current agreed instruction.

Read current configuration before judging it

Read the relevant current setting for the resolved target. An old report or a remembered configuration can provide discussion context, but it should not establish what exists now. Another account user may have changed the value.

Use the appropriate entity level. Account auto-tagging, a campaign location option and an ad-group setting belong to different resources. A generic request for all settings can retrieve more than the decision needs while still missing the relevant relationship.

The reporting tools can discover readable fields and compatible plans. A local catalogue describes support; a successful provider read establishes the current value. The distinction helps the assistant avoid unnecessary calls without inventing account facts.

Keep unavailable evidence unknown. A failed setting read does not prove the value is off, absent or zero. The next task is to resolve access or choose a supported plan rather than prepare a mutation from an assumption.

Preserve the resolved target in follow-ups. Asking whether the same campaign uses another option should retain its account and name context. Ambiguity needs clarification, but established context should not be discarded routinely.

Treat audit findings as defined recommendations

An audit check has a defined population and criterion. A fail means that selected evidence did not meet that criterion. It does not automatically prove the account is commercially unprofitable or identify a guaranteed saving.

Read the explanation and relevant target context. A location heuristic, negative-presence check and conversion-counting recommendation answer different questions. Their operational implications depend on the client's offer and measurement design.

Keep unavailable sections visible. A partially completed audit should not become a clean bill for unread areas. The agency can investigate successful findings while retaining the missing coverage as an open task.

Audit rendering and conversational edits also have different support paths. A web finding can have a specific correction mechanism without a universal equivalent exposed to the assistant. The operation must be checked in the current conversational catalogue.

Use the audit to ask the next useful question. Does this setting match the client's intention? Is the recommended operation supported for this target? Has the relevant current state changed? This turns a checklist into a precise review rather than automatic blanket edits.

Discover supported operations locally

The edit catalogue describes action names, valid target kinds, parameters and prerequisites. The dependency framework describes producers and relationships for supported sequences. These local structures help prepare a plan before unnecessary provider reads.

An account-wide action should not discover campaigns merely because some other operations need campaign parents. The app's account app-category exclusion entry is one example of a scope that does not require that search.

Some actions are unavailable in the ordinary conversational flow. Auto-apply subscriptions and a universal HTTPS destination conversion are not safe promises from their catalogue names alone. Read their actual support notes.

A provider endpoint can exist without every operation being exposed through this path. Current tool support, compatibility and verification requirements determine what can be prepared. A readable API schema is not an unrestricted execution interface.

Explain the real boundary when it affects the user. The assistant can identify the appropriate alternative workflow or missing prerequisite without claiming that Google universally cannot perform the action.

Choose the correct entity scope

Scope determines who and what the operation affects. An account setting can influence more than one campaign. A campaign change can be wider than an ad-group or keyword change. The proposal should make that entity level obvious.

Resolve repeated names and relevant parents. A keyword belongs to an ad group in a campaign; a Performance Max asset-group item has another hierarchy. A guessed parent can turn a reasonable recommendation into the wrong target.

Shared assets and lists need usage context. Changing a shared list can affect attached campaigns, while an unused list may not affect the intended campaign at all. The name alone does not establish scope.

Do not broaden the user's request silently. A confirmed mismatch in one group is not an instruction to exclude the same phrase account-wide. The agency should preserve the selected target and commercial reason.

Use recognisable names in the human review and keep resource identifiers internal. The client needs to assess which asset and parents will change, not decode technical syntax. Precise scope is part of readable output.

Review location options with commercial context

Start with whom the client intends to reach. A business serving local customers can have a different objective from one attracting travellers or buyers interested in a place. The setting cannot be judged without that intention.

Google's advanced location guidance distinguishes presence and location-interest options and notes that location signals are not perfectly accurate. A presence setting is not a guarantee that every person physically matches the advertiser's expectation.

The audit's location recommendation is a heuristic. It does not establish the amount of irrelevant geographic spending or prove that location interest is wrong for every client. Confirm the offer and serviceability context.

Target settings and geographic performance are different evidence. If the decision needs observed outcomes by location, select a supported report with the appropriate dimension rather than use the configured targets as a substitute.

Prepare an exact requested campaign operation after the decision is clear. A location-option change and adding or removing geographic criteria are different actions. The human should see which one the proposal contains.

Assess network and channel choices

Network options can be reviewed against the campaign's purpose and supported evidence. The current catalogue includes specific actions such as disabling Search Partners or Display Network where applicable. Their availability is not a universal recommendation to apply them.

Read the current relevant setting and, when needed, the compatible performance population. A broad account outcome does not independently establish the contribution of a particular network. Keep the evidence scope explicit.

Do not treat one client's preferred setup as a mandatory configuration for every account. The agency can reuse the question and review method while preserving each client's objective and campaign compatibility.

Keep other automation settings within their own support notes. A named action can need provider preflight or an existing prerequisite. The assistant should not promise an accepted mutation from the label alone.

After a supported change, verify the configured result. Delivery and selected outcomes need a later performance review. A checked network setting does not prove that all subsequent cost or conversion movement was caused by the edit.

Separate tracking settings from implementation health

An account flag can be read and changed without proving the entire measurement journey works. Auto-tagging, tracking templates and landing-page behaviour are connected, but they are not one interchangeable health state.

Enabling auto-tagging can establish the selected account setting. It does not prove that every redirect preserves click identifiers or that the destination implementation records the intended event. Those questions need their own relevant evidence.

Clearing a tracking template also needs context. The existing template may serve a deliberate measurement purpose. A broad cleanup recommendation should not remove it without understanding what the client expects it to do.

Do not advertise universal automatic URL repair. The ordinary conversational HTTPS-fix action is unavailable because destination verification is required. Normalising a provided URL during supported creative preparation is narrower.

Assign implementation investigations to the right owner. The agency can prepare a supported account-setting operation while the website or measurement specialist checks actual behaviour. The receipt should not claim both tasks were completed by one write.

Review conversion roles and counting deliberately

Primary status and counting method affect measurement and optimisation context. A client-specific review needs the intended business action, current type and origin, and applicable support rules. Do not infer the correct role solely from a friendly display name.

The current conversion policy excludes certain locked types and origins from these edits. The assistant should respect that boundary instead of preparing a universal role or counting change for every action in the account.

For a bulk role operation, the explicit selected targets and primary set determine the effect. The reviewer needs to understand which selected actions become primary and which become secondary. It should not read as an account-wide hidden rewrite.

Counting recommendations also need the intended event semantics. The audit can identify its heuristic, but it does not independently prove duplicate tracking or reconstruct the client's business process. Confirm the objective before applying the supported change.

After execution, verify the relevant field. Later aggregate conversion reports can change definition when roles change, so the agency should record that amendment rather than compare old and new totals as though nothing changed.

Keep budget and bidding decisions distinct

A budget amount and a bidding target have different purposes. The ordinary conversational budget action covers non-shared daily budgets. Target CPA and target ROAS actions work with existing supported targets and their defined parameters.

Set amounts use the declared unit. Budget and CPA set values use account currency, while target ROAS uses its display-percentage meaning. A value without that context can be misread during a client review.

A pacing projection is not an automatic bidding instruction. It estimates cost under a recent pattern and configured period target. The agency still needs the client decision and exact campaign operation.

Do not create an unsupported strategy transition implicitly. Changing an existing target differs from choosing a new strategy or inventing a target where none exists. The proposal should match current support and target evidence.

Verify the setting after the requested operation, then review later outcomes separately. A saved value does not guarantee daily cost, acquisition volume or return. The client's agreement and provider configuration remain distinct records.

Review exclusions and keyword diagnostics

A negative-keyword presence check asks whether relevant exclusions exist in its defined context. It does not prove the list is suitable or quantify wasted cost. The agency needs actual intent and offer context before adding exclusions.

The standard conversational negative actions use exact match at campaign, group or attached shared-list scope. Show the text and scope plainly. Do not imply that one exact negative excludes every related phrase.

Google's Quality Score guidance describes the score as a diagnostic, not a business-performance KPI. A low value can prompt investigation without proving the keyword should be paused.

Use a selected outcome and meaningful dates when performance informs the decision. Zero recorded purchases, unavailable value and a confirmed offer mismatch are different observations. Keep those distinctions in the recommendation.

Prepare the requested supported operation only after the target and action are clear. Pausing a keyword, adding a negative and changing the parent group's status have different effects and should not be hidden under one cleanup label.

Resolve product and asset-group context

A Shopping product decision needs more context than an item name. Performance Max listing selections belong to asset groups; Standard Shopping selections belong to ad groups. The same item can appear in several groups with different inclusion decisions.

Ask which campaign and group are in scope before proposing a change. A product excluded from one enabled asset group can remain included elsewhere. Counting each group-level occurrence answers a different question from counting unique items in the catalogue.

The current reporting framework can work with those listing-group relationships. Individual item leaves and broader category or brand selections are different tree nodes. A broad exclusion must not be described as a verified list of every individual item beneath it.

When performance supplies a filter, define the dates and outcome before reading the product selection. “Excluded items with purchase ROAS above four last month” combines current targeting with historical performance; it does not prove that excluding an item caused its later performance.

For a supported edit, preserve the requested group and exact selected products. Do not silently widen an individual inclusion request into replacement of the whole asset group's selection. The proposal should disclose replacement when that is the operation being prepared.

Inherited bids and ambiguous tree structures can prevent a safe Standard Shopping replacement. A refusal with a specific reason is preferable to flattening a tree and losing a deliberate configuration. Review the relevant hierarchy before authorising a broader restructuring.

Preserve the complete creative effect

Creative settings are often discussed as small adjustments even when the underlying operation replaces a complete asset. An agency reviewer needs to know whether the proposal changes one field, replaces a text collection, clears paths or creates a separate paused ad.

For a standard responsive search ad replacement, review the desired headlines, descriptions, final URL and display paths together. Current replacement behaviour also removes pins. Retaining the serving status does not mean every other part of the previous creative survives.

Provide the actual approved landing page. Normalising a bare domain into a public HTTPS URL can prepare its syntax, but that does not independently verify destination availability, checkout operation or conversion tracking. Those require their own appropriate checks.

An agency can brainstorm headlines conversationally before asking for any live action. Suggested language is a draft, not evidence of a client's guarantees or credentials. Supply the factual claims that may be used and review the proposed text before creation.

Creation starts a standard RSA paused in the supported Search context. Creating, enabling and passing Google's policy review are separate outcomes. A content verification receipt should not imply that an ad is approved or currently delivering impressions.

When an audit suggests unpinning, assess why the pin exists. It might reflect a contractual requirement or an intentional message order. The recommendation can identify a configuration worth reviewing without deciding the client's legal and brand obligations.

Work through an invented settings review

The following situation is invented for explanation. An agency manages advertising for Harbour Repairs, a fictional local service business. The client confirms that technicians accept work only within a defined service area and that completed appointment requests are the chosen lead outcome.

A review identifies three questions: whether location options fit that service area, whether an old tracking template is still required and whether a particular negative keyword blocks a newly offered service. These are different decisions with different owners.

The account specialist reads current campaign locations and the location option. The client confirms the commercial restriction. If a supported presence-setting change is appropriate, the specialist prepares that named campaign action without changing its selected country or inventing a radius.

The tracking question remains open until the agency's measurement specialist identifies the template's purpose. A template can belong to a third-party measurement arrangement. Its age alone is not sufficient justification to clear it across the account.

The negative-keyword question is reviewed against the revised service offer. The specialist resolves whether the negative belongs to a campaign, ad group or attached shared list. The proposal uses that actual target rather than claiming to remove every matching exclusion.

The team approves each settled decision under its normal authority process. An uncertain tracking question is left unchanged. This is a successful review because it makes the justified changes and preserves the configuration whose purpose has not been established.

Keep the fictional review record specific

The record says that the location option was changed for the named campaign after the client confirmed its service-area requirement. It does not say that irrelevant traffic was eliminated or that lead quality improved immediately.

For the negative, the record names the removed entry and its parent scope. If an equivalent exclusion remains elsewhere, that remains a separate observation. “The requested negative was removed” is narrower and more useful than “the account now covers the service.”

The unresolved template gets an owner and a question: confirm the downstream system before deciding whether to clear it. That handoff prevents another specialist from treating an open investigation as an approved account-wide cleanup.

Prioritise one coherent decision

Prioritisation should consider the client's objective, strength of evidence, affected scope and reversibility. A large number of audit findings does not tell the agency which action matters most. Several low-priority recommendations can coexist with one important measurement question.

Separate factual corrections from performance experiments. Removing an obsolete exclusion after the client confirms a new service is a different type of decision from testing a network option because recent outcomes look weak. The second requires a defined observation plan.

Do not bundle unrelated changes merely because one reviewer is available. If a tracking adjustment and budget increase happen together, a later shift in recorded CPA becomes harder to interpret. Coherent sequencing can be more informative than a broad cleanup.

Where actions genuinely depend on each other, describe that dependency. Creation may require the parent entity first. A change to an existing campaign budget usually does not need discovery of unrelated asset groups. The local operation catalogue helps establish those requirements.

Use the client's operational calendar to choose the moment. A planned promotion, service interruption or landing-page release can change the interpretation of performance. These facts should be supplied by the team rather than inferred from settings alone.

Leave a clear outcome for deferred items. “Not changed because measurement ownership is unresolved” is a useful decision. A silent omission forces colleagues to repeat the investigation and can lead to a later change without the original caution.

Ask for supported evidence in Slack

A connected account channel gives the agency a place to ask for the relevant configuration and performance context. Be explicit about the client account, selected campaign and date range when those determine the answer. A general account question should not be silently expanded across clients.

A useful opening request is “Show the current location option and selected locations for our local service campaign.” That establishes the current setting. Follow with the commercial requirement before asking the assistant to prepare a change.

For performance evidence, name the chosen outcome. “Compare primary lead conversions and cost for the last thirty completed days” has a clearer definition than “show performance.” If comparing GA4 events, confirm the event that represents the site's lead outcome.

The assistant can discover available fields and operations from its local catalogues. Reading a supported capability does not need to fetch every campaign. Actual configuration and account performance still need the relevant account data.

Ask for a table or chart when that helps the decision. The framework formats returned data rather than inventing values to fill a graphic. Unavailable results should remain unavailable, and a limited preview should not be presented as a complete account total.

For a broader introduction to that conversation, use the agency account-question guide. It explains why current configuration, historical measurements and creative suggestions need different kinds of evidence.

Keep creative help separate from account evidence

The agency may ask for a clearer explanation of a setting, a draft client note or alternative headlines. Ordinary creative conversation does not require a provider read merely because Google Ads is being discussed. Treat those responses as authored suggestions.

A request for the client's actual spend or current language targeting is different. It requires the account's information. The assistant should not substitute a plausible default when the requested live setting could not be read.

Prepare readable targets and steps

An edit proposal should explain what will change in terms the agency reviewer can understand. Display the target's name and the required parents. A keyword needs its ad group and campaign; an account setting needs the connected account context.

Describe each logical action on a separate row or section. “Clear the account tracking template” and “set this campaign's location option to Presence” should never be compressed into a string whose target is unclear. The action verb belongs beside its own target.

State the intended value and the important replacement effects. A complete language selection replaces the desired set. An audience signal addition is an addition. A creative replacement can clear omitted paths. Reviewers need those distinctions before confirming.

For a budget change, show the currency amount or requested percentage with clear units. A target CPA is a currency value; a target ROAS percentage is not the same kind of number. Avoid a technical numeric representation that forces the reviewer to guess the unit.

The proposal is a pending instruction, not evidence of a completed change. It should say that no changes have been made yet and make the confirmation step obvious. Preparing a request must not itself apply the live edit.

If the requested action is unavailable, explain the specific boundary. A supported HTTPS URL normalisation is not a universal destination repair. An audit recommendation is not automatically an executable operation. The agency can then choose the appropriate manual path.

Use the agency approval handoff

ppctor's conversational edit confirmation is tied to the connected owner and the proposal context. It does not implement an agency-wide delegation chain in which any team member can approve for every client. Arrange the appropriate authorised handoff before relying on the button.

The reviewer should check the current target, requested values and complete effect against the settled client instruction. Human confirmation is especially important when a recommendation came from a heuristic or when the client supplied commercial information outside the account.

Pending proposals expire after fifteen minutes. A later approval should use a fresh proposal rather than assume that the earlier target and settings remain suitable. This short-lived approval window is distinct from the agency's longer-term decision record.

Execution checks the relevant settings again and can stop when they have changed. A second specialist may have edited the campaign while the proposal was pending. The earlier text must not override a conflicting current configuration without a renewed decision.

The visible change-in-progress state acknowledges execution. A repeated click is not a request to perform the same live mutation again. Wait for the outcome rather than creating duplicate proposals because the receipt has not yet appeared.

Use the agency approval workflow for the detailed distinction between authority, pending proposals, cancellation and final receipts. Cancelling a pending request is not rollback of an already applied change.

Read verification and uncertainty accurately

A verified execution receipt establishes the outcome checked for that operation. It can confirm the intended setting or created content. It does not establish future profitability, successful policy review or a functioning conversion implementation.

A provider-accepted result can have a narrower meaning than a fully verified one. Read the actual status and its detail. The agency should not replace a qualified outcome with a stronger client-facing claim simply because the request was sent.

When execution is uncertain, some changes may already exist. Do not repeat creation or another potentially duplicating action without checking the account. A communication failure and a mutation failure are different events with different recovery steps.

Inspect the specific target and desired fields when verification could not finish. If the account contains the intended result, record that observation and its timing. If it does not, resolve the uncertainty before deciding on a new approved request.

Keep the receipt attached to the decision it implements. A later colleague should be able to distinguish “the proposal was prepared,” “the setting was verified” and “the later outcome was reviewed.” Those are three separate stages.

Do not use a green execution status as a blanket audit pass. The operation checks its own result. Other findings, unread sections and unrelated settings remain in their original state until separately examined.

Measure the later outcome under a defined question

Follow-up begins with the reason for the change. A location-option adjustment should be reviewed against the client's geography and chosen lead outcome, not judged solely by a higher click-through rate. A tracking change first needs measurement validation.

Record the effective date, selected campaigns and the measurement definition. Compare suitable periods with adequate observation time. Recent outcomes can still be incomplete, and a short interval after a change rarely establishes its isolated effect.

Preserve commercial context such as promotions, stock changes and landing-page releases. If several things changed together, disclose that limitation. The agency can describe what happened after the setting edit without claiming the edit caused every movement.

For ratios, use the aggregated inputs for the selected period. Purchase value divided by cost is a different measure from all-primary conversion value divided by cost. Changing the numerator between the initial review and follow-up obscures the comparison.

A client may care about total-site revenue relative to advertising cost. Label that as blended ROAS or MER and confirm the cost scope. It should not be silently substituted for a Google Ads purchase ROAS question.

Decide the next action only after interpreting the evidence. Keeping the setting, revisiting it or collecting more information are legitimate outcomes. A review workflow should not force a fresh live edit every time new results arrive.

Separate review from configured automation

A recurring agency review does not itself authorise recurring mutations. The conversational approval applies to its proposal. It should not become permission to repeat the action whenever a later metric crosses an imagined threshold.

Configured Rules are a separate workflow with their own supported actions and safeguards. Current condition-based Rules ask for approval; automatic mode is restricted to a time-based primary trigger. Describe the selected mode accurately when documenting an agency process.

Alerts and budget pacing provide monitoring signals. They do not enforce an account-wide spending cap or automatically choose an optimised target. Their usefulness is bringing a defined condition to the team's attention for the appropriate review.

Use an audit for a periodic configuration check and a focused question for a specific current setting. Repeatedly reading the entire account to answer a narrow question wastes review effort and can obscure the relevant target.

A routine can therefore have separate steps: observe the configured signal, investigate the relevant account context, decide whether a supported change is justified, and confirm that proposed change. Each step has a distinct result and owner.

Review the routine when the client's objective changes. A newly introduced subscription offer, different service area or revised lead qualification can invalidate an older interpretation even if the underlying settings have not changed.

Retain the decision and its context

A useful handoff preserves the business reason, current evidence, exact action and verification result. It should allow another specialist to understand why a setting was retained or changed without reconstructing the entire conversation.

Include the affected named entities and parents where needed. Keep customer identifiers and internal technical details out of public examples and client prose when they add no decision value. Human-readable targets make review easier without weakening the execution's internal identity checks.

Document unresolved dependencies beside the decision. The measurement specialist may need to confirm a template, or the client may need to approve a service-area policy. Do not let an open dependency disappear into a general statement that optimisation is complete.

Assign the later review question and an appropriate owner. A date alone is not sufficient: the reviewer also needs the selected outcome, scope and evidence needed to judge it. This prevents a later report from measuring a different business goal.

Keep the distinction between a proposal, a live result and business performance in client reporting. The agency can confidently describe verified work without overstating its effect. That precision is part of a repeatable settings-review service.

Agency settings-optimisation questions

Does an audit finding mean every client should use the suggested setting?

No. The finding identifies a defined configuration or heuristic. Review the client's objective, commercial requirements and current account context before deciding whether that recommendation fits. A deliberately retained setting can be a documented review outcome.

Can ppctor apply every setting shown in Google Ads?

Use the current supported-operation catalogue rather than assume universal coverage. Availability depends on the action, target, campaign context and provider requirements. Some audit issues remain manual or need a different supported operation.

Does discovering an operation require reading all campaigns?

No. The capability catalogue is local. Preparing a specific account change then reads the information required by that action. An account-wide operation with no campaign target should not need unrelated campaign discovery.

Can the agency ask for creative ideas before preparing an edit?

Yes. Ordinary explanations, draft copy and brainstorming can remain conversational. Actual account numbers and live settings require account evidence. Asking for an idea is not approval to create or modify an asset.

What if the proposal's settings change before approval?

Execution checks the relevant current state again and can stop. Review the new state and prepare a fresh decision. A pending proposal should not silently override another specialist's intervening work.

Does a verified edit prove that the optimisation improved performance?

No. It confirms the checked execution outcome. Evaluate later business results under the chosen dates, outcome and scope, with the other changes and measurement limitations disclosed.