Ask Google Ads account questions for freelancers: get evidence you can use

Start with the question that matters

A freelance consultant often receives a broad client question in the middle of other work. “What is happening with the account?” may hide a concern about spending, sales, a particular campaign or a recent change. Identify the practical question before opening a large report.

A useful account question asks for an observation that can be supported by the connected data. It might concern the number of enabled campaigns, purchase performance during a completed period or the current targeting of a named campaign.

Keep the intended decision in view. A question about whether to increase spend needs different evidence from a question about which ads are paused. Both can be answered accurately, but the second does not settle the first.

ppctor's conversational tools can retrieve measured data and current settings through defined reporting plans. The freelancer does not need to construct a separate bespoke report for every ordinary question, but the requested meaning still needs to be clear.

When the client's wording leaves a consequential ambiguity, ask a short clarification. It is better to establish whether they mean purchases or all conversions than to provide a confident number with an unintended denominator.

After receiving the answer, check whether it resolves the original concern. An accurate total may be only the first observation. The next question should target the remaining uncertainty rather than retrieve unrelated account detail.

Distinguish account facts from advice and ideas

A conversation can contain several kinds of requests. “How many campaigns are enabled?” requires current account evidence. “Suggest a campaign name” is a creative task. “Explain cost per purchase” can be general guidance without a provider read.

The assistant's rules require account numbers to come from tools in the conversation. They also allow ordinary brainstorming, copy drafts, explanations and general advice to remain conversational. Available data tools are not a reason to call them for every sentence.

This distinction matters for freelancers who move quickly between analysis and creative work. Asking for five headline ideas should not trigger keyword research or introduce unsupported claims about search volume, competition or likely performance.

Use the existing product context for creative suggestions. If an essential fact is missing, clarify it. Do not invent a guarantee, ingredient, delivery promise or commercial offer simply because it would make the draft more persuasive.

If the request includes both ideas and measured evidence, keep the two parts identifiable. Suggestions can be created conversationally, while requested traffic or account metrics require the relevant tools. Do not label the invented wording as a provider finding.

A request to upload or add the suggestion is another category again. It requires a supported edit proposal and the proper confirmation. Creative preparation, factual retrieval and live execution should remain distinct throughout the exchange.

Keep the connected account explicit

A freelancer can work with several clients while each Slack channel has its own connected account. Before interpreting a result, make sure the conversation is taking place in the intended account context. A familiar campaign name does not establish the account by itself.

The assistant's account tools are bound to the channel's account. Ownership of several accounts does not mean one channel should silently query all of them. The connection provides a practical boundary for the conversation.

Use a recognisable client or account label when coordinating work, without copying private identifiers into the client response. The reader needs to know the scope, not reconstruct an account from a technical resource name.

If the account connection is missing or incorrect, resolve that before asking for current metrics. The assistant should not replace the absent evidence with remembered figures from a different client or a previous engagement.

Keep multi-client comparisons outside an accidentally account-specific question. A request about “our campaigns” in one client's channel should not be interpreted as permission to disclose another client's results.

This boundary also helps with later actions. A target resolved in the correct account can be prepared for review with its current context. A conversation that begins in the wrong account creates avoidable risk before any edit button appears.

Resolve the observation period

A performance question needs an observation period. “Which campaign spent the most?” is incomplete when the answer could concern yesterday, a rolling month or the client's entire historical relationship. State the dates or a clear completed-period label.

Keep rolling periods distinct from calendar periods. The last thirty days crosses month boundaries and differs from the previous month. If a comparison matters, retain its actual start and end rather than relying on a conversational shortcut.

Current settings and historical performance are different observations. The account may contain a campaign that is paused now but spent during the selected period. Decide whether the question concerns its present status, its past activity or both.

For today's results, acknowledge the partial day and the relevant time zone. A few hours of spending should not be interpreted against a full prior day without a matched comparison. Completed periods can make routine client questions easier to interpret.

Daily conversion outcomes in the app's business measurements default to conversion date. Interaction-date reporting is a separate explicit request. This can affect how a daily series is interpreted, so preserve the selected date basis in a follow-up.

When the client changes only the period, keep the metric and scope intact. A comparison becomes unreliable if “show the previous week” also changes purchases into all conversions or silently broadens the campaign population.

Define which entities belong in the answer

The population is the set of entities the question concerns. Account-wide campaigns, enabled Search campaigns and one named ad group's keywords are different populations. Specify the narrowest relevant scope without excluding evidence merely because it complicates the answer.

Current status is an important choice. A question about enabled campaigns can be answered from the enabled population. A question about all spending during a past period may need entities that are currently paused but were active then.

A campaign-type restriction also changes the result. Do not generalise a Search-only finding to Performance Max or Shopping. Their structures and available measurements can differ even when the business objective is shared.

For a named asset, retain its parent. An ad group name may recur in several campaigns, and a keyword text may appear in several groups. Parent context distinguishes the intended target without requiring the client to read an identifier.

Use exclusions explicitly when needed. “All campaigns except the brand campaign” defines a different population from “all non-brand searches”. The former is an entity selection; the latter can concern the actual searches and their interpretation.

The assistant's reporting plan applies shared scope to the requested series. A series-specific population can deliberately differ, but that difference should be named. Hidden changes in population are a common reason a follow-up answer appears inconsistent.

Choose a meaningful measurement

A metric should describe the business question rather than simply be available in the provider. Spend answers how much was recorded as advertising cost. Purchase count describes selected purchase actions. Neither one alone establishes whether the client should change a campaign.

ppctor exposes reviewed business measurements alongside provider fields. These definitions specify the conversion population, relevant ratios and source-specific requirements. Use them when asking about conversions, revenue, ROAS or cost per result.

All-primary conversion value includes values outside purchase actions. Purchase revenue is explicitly narrower. A freelancer responding to a sales question should make this selection deliberately instead of treating the two terms as interchangeable.

When performance is unspecified, the app's automatic measurement selects ROAS when the account records conversion value and CPA otherwise. This gives an initial interpretation, but the freelancer should still review whether that metric fits the client's objective.

A ratio needs valid inputs. With no spend, ROAS cannot be calculated normally. With no relevant outcome, cost per result may be undefined. Missing or undefined values should remain visible rather than being converted into a convenient ranking number.

If the client requests a specific provider column, retain that definition. A raw field request and a business measurement request can intentionally differ. The answer should not disguise one as the other merely to simplify the wording.

Ask for a count rather than counting examples

A list preview and a population count serve different purposes. A conversational response may show a limited number of examples for readability. Counting those examples does not establish how many matching entities exist in the selected account scope.

The reporting instructions require a count request for counts. The result should concern the complete defined population, with completeness information retained. This protects the answer from turning a compact preview into an apparently authoritative total.

Ask for the population and condition together. “How many enabled campaigns are there?” is clearer than asking for a list and then trying to infer a count from the visible names. If you also want names, request them as supporting examples.

Check what the counted unit represents. An entity, a relationship entry and a unique product are different units. Repeated appearances can be correct in a relationship count even when a unique-item count would be smaller.

For product targeting, the relevant hierarchy can span campaigns and several asset or ad groups. A count should explain its group scope and inclusion rules. It should not silently collapse distinct product occurrences into a single unique identifier.

If the result is incomplete, narrow the question or resolve the retrieval limitation before presenting an exact account-wide claim. A known subset can still support useful examples, but it should not be described as the complete total.

Give a ranking a fair definition

“Best” and “worst” need a measurement, a period and an eligible population. A high-spend campaign may have strong results, while a low-volume campaign can show an unstable ratio. Start by defining the intended comparison.

Use the client's goal to choose the metric. A purchase-focused review can rank purchase return or cost per purchase, while an all-primary conversion review uses a broader outcome. State which one the result represents.

Some business rankings require a relevant outcome above zero. This avoids treating an undefined ratio as a normal value. If campaigns without outcomes are important, inspect them as a separately defined group rather than assuming they appear in the ratio ranking.

Look at the underlying volume before making a consequential recommendation. A ratio based on one purchase can be mathematically correct while offering a limited basis for a major budget decision. The retrieved count helps explain that limitation.

Do not treat rank as a complete diagnosis. The lowest selected ROAS identifies a place for review. It does not establish a cause, quantify profit or prove that pausing the campaign is the best business choice.

If the client follows with “pause it”, resolve the target from the defined result and prepare the supported proposal. The ranking answer should not itself execute the change. The confirmation must show the exact selected campaign.

State thresholds precisely

Threshold wording changes which rows qualify. “Above 3” excludes a value equal to 3, while “at least 3” includes it. A freelance consultant should preserve that distinction when answering a client or preparing a later selection.

The shared condition framework includes equality, inequality, greater-than and less-than operations, including inclusive forms. It normalises common expressions such as at least and at most. Provider-specific compatibility still determines where a particular condition can be used.

State the unit as well as the threshold. A monetary value, a percentage and a ROAS multiple require different interpretations. “Over 300” does not identify the intended measure without that context.

Keep null values separate from zero. A missing performance value should not satisfy a zero-result condition unless the data and request genuinely define it that way. Absence of a row can have a different meaning from a returned zero.

For a range, specify both bounds in a sensible order and confirm whether the requested meaning includes the boundaries. Avoid translating conversational shorthand into an arbitrary limit that the client did not choose.

Read the final selection back when it can lead to an edit. A precise filter can still produce unintended targets if the population is wrong. The metric condition, period and parent scope together determine what the client is reviewing.

Combine conditions without changing their meaning

Several conditions can define a useful investigation. “Campaigns with cost above 200 and no primary purchases” requires both conditions. Changing and to or would produce a much broader selection and a different answer.

The condition syntax supports grouped AND, OR and NOT logic. Keep the grouping clear when ordinary language could be read in more than one way. A short clarification is appropriate when the intended selection affects a live decision.

For example, a client may want enabled campaigns that meet either a cost threshold or a particular name condition. The status restriction should apply to the whole group if that is the request. Parentheses in the reasoning prevent one branch from escaping the shared scope.

A text condition can also depend on case and matching behaviour. Contains, begins-with and exact equality do not select the same names. Ask for the recognisable target or clarify the pattern instead of treating every similar name as interchangeable.

Keep the logical explanation readable in the client answer. The user needs to understand the included set, not inspect a technical expression. “Enabled campaigns meeting both conditions” can communicate the selection while retaining exact supporting evidence.

Supported syntax does not mean every operator works on every provider field. Use the field catalogue and structured validation response. An invalid combination needs a corrected plan, not a claim that the provider has no relevant data.

Retain the parent of a deeper target

Advertising assets exist within a hierarchy. A campaign contains groups, and a group can contain ads or keyword criteria. A question about a deeper entity should retain enough parent context to distinguish it from similarly named entities elsewhere.

Use the correct structure for the campaign type. Performance Max uses asset groups, while Standard Shopping uses ad groups. Their product-listing selections can form trees, and one campaign can have several groups with different selections.

A product can appear in several groups. A question about excluded product occurrences can legitimately count it more than once when the requested unit is each group-specific exclusion. A unique-product question asks for a different aggregation.

For a keyword, name its ad group and campaign in the relevant answer. The same keyword text can exist in several contexts and match types. Showing only the text may leave the freelancer unable to identify the intended entity.

Parent context should be meaningful rather than mechanically repeated. An ad group's parent is its campaign, not another ad group. A missing name should lead to an unresolved-context explanation instead of a fabricated parent label.

This matters especially when a read leads to an edit. The proposal must preserve the exact target resolved by the question. A clear hierarchy allows the connected owner to review the intended change without seeing internal asset identifiers.

Separate inventory from performance

An inventory question asks what exists or how it is configured. A performance question asks what activity was recorded during a period. The two can be combined, but neither automatically supplies the other's answer.

A keyword may exist without a returned performance row for the chosen dates. That absence should not be treated as proof that the keyword does not exist. Conversely, historical activity can remain relevant after the keyword has been paused.

Current status and effective delivery are also different. An enabled entity can still face other eligibility or parent constraints. A current status answer should not imply that the entity is spending at this moment unless the requested evidence establishes that activity.

For product targeting, listing-group entries describe the selected hierarchy. Merchant inventory describes products in the feed, and performance describes observed activity. A question about excluded products may require relating these surfaces rather than counting a generic product list.

Ask which observation the client needs. “Does this ad group contain an RSA?” concerns inventory. “Did this RSA record purchases last week?” concerns performance. “Can we create an RSA here?” introduces supported creation requirements and approval.

Keeping these questions separate makes the next step clearer. The freelancer can resolve the existing target, retrieve relevant activity and then prepare a specific proposal without confusing a lack of performance with a missing asset.

Ask what the search-term evidence covers

Search terms and keywords answer different questions. A keyword is an advertiser's targeting criterion, while a search term is an actual query associated with the provider's reported activity. Asking what people searched for should not return a keyword inventory under a misleading label.

Google's search-terms guidance explains that some queries with insufficient activity are omitted for privacy. The available report therefore does not expose every individual search that occurred.

Use the available search-term evidence to investigate relevance within its stated scope. A suspicious phrase can be a useful starting point, but a compact list of examples is not a complete account-wide waste estimate.

Keep the matched keyword and reported search-term match type distinct. Their relationship helps explain what happened, but a label on the search-term row should not automatically be treated as the configured keyword match type.

Interpret the term with the client's actual offering. An apparently unrelated query may be relevant to a product or service the freelancer has not yet learned about. Obtain that fact before recommending an exclusion.

If the review leads to a negative-keyword proposal, specify the exact supported target and match behaviour. A term's appearance in the report does not itself authorise a change. The connected owner still needs to review the proposed exclusion.

Identify the conversion population

A conversion question needs a defined action population. ppctor's all-primary measurement includes all selected primary conversion actions. Purchase, lead and call measurements select their relevant groups, while a named-conversion request selects one action by its name.

Secondary actions are excluded by default in these business measurements. Include them only when the user asks for that population or a secondary action. A changed selection should be visible in the answer rather than silently widening the total.

Primary actions are added together without a deduplication repair. The requested report describes configured account measurements. It does not attempt to decide that two separately recorded actions were really one business transaction and remove one.

Offline conversions can be listed and selected where the account exposes uploaded or imported actions. They can overlap purchase, lead or call categories. Do not add overlapping groups together as though each were an independent pool of outcomes.

For a named action, confirm ambiguity when similar names exist. A campaign-level performance question is not a reason to invent the conversion definition. The assistant can read action metadata to identify the available selection.

Keep value distinct from count. A primary call can carry conversion value, and a non-purchase primary action can contribute to all-primary value. Ask for purchase revenue when the client's question specifically concerns purchase actions.

Confirm a cross-source event comparison

A Google Ads conversion action and a GA4 event can be compared when the request identifies the corresponding measurements and the workflow supports the composition. The freelancer should not assume that a familiar event name proves an equivalent business action.

For leads, the user must name or confirm the GA4 event. The assistant can list bounded observed event names and suggest a plausible candidate. A suggestion alone is not the user's selection and should not be treated as confirmation.

All-primary Google Ads conversions and GA4 key events form a broader aggregate comparison. Each retains its source-specific definition. The result is a comparison of reported measurements, not a reconciled count of unique business outcomes.

Check traffic scope for purchase revenue. The app's GA4 purchase measurement defaults to all traffic unless Google Ads traffic is explicitly requested. An all-site series should remain labelled as such when shown beside advertising conversion value.

The current named-comparison release gate remains separate from available metric definitions. Do not promise that a selected action/event composition is deliverable merely because both entries exist in the catalogue. Respect the current validated capability.

If the requested comparison cannot be delivered, explain the specific missing selection or capability. Avoid saying that the provider lacks daily data when the actual problem is an invalid plan or a gated composition in the app.

Work through an invented freelancer exchange

The following exchange is invented and contains no client data. A freelancer working with Willow Lighting wants to investigate purchase performance before a client review. The connected account contains three illustrative enabled campaigns in the selected population.

The first request is to rank those campaigns by purchase ROAS for a completed seven-day period. In the example, Lamps records cost 300 and purchase value 1,200, Shades records cost 200 and value 500, and Fixtures records cost 100 and value 500.

The corresponding purchase ROAS values are 4.00, 2.50 and 5.00. Shades has the lowest selected return. This is a mathematical ranking for the defined population, not proof that the campaign is unprofitable or should be paused.

The freelancer then asks which campaigns have purchase ROAS at most 3.00 and cost at least 150 during the same period. Shades qualifies. Both inclusive conditions apply, and the period and enabled population remain unchanged.

A useful follow-up requests the purchase count and the client's relevant offer context for Shades. The count requires account evidence; the offer context requires a confirmed business fact. Neither should be fabricated to make the original ranking sound explanatory.

If the client ultimately wants to pause Shades, the assistant prepares a supported campaign-status proposal. The exact campaign and action appear for confirmation. The read, interpretation and live instruction remain separate stages of the exchange.

Change one part of a follow-up deliberately

Follow-up questions often use shorthand such as “show it for fourteen days” or “put that in a table”. Resolve what changed against the preceding plan. The assistant's instructions preserve scope and series populations unless the user changes them.

A period change should retain the metric and target. A format change should retain the evidence definition. A new filter should apply to the intended population, with the rest of the request remaining recognisable.

If the client refers to “that campaign”, the prior result needs to identify one clear target. When several campaigns were listed, clarify which one they mean before preparing an action. Do not choose an arbitrary visible example.

The reporting tool can reuse returned data within the request when only the presentation changes. A different period, grain or measurement may require another read. Distinguish efficient reuse from pretending that old evidence answers a new question.

If a stored result is unavailable, supply the report plan again rather than inventing its numbers. Thread context can preserve meaning, but it is not a substitute for current provider evidence when the request requires a fresh observation.

Read the follow-up answer for unintended drift. The metric labels, period and population should still match the user's changed request. This simple review is especially valuable before forwarding the result to a client or turning it into a proposal.

Interpret missing and incomplete results

A failed read, an empty result and an undefined metric are different outcomes. A failed read means the requested evidence was not obtained. An empty result means the valid query returned no matching rows. An undefined ratio means its inputs do not permit the calculation.

Do not translate all three into “nothing happened”. The freelancer needs the actual outcome to decide whether to correct the request, check access, broaden a population or investigate a valid absence of activity.

The reporting tools return structured errors and completeness information. A plan error should be corrected according to the reported selection problem. Repeating an unchanged non-retryable request does not improve the evidence.

A compact preview can omit rows while preserving a full supported result behind its data reference. Conversely, an incomplete selection cannot safely authorise an edit of only the visible rows. Keep preview and completeness distinct.

When a returned area is unread, state which conclusion it prevents. Other successfully obtained evidence can still be useful, but it should not become a claim that the entire account is clear or fully checked.

Give the client a specific next step. “I could not retrieve the selected action; please confirm its name” is more actionable than a generic apology. A limitation should lead to a defined clarification or investigation rather than an invented account fact.

Use settings discovery for a settings question

A question about current settings needs a settings read, not an inference from performance. A campaign recording clicks does not establish its search-partner setting, and a geographic report does not by itself describe its configured targeting option.

The assistant can discover settings through a local catalogue that describes available fields and entity levels. That discovery does not make a provider request or prove how the connected account is configured.

The actual current-value question then uses the relevant read. Keep capability discovery and account observation separate in the answer. “This setting is supported” is different from “this campaign currently uses this value”.

Select the correct level. An account-wide setting should not require an invented campaign target. A campaign setting needs the campaign, and an ad-group setting needs its relevant parent context.

When the client asks whether a setting should change, separate the retrieved value from the recommendation. A current option can be valid in one business context and inappropriate in another. The catalogue does not supply the client's business reasoning.

For a live change, use the supported action and its target rules. Some visible settings may not have an available safe write workflow. Explain the actual boundary rather than implying that every readable field can be edited from Slack.

Ask and review in the Slack thread

A focused Slack thread gives the freelancer a place to retrieve evidence and discuss the next question. Use the channel connected to the intended account, then phrase the request in the client's business terms with enough scope to avoid ambiguity.

A first request can be concise: ask for the number of enabled campaigns or the purchase results of a named campaign during a completed window. Add conditions only when they matter to the investigation.

Review the answer's labels before accepting the conclusion. The returned measurement, period and target should match the request. A fluent explanation cannot compensate for a different population or an unintended conversion definition.

Keep clarifications in the same thread when they refine that investigation. This allows the follow-up to refer to the established plan. Introduce a distinct target or business question explicitly when the conversation changes direction.

Use creative assistance when the task becomes wording rather than measurement. The assistant can draft a short client explanation from the existing evidence without running a fresh provider query solely to change the tone.

Before requesting a write, restate the intended target if the conversation has become long or ambiguous. The confirmation proposal should show the same campaign or deeper asset that the freelancer intends to change.

Request a useful supporting visual

A chart or table should help inspect the result, not merely decorate it. Ask for the format when the question benefits from one. The assistant's instructions reserve visual rendering for requested visuals rather than attaching charts to every answer.

A line chart can show a time trend, while a bar chart can compare categories. A table can make exact values and several conditions easier to inspect. The requested format takes precedence when the workflow can support it.

The renderer uses returned data references rather than user-invented numeric arrays. This connects the visual to the evidence plan. Reformatting the same result should retain its definitions and scope.

Native Slack charts in this workflow allow up to twenty points per series and up to three series. A longer automatic period can use a coarser grain to keep one readable chart. An explicit daily request needs the daily-detail constraint explained.

Use compatible units and clearly named series. A shared revenue comparison should retain source labels and should not add attributed values together. An unrelated metric may need another presentation rather than a confusing single scale.

When evidence is incomplete or bounded, a table can be more honest than a fabricated continuous trend. Missing days should not become invented zero values merely to make the graph look complete. Keep any relevant coverage note visible.

Move from an answer to a reviewed action

An account answer can lead to a live-edit request, but it does not authorise one automatically. The assistant resolves the intended target and prepares the supported operation for review through the approval-based edit flow.

Read the exact action and parent context. A campaign pause differs from an ad-group pause. Adding keywords differs from enabling the group that contains them. Multi-step requests should make each intended logical action visible.

The connected authorised owner must click Confirm and Execute. A proposal expires after fifteen minutes, and current settings are checked again before execution. A conversational statement of intent does not remove these controls.

During execution, distinguish progress from completion. The returned outcome can confirm verification, describe acceptance without complete verification or report uncertainty. Use that exact status when explaining what happened to the client.

An uncertain write may already have changed the account. Do not request the same creation again merely because the final verification failed. Inspect the relevant target and outcome before deciding the appropriate follow-up.

A later performance question is a separate review. Verification of an edit confirms its account result, not its commercial effect. Choose a relevant period and consistent measurement before evaluating whether the subsequent activity supports the intended objective.

Turn the answer into a client response

A client response should state the observation, its meaning and the next step. Keep the evidence definition close to the conclusion. The client should not need to inspect the entire thread to discover which period or conversion population was used.

Use readable target names with parent context where needed. Remove internal resource identifiers from the narrative. The purpose is to help the client recognise the affected part of the account and understand the recommendation.

Preserve a consequential limitation. If a ratio excludes campaigns without qualifying outcomes, explain that when presenting a ranking. If a read failed, state which question remains unanswered instead of presenting a complete account judgement.

Separate an observed difference from its cause. The answer can show that cost rose or purchase value fell. The reason may need business context, another defined read or website investigation. Avoid making the response sound more conclusive than the evidence.

State the required client response precisely. They may need to confirm an event, identify a campaign, supply a business fact or review a supported proposal. A specific request makes the conversation easier to advance.

The freelancer's contribution is the judgement connecting a clear observation to a useful decision. Efficient retrieval helps, but the quality of the question and the honesty of the interpretation determine whether the answer is useful to the client.

Freelancer account-question answers

Do all questions need a Google Ads read?

No. Current account facts and measured results need evidence. Ordinary explanations, brainstorming and copy drafts can remain conversational. Do not attach provider claims to creative suggestions unless the relevant research was requested and obtained.

Can I count the campaigns shown in a preview?

A preview may be capped. Ask for the count of the defined population and retain completeness information. The visible examples help identify entities but do not necessarily establish the account-wide total.

Does at least include the boundary?

Yes. At least is an inclusive condition, while above is strict. The shared condition framework preserves that distinction, subject to field and provider compatibility. State the metric and unit with the threshold.

Can the answer identify one named conversion?

The measurement catalogue supports selecting an action by name, with ambiguity resolved as needed. A GA4 comparison requires the corresponding event to be named or confirmed and the current composition capability to be available.

Does the lowest ROAS campaign need to be paused?

A ranking identifies a defined result for review. It does not prove the best business action. Consider the objective, volume and relevant context, then prepare a supported proposal only when the user requests the change.

What makes an account question reliable?

A clear account, period, population and measurement make the result interpretable. Accurate conditions, complete counts, meaningful parent context and honest limitations keep follow-ups from drifting into a different question.