Google Ads reporting for agencies: build a clear client decision from defined evidence

Start with the agency's reporting constraint

An agency's reporting problem is partly a definition problem. Different clients can use the same word for different outcomes, and different colleagues can inherit a report without its original context. A readable table is useful only when its account, period and selected measures remain clear enough for the next person to interpret.

Consider a fictional agency, Cedar Lane Digital, managing a parts retailer and a commercial installer. Both ask about conversions. The retailer may mean purchases, while the installer means submitted specification requests. An identical report heading does not make those outcomes comparable. Keep the client-specific definition before trying to standardise the presentation.

The agency needs consistency in the reporting method rather than pretending that every client has the same commercial model. Identify the source, scope, period and outcome, retrieve the supported evidence and explain the decision it informs. The framework can stay consistent while the selected business definition changes.

This approach also makes handoffs less expensive. Another colleague can read the selected measures and ask a focused follow-up without rebuilding the whole account investigation. The goal is a reproducible client decision, not a large collection of charts whose labels hide incompatible populations.

Choose the client decision

Start with the question the report should answer. An overview of spending, a campaign comparison and a review of one contact action need different selections. “How did the account perform?” is broad; “Which enabled campaigns spent money without the agreed contact action?” identifies a more precise population and condition.

Define the next step the evidence could support. The report might help prioritise an investigation, explain a period's movement or review a configured budget. It does not automatically establish a cause or authorise a live change. A commercial decision can need business context beyond the connected advertising and Analytics sources.

Keep owner-supplied context distinct from returned metrics. A client may explain that an offer closed or a contact route changed. That information helps interpretation but should not be labelled an API-derived finding. State its scope and date rather than turning a qualitative explanation into an invented numerical adjustment.

Choose the smallest useful report for the decision. A status question does not need every daily performance series. A trend question needs the requested segmented values, not only a total. Let the evidence requirements determine the read rather than requesting every available dataset because the tool can access it.

Confirm the account and campaign scope

Confirm the channel's connected account before interpreting a report. An agency can have similar campaign names across clients, so a name alone should not replace account context. The assistant should resolve the target under the actual connection rather than infer the client from a familiar phrase.

Account scope and campaign scope answer different questions. An account total can provide an overview, while a named campaign read supports a narrower investigation. An enabled-only population is different from every campaign with activity in a period. Status selection and dated performance need their own explicit meaning.

For an entity comparison, state the included population and important exclusions. A top-five preview is useful for examples but is not the entire account. A count question needs the complete matching-count flow. Do not identify the account's total number of poor-performing campaigns by counting a capped list.

Keep names and relevant parents in the explanation when they identify the target. Internal technical identifiers are usually unnecessary for a client decision. The agency should be able to understand which campaign the report concerns without deciphering a provider resource name.

State the reporting period

A rolling thirty-day window, last month and the current month through yesterday are different periods. Name the actual start and end of the selected range. A client who asks for recent complete days should not receive last month's dates merely because that was the dashboard's previous selection.

Keep time-zone and date-basis context when it affects the question. Current daily conversion outcomes use conversion dates by default; interaction dates are selected explicitly. A comparison of later outcomes and advertising interactions should not silently assume the same timing basis.

A follow-up asking for a chart should retain the established period. A change from daily to weekly can alter presentation while preserving the window; switching to another month changes the question. Make that change explicit if it is requested rather than letting the renderer choose it without explanation.

Record important offer or tracking changes alongside the period. A chart can locate movement, while a dated operational note can explain a known change. Neither should replace the other. The reporting source cannot infer every promotion, closure or implementation decision from account totals alone.

Define the selected outcomes

Ask what the client means by the reported outcome. A purchase, information request, call conversion and submitted application represent different observations. A conversion action's name can help locate it, but its trigger determines the business meaning. Confirm that definition before labelling a count customers or revenue-generating sales.

Use a broader primary aggregate for an overview when that is the request. Use a named action when the decision concerns one specific outcome. The reporting framework supports selected business measurements and action names rather than requiring an agency-only metric that guesses the client's commercial process.

Keep the definition in the report and follow-up. A request to show the same data as a graph should not replace purchases with all primary conversions. A request to add another measure should keep the original series definition. Consistency means preserving the selected population, not repeating the same vague heading.

Resolve ambiguity before presenting a definitive result. If similar contact names exist, show the relevant choices. If a GA4 event correspondence is missing, ask for it. A brief selection question is more useful than a polished report whose central outcome silently changed.

Read primary totals accurately

The Google Ads Conversions business measure selects all primary actions by default. The Leads measure combines supported primary lead categories without inventing separate lead-quality stages. Purchases and Calls have their defined populations, while a named action can narrow the question further.

Several primary actions can contribute to one total. That aggregate reports the configured actions; it does not automatically deduplicate people, orders or companies across separate actions. The tool should not repair a tracking history by guessing which records overlap. The agency can review implementation separately when needed.

Secondary actions remain outside the ordinary primary aggregate unless the user explicitly requests them through supported selection. A difference between a primary total and a wider provider count can therefore be expected. Inspect the selected population before claiming that a report lost conversions.

Imported outcomes can also overlap lead, call or purchase categories. Do not add an offline group to those totals without checking the intended population. Labels that describe the same action in different ways are not automatically independent counts. A correct aggregate starts with a coherent selection.

Choose the revenue definition

Conversion value and purchase revenue are different selections. The broader primary-value measure includes values from all selected primary actions, including non-purchases. Purchase revenue selects primary purchase-category values for the Ads side. An assigned contact value can therefore contribute to one total without becoming recorded purchase revenue.

Ask what the implementation's value means when the client uses a commercial label. A payment, deposit, expected contract amount and margin are different quantities. The advertising connection does not establish their accounting treatment. Retain the source definition and the owner's explanation rather than inventing a profit claim.

Choose the matching return ratio. Primary ROAS uses all primary conversion value against cost. Purchase ROAS uses purchase value. All-traffic GA4 purchase revenue divided by an agreed advertising-spend source is blended ROAS or MER. The assistant should ask for the spend source when the all-traffic question leaves it unspecified.

A contact-led client may not have recorded purchase revenue. In that case, cost per the agreed outcome can answer the question without manufacturing a monetary result. The agency report should preserve missing evidence rather than assign a speculative revenue amount to every request.

Compare Ads and Analytics deliberately

A comparison can place related Ads and Analytics measurements in one requested visual while keeping their definitions distinct. Different attribution is allowed; it is not an automatic reason to refuse a chart. The assistant should choose compatible measurements from the framework and label each source clearly.

Do not add the two sources together as if they were independent sales. Ads-attributed purchase value and GA4 purchase revenue can concern overlapping business activity under different measurement scopes. A shared time axis helps comparison, but it does not establish that the lines should have identical totals.

Keep traffic scope explicit on the Analytics side. All-traffic purchase revenue is not silently narrowed to the linked advertising account. If the user requests Google Ads-attributed Analytics activity, select that supported scope deliberately. The client should know what population each series describes.

Cost definitions also need their source. The advertising cost available in Analytics may have a different scope from the selected Ads account. A ratio should retain the agreed denominator. The agency's job is to make these choices understandable, not to hide the source difference under a generic revenue or spend heading.

Confirm a named-event comparison

A named Ads conversion can be compared with a specific GA4 event. For a lead question, the event must be named or confirmed by the user. The assistant can list available observed events and suggest a probable match, but a suggested name alone is not confirmation of the correspondence.

Ask what each trigger observes. An application success event and a form-start event are different steps. A phone-link click is not automatically a qualifying call. The source names are useful locators, while the implementation owner explains the business meaning.

Retain period and traffic scope when the comparison is rendered. A difference in values can prompt a measurement investigation without proving an error. The report should state the two populations before recommending further work. A broader key-event total is not an unrequested substitute for the selected event.

The aggregate Conversions comparison is a separate choice: all primary Ads conversions against all GA4 key events. It can answer an overview question, while a named-action comparison answers a narrower one. Preserve this distinction when a colleague inherits the report.

Aggregate ratios from components

A combined cost-per-outcome ratio should use combined cost and the selected outcome count. CTR should use selected clicks and impressions. A simple average of campaign ratios can misstate the total because campaigns contribute different volumes. Retain the components so the calculation matches the defined scope.

For a client comparison, keep currencies and units clear. A count and a currency value are not interchangeable. A report should not manufacture a summed series from incompatible units merely because the visual can display several lines. Choose labels and presentation that preserve the actual measures.

Undefined ratios need an explanation. Cost with no selected conversions does not yield a finite cost per conversion. No spending means a return ratio has no valid denominator. An unavailable source is different from either state. Do not fill these cells with zeros that imply a measured result.

A concise table can show the relevant components and ratio with sensible decimal formatting. The agency does not need dozens of technical columns to make the arithmetic credible. The selected population, period and definitions supply the important context for the decision.

Preserve coverage and unavailable evidence

Coverage says what the report actually examined. A complete matching count and a capped example list serve different jobs. A top-campaign preview can illustrate the result, while a count query establishes how many entities meet the selected condition. Do not use one as proof of the other.

Missing data must remain distinct from zero. A campaign whose required outcome could not be read should not satisfy a zero-conversion condition by default. Preserve the unavailable population and explain what selection or read is needed next. This prevents a polished report from making a false claim of complete coverage.

Partial provider evidence can still be useful when its scope is clear. An inspected surface or returned subset can support a bounded observation. It should not become a statement about every asset or every period without evidence. The agency can decide whether the missing coverage matters to the client question.

Keep structured tool failures attached to the request. A rejected report plan is not automatically a provider limitation. Resolve the missing selection or incompatible field through the available catalogue, then make the supported read. Repeating an unchanged invalid request adds delay without establishing new evidence.

Explain a later updated value

Google's data freshness guidance explains that statistics are not instant and that some values can be adjusted later. A report is an observation from available data at a stated time, not a guarantee that the newest period has received every outcome.

When a client sees another number later, first compare definitions. Check the account, scope, period, action selection and date basis. If those match, consider the fact that the later read can include updated data. Do not jump straight to an explanation about processing when the two reports actually selected different populations.

Preserve the original report's context instead of declaring it fabricated because a newer read changed. The agency should be able to explain what it reported and when. A fresh report can update the decision, while the earlier observation remains part of the review history under its original definition.

Keep this explanation proportionate. Freshness does not excuse an incorrect query or a wrong date range. A known reporting error needs correction. A legitimate later update needs a clear note about the changed observation. The two situations require different actions even when the client's first question is simply why the total moved.

Work through an invented agency report

The following names and figures are invented for illustration, not a customer result. Suppose Cedar Lane Digital compares two fictional campaigns for a parts retailer over the same complete period. The selected outcome is primary purchases, and the table preserves cost and purchase count before calculating cost per purchase.

The combined cost per purchase is 2,000.00 divided by sixty purchases, or 33.33. Averaging the two campaign ratios would give 35.00 and would answer the aggregation incorrectly. The components establish the intended combined measure without needing a large technical export.

Now consider the installer's report. Its selected outcome is a submitted specification request rather than a purchase. The agency can use the same reporting method but should label the different outcome. It must not compare the installer's cost per request directly with the retailer's cost per purchase as if both were a universal customer-acquisition measure.

A useful next question might ask for the retailer's purchase revenue, or for the installer's named request beside a confirmed Analytics event. Each requires its own explicit selection. The example shows why a reusable framework is valuable while a single vague conversion heading is not enough.

Use a table for the decision

A table is useful when the client needs to compare entities or components. Choose columns that support the question: campaign, selected spending, selected outcome and the relevant ratio. Avoid adding every available field merely because it exists. A concise table with a clear definition can be more useful than an export-shaped wall of data.

Keep names and units readable. Currency should identify the money measure, and counts should retain their actual outcome labels. Metrics should not carry unnecessary decimal precision. Undefined ratios and unavailable values need explicit states rather than zeros that imply a measured result.

State whether the rows are the complete matching population or examples. A top list can help identify where to investigate, but it is not proof of the total number of matching entities. When the client asks for a count, retrieve the complete count under the selected condition.

A table can also expose the selected definition. Showing cost and outcome components makes a ratio assessable. Keep the period and source nearby so another colleague can reproduce the selection. The table's layout should make the decision clearer without adding internal account identifiers to a client-facing explanation.

Choose a readable requested chart

A line chart can show movement over time, while a bar chart can compare categories or selected time points. Honour the requested format where the supported renderer permits it. The assistant should render returned data rather than supply imagined arrays to create a visually plausible story.

Keep multiple series within the supported readable limit and preserve their definitions. A combined Ads and Analytics purchase-revenue chart can retain distinct source labels and scope. A count and a currency measure need clear units. Sharing a chart does not justify summing overlapping sources or calling unlike outcomes the same thing.

Native Slack charts have bounded points. When daily detail was not required, a longer window can use a coarser grain to remain in one readable chart. Explicit daily detail can need consecutive panels with the point limit explained. Preserve the actual window; do not replace it with last month to simplify rendering.

A total does not supply daily points. If the available result lacks the requested segmentation, make the supported segmented read. If the data already contains the needed points, reuse it. The agency should not wait for unnecessary duplicate reads merely because a follow-up changes the presentation.

Ask reporting questions in Slack

Connect the channel to the intended account and ask for the period, scope and outcome the client decision needs. “Compare purchase cost and purchase revenue for these enabled campaigns over the last fourteen complete days” gives the assistant a much clearer selection than “Send a performance report.”

The assistant can inspect available fields and measurement definitions before preparing a supported read. A named action or confirmed event can narrow the question. The account evidence must come from returned tools rather than a guessed number based on the conversation's tone or general industry advice.

Ordinary explanations and creative ideas can remain conversational. Asking what a metric means does not always require a provider read. Asking for a headline suggestion does not require Keyword Planner merely because it is available. The assistant should distinguish supplied-context advice from measured account questions.

Use the Slack reporting guide for the complementary workflow. A normal conversation supports one-time questions and requested visuals; saved reporting supports deliberate recurring delivery. The agency should choose the workflow the user actually asked for rather than create unsolicited schedules.

Preserve definitions in follow-ups

A follow-up often changes only one part of the report. “Show it weekly” changes grain; “add primary calls” adds a selected measure; “use this event” confirms a correspondence. Preserve the other parts unless the user changes them. This makes conversational reporting useful instead of forcing a new setup on every message.

Check whether the returned data contains the new requirement. A chart of the same points can reuse the data. A new action, dimension or period can need another supported read. The assistant should know the difference and avoid both invented values and unnecessary provider requests.

Conditions need exact meaning. Above differs from at or above; below differs from at or below. AND and OR produce different populations. A query for entities matching a condition must retain the selected fields and missing-value handling. A missing outcome cannot become zero merely to satisfy a filter.

Explain a changed definition when it matters to the client. If the follow-up narrows a broad primary total to one action, say that it is now a specific outcome. The new report can be useful without being presented as an identical reproduction of the old population.

Configure a recurring report

The current saved Reporting workflow supports Ads and Analytics reports with deliberate configuration. Choose the source, supported target, metrics, range and delivery schedule. Daily, weekly and monthly frequency are different choices from the reporting window itself.

The available saved ranges include yesterday, rolling seven, fourteen and thirty days, the current month through yesterday and last month. Choose the range that fits the client review. An agency should not label a current-period report last month or assume that every monthly schedule automatically means the previous calendar month.

Use the configured time-zone and receiving account context. The Slack workflow supports report management and Run Now for an enabled saved report. A paused report needs to be resumed before that queued run. Queued does not mean that delivery has already completed or that every upstream read succeeded.

Review what was saved, especially after changing the source or target. A recurring overview should retain the client-specific metric definitions and important limitations. It is a reporting configuration, not automatic authority to change budgets, status or creative when a number moves.

Choose the comparison period

The saved Reporting model supports no comparison, a previous-period comparison and year-over-year comparison. Each serves a different question. A previous period can show recent movement; year-over-year can add seasonal context. Neither automatically explains why a campaign changed.

Keep both periods and definitions visible. A comparison of different lengths or changed outcome triggers needs context. If an offer closed or measurement changed, record that fact rather than treating every percentage movement as directly comparable performance.

Ask whether the comparison matches the client's decision. A seasonal retailer may find a prior-year view useful; a newly launched service may have no comparable earlier evidence. Do not invent a baseline when the source does not provide one. An unavailable comparison can still leave the current-period report useful.

Retain the scope when comparing. A report restricted to currently enabled campaigns may differ from a report of all campaigns with activity in both periods. The agency should understand that population choice before claiming account-wide growth or decline. The framework's explicit selections make that distinction assessable.

Write the client explanation

Lead with the observed result and the decision it informs. Name the period and selected outcome in ordinary language. The client should understand whether the number concerns purchases, contact requests or all primary conversions without needing to decode provider column names.

Then explain the relevant limitation or context. A source difference, changed trigger or unavailable population can materially affect the conclusion. Keep this explanation concise but concrete. Do not announce a cause that the evidence did not establish or bury the result in technical narration.

Separate observations from recommendations. The report may identify a campaign worth investigating. A recommendation needs the offer and operating context, and a live change needs its own reviewed action. These steps can appear in the same discussion without becoming an automatic execution chain.

Use supplied client context accurately. If the owner says a product was unavailable, attach that fact to its actual scope and date. Do not claim the reporting tool verified every inventory state. A clear explanation shows which facts came from tools and which came from the business.

Transfer the report to another colleague

Resolve two reports that appear to disagree

Suppose two colleagues bring different totals to a client meeting. One report uses all primary conversions for the account; the other selects a named contact action for enabled campaigns. Start by writing those selections side by side. The disagreement may be a population difference rather than an arithmetic error. A new broad provider read is not the first requirement when the existing definitions already explain the difference.

Then compare the periods and date bases. One report may use a rolling window while the other uses last month. One may use conversion dates while another explicitly uses interaction dates. These choices can be legitimate individually, yet they cannot be described as the same report. Correct the labels and choose the definition the client decision actually needs.

Next compare the source and traffic scope. An Analytics all-traffic event total and an Ads-attributed conversion total describe different populations. Do not force equality by narrowing one series without the client's request. The appropriate result may be a clearly labelled comparison, with the source difference preserved rather than erased.

If the selections really match, inspect the read timing and returned components. A later update can explain changed values, while a query or calculation defect needs a correction. Keep the investigation evidence-based: identify the specific difference rather than suggesting several possible causes in place of checking the actual report definitions.

Close the issue with a shared definition record. State which report answers the meeting's question, what the other report measures and whether any defect was corrected. The client should not be left with a vague instruction to trust one number. A concrete explanation lets the agency reuse the correct selection in future reports.

Handle a client question that expands the evidence boundary

A client may see a contact report and ask how many of those requests became contracts. That is a reasonable next question, but the existing contact count cannot answer it by inference. Check whether a corresponding later action is available in the connected reporting source. If it is, select and define it deliberately.

If the later evidence is not available, explain exactly what the current report establishes: the selected recorded contacts under its period and scope. Ask the commercial owner for the missing business context or implementation definition. Do not invent a conversion rate from contacts to contracts to make the response sound complete.

Keep supplied commercial information distinct from retrieved account evidence. The owner may provide a later outcome total for discussion, but that does not mean the tool inspected the underlying records or matched them to every contact. State its source and scope rather than presenting it as a verified joined dataset.

The next reporting task should follow the established evidence. A supported named-action read can extend the report; an external operational task can clarify missing outcomes. The agency maintains the decision's meaning by identifying this boundary, rather than hiding it behind a broad heading such as pipeline performance.

A reporting handoff should contain the selection, observation and next task. Name the account context, scope, period, selected measures and important source limitations. Another colleague should be able to ask a focused follow-up without reconstructing the client's whole setup.

Hand over a changed contact definition

Suppose a fictional service client replaces one general contact form with separate installation and maintenance routes. The agency's old report selected the general contact action. The next review needs the intended new population, not a silent choice of whichever action has the most familiar name.

The implementation owner confirms which events now fire and whether the old action remains. The report owner decides which selection answers the client question. If the desired total uses both new actions, confirm that the population is coherent and does not overlap a retained general action. Do not repair the definition by guessing which counts represent unique people.

Record the implementation date and the selected measure before comparing periods. A before-and-after chart can remain useful, but a changed trigger may limit its interpretation. A movement in counts cannot be called a performance improvement solely because the new definition records more events.

Keep the work record scoped and free of private contact narratives. The relevant facts are the route, trigger, selected action, period and owner. State any unresolved implementation question. A broad “tracking updated” note does not give the next colleague enough evidence to reproduce the report.

Carry a revenue definition into the next review

If the client changes from assigned request value to a later recorded commercial value, preserve that distinction. The report should say which measure it selects and what the owner means by the value. Do not treat the new amount as a continuous purchase-revenue series unless that definition is actually established.

Review the ratio's denominator at the same time. All-traffic Analytics purchase revenue with selected advertising spend is blended ROAS or MER, while an Ads primary-value ratio has another numerator. A familiar label such as ROAS should not hide the change during a team handoff.

The incoming colleague can then retrieve current evidence when needed without rebuilding every unrelated report. A new period or changed selection justifies a fresh read; a different person viewing the same returned points does not automatically require another provider request. The handoff preserves both accuracy and useful efficiency.

Keep recommendations and execution separate

A report identifies observations under selected definitions. A recommendation connects those observations to business context. A supported live edit needs a concrete target and operation. Keep these stages clear so the client does not mistake a low-performing row for an automatic campaign pause.

For conversational account changes, ppctor prepares a proposal naming the target, relevant parents and understandable steps. A person confirms before execution. Proposals expire after fifteen minutes, and relevant settings and authorisation are checked again. The reviewer should see the actual action rather than a wall of technical identifiers.

The result distinguishes verified completion from an operation sent without full verification. An uncertain creation should not be blindly repeated. Inspect current account state before another attempt. A subsequent reporting movement also does not automatically prove that the edit caused it.

The confirmation workflow explains the account path. External destination, implementation and commercial tasks need their own owners and evidence. A verified account edit cannot certify all of them merely because it followed the report.

Agency reporting questions

Should every client use the same conversion definition?

No. Keep the reporting method consistent while selecting the outcome that fits each client's question.

Are primary totals unique customers?

No. They add the selected primary actions without guessing which records overlap.

Is primary conversion value purchase revenue?

Not necessarily. It can include non-purchase values; purchase revenue is a narrower selection.

Can related Ads and Analytics values share a chart?

Yes, under compatible selected measurements with clear source and scope labels. Do not add overlapping sources together.

Does a requested chart need daily points?

A daily chart needs returned daily values. A period total cannot supply that history by itself.

Can a saved report run immediately?

Run Now queues an enabled saved report. A paused report must first be resumed.

Does a report authorise a live change?

No. A conversational edit needs a separate concrete proposal and confirmation.

Does monthly delivery always report last month?

No. Delivery frequency and reporting range are separate saved choices. Review both so the client knows which period the recurring message describes.