Google Ads reporting for in-house teams: agree the definition before the decision

Start with the decision due at the review

An in-house report should help the team understand a result and make a defined decision. A large collection of advertising figures can be accurate while leaving nobody sure what needs attention.

Identify the decision before selecting metrics. The team may need to review a spending plan, investigate purchase value, understand a campaign's contribution or confirm the outcome of completed work. Each question requires a different set of evidence.

ppctor's saved reporting provides a consistent delivery routine, while conversational tools can retrieve scoped follow-ups. The team should use both around its actual review question rather than treating available fields as a checklist to include in every message.

Keep the decision proportional to the observation. A short period may support further investigation rather than a major budget change. A current setting read can confirm configuration without explaining commercial performance.

State who needs to respond and what information is missing. A report that asks for a specific business fact or a clear approval is easier to use than a narrative ending with no identifiable next step.

The useful starting point is a short sentence describing the report's purpose. It guides the selection and helps different readers understand why the evidence matters to the team's current plan.

Align marketing, finance and ecommerce readers

Different in-house readers bring different responsibilities. Marketing may focus on campaign outcomes, finance on spending context and ecommerce on product availability or purchase value. They can share evidence without needing identical detail.

Choose a principal reader for the opening explanation. Give that person the result and decision they need, then retain supporting definitions for colleagues who will investigate the figures more closely.

Keep the meanings consistent across versions. A short executive note and a detailed campaign table should describe the same source, period and measurement population. Simplification should not change the number's definition.

Explain consequential terminology at first use. A cost-per-purchase measure is clearer than an unexplained CPA label when another reader assumes the denominator includes all conversions.

Do not treat a platform Revenue label as a finance-certified sales total. The selected measurement can have a different scope from the business's accounting records. Identify the distinction before the review becomes an argument about which team is correct.

Assign the follow-up to the reader who can provide the needed context. The ecommerce team may confirm stock changes, while analytics can investigate implementation. A shared report works best when it also makes responsibility clear.

Resolve a disagreement about meaning before requesting more data

When colleagues disagree about a total, first ask what each person believes it includes. One may be discussing attributed conversion value while another means all-site purchase revenue. Another may use a finance record adjusted for a different business period. Retrieving the same number repeatedly does not resolve those different meanings.

Write the two definitions beside each other and identify the business question being decided. The team can then choose the appropriate observation or deliberately compare both. This is more useful than selecting whichever source produces the more favourable result.

If the definitions really match, continue with the actual dates, currency and population. A precise disagreement gives the next investigator a defined task instead of an open-ended demand to make every system agree.

Write a small measurement agreement

A measurement agreement records what the team's principal report means. It can be concise: source, account or property, period, scope, outcome definition, comparison and responsible reviewer.

Specify whether conversions mean all primary actions, purchase actions or a named action. These populations can produce different totals. The agreement prevents a later reader from assuming that a familiar label always means purchases.

Record the intended traffic scope. All-site GA4 purchase revenue and Google Ads-attributed traffic answer different questions. A paid-media review should not silently adopt an all-traffic numerator.

Include the spending source for blended ROAS or MER. All-traffic purchase revenue divided by Ads cost is a different scope from dividing by all advertising cost available in GA4. The chosen denominator needs agreement.

Keep the reporting period and comparison in the note. A rolling period and a calendar month can both be valid, but their figures are not interchangeable. Actual dates make later discussions easier.

Update the agreement when objectives or tracking change. Preserve the previous definition for historical interpretation rather than rewriting old results as though the new selection always applied.

Select the source and account scope

Saved reporting distinguishes Google Ads and Analytics sources. Each has its own selected data and valid scope. A source choice should match the team's question rather than the interface the reviewer happens to open first.

For Google Ads, an account report and a campaign report cover different entity populations. Name the selected campaign when the result is scoped there. Do not generalise it into an account-wide conclusion.

For Analytics, the linked property and segment matter. A property can describe more activity than the connected advertising account, and a traffic segment can deliberately narrow that population.

The Slack conversation is bound to its connected account context. Ownership of several accounts does not mean one channel should silently query or disclose all of them. Verify the intended connection before interpreting a result.

Use recognisable account and campaign names in the team discussion. Internal identifiers can resolve resources behind the scenes without becoming the client's primary way to understand the scope.

If the connection is unavailable, resolve it before presenting current figures. A remembered value from another account or earlier review cannot replace the missing current observation.

Distinguish rolling and calendar periods

The saved completed-period choices include yesterday, the last seven, fourteen or thirty days, the current month through yesterday and the previous month. These windows answer different review questions.

A last-thirty-days report ends yesterday and reaches back thirty completed dates. It is not a last-month report. The difference is especially visible near the beginning of a month.

The current month through yesterday excludes the unfinished current day. On the first day of a month, that period is empty rather than silently using the previous month's data to make the report look populated.

Keep the observation dates separate from the delivery date. A report arriving on a particular morning can contain a completed period chosen independently of the schedule.

If the team requests today's activity, describe it as partial and retrieve the relevant current-day evidence. Do not interpret a few hours against an entire completed day without an appropriate matched comparison.

Retain the date window in follow-ups. A request to change the chart format should not also change the period. A request for another period should preserve the metric and population unless the user changes those too.

Read the actual comparison dates

Saved reports can use no comparison, a previous period or the same period last year. Read the returned comparison dates before interpreting the percentage movement. A conversational label can conceal a calendar difference.

For a completed rolling window, the previous comparison is the immediately preceding interval of equal length. This produces a clear adjacent-period comparison without assuming calendar-month boundaries.

For a previous-month report, the previous-period calculation is also a preceding equal-length interval. It is not necessarily the entire month before that. The actual dates are the authoritative comparison scope.

The current-month-through-yesterday comparison uses the corresponding portion of the previous month, clamped if that month is shorter. A clamp can change coverage and should remain visible when it affects the interpretation.

Year-over-year date shifting also respects valid calendar dates. A leap-day boundary can require a clamped date. Do not imply that every shifted comparison has identical weekday composition or business conditions.

The team should choose the comparison that answers its decision. Adjacent periods, calendar expectations and year-over-year context can each be useful, but none should be treated as a universal explanation of seasonality or causation.

Separate schedule time from data time

The reporting schedule determines when a saved report is due. The data's calendar context determines which dates belong in its observation. These choices should not be merged into one vague statement about local time.

The current reporting evaluation uses the Google Ads account time zone for delivery cadence and the Analytics property's time zone for its data period. A team spanning locations should understand that distinction.

A consultant's or employee's computer time does not redefine the connected account's day. Use the report's returned dates and time-zone context rather than assuming every user sees the same current-day boundary.

A completed-day window is useful for routine comparison, but it does not mean the provider's values are permanently final. Processing and later updates remain separate from the calendar boundary.

When colleagues compare two reports, check source time zones along with dates. An apparent disagreement can concern coverage before it concerns the underlying measurement.

Record the schedule choice in the reporting agreement. The team should know when to expect the message and who reviews it, without assuming that delivery establishes the finality or completeness of all provider data.

Define conversions and purchase outcomes

The conversational business catalogue distinguishes all-primary conversions, purchases, leads, calls, offline actions and named conversion actions. These selections make the intended outcome explicit.

Primary measurements exclude secondary actions by default unless the user asks for the secondary population. They add the selected primary actions together without deduplicating separately recorded business outcomes.

Purchase counts select primary actions in the Purchase category on the Ads side. All-primary conversions include a broader group. A changed selection can alter the count without proving that the business suddenly gained or lost purchases.

Lead reporting combines the supported primary lead categories without a lead-stage hierarchy. A site-specific GA4 comparison needs the corresponding event to be named or confirmed by the user.

Calls can carry value, and a phone-link click is not automatically a completed call. Offline uploaded or imported actions can overlap purchase, lead or call groups, so overlapping categories should not be added as independent totals.

Keep a saved reporting metric's definition accurate too. Its catalogue should not be silently reinterpreted as every conversational recipe. The team needs to know which workflow produced the number it is discussing.

Make the revenue definition explicit

All-primary Google Ads conversion value can include values from non-purchase actions. Purchase revenue selects the values of primary Purchase actions. Both can be useful, but they describe different outcomes.

The saved reporting catalogue labels its conversion-value metric Revenue. That label alone does not make it purchase-only revenue. Confirm the actual selected field before explaining the number as sales.

GA4 purchase revenue in the app's business measurement defaults to all traffic unless Google Ads traffic is explicitly requested. Keep that scope visible when it is shown beside an advertising measurement.

The team can compare compatible revenue series while retaining distinct source labels and definitions. It should not add attributed Ads value to GA4 purchase revenue as though they represent independent pools of sales.

Revenue is also different from profit or billed cash. Product margin, refunds and finance definitions can require additional business evidence. A platform figure should not be presented as a complete accounting reconciliation.

Use a brief explanatory sentence where the distinction matters. The principal reader does not need every implementation field, but they do need to know whether the result includes purchases, other actions or all-site traffic.

Calculate ratios from the relevant inputs

A ratio should retain its numerator, denominator and population. ROAS concerns a selected value divided by a selected cost. Cost per result concerns cost divided by the specified outcome count.

For a period summary, combine the period inputs before calculating the ratio. Averaging daily ratios can give small days disproportionate influence and answer a different question from the overall period's return.

Keep undefined values explicit. With no spend, the relevant return ratio cannot be calculated normally. With no selected outcomes, cost per result may have no valid denominator. An undefined value is not an invented zero.

Explain a movement through the inputs when the team needs a decision. Value can rise while ROAS falls if cost rises faster. A bare ratio change can conceal that useful business interpretation.

For blended ROAS or MER, use all-traffic purchase revenue and the agreed cost source. Name the denominator. This is a different scope from an advertising-specific purchase-return measure.

Keep units readable and consistent. A multiple and a percentage can express related ratios, but the presentation should not leave readers guessing. Monetary amounts should retain the relevant currency context.

Retain traffic and entity populations

Shared scope applies to every requested series in the conversational plan. A series-specific population can deliberately differ, but that difference must remain identifiable in the report.

Enabled campaigns now and campaigns with historical activity are different sets. A currently paused campaign may still contribute to spending during the selected period. Choose the population that answers the team's question.

A channel, source or campaign segment in Analytics also changes the population. Do not describe a filtered result as all-site performance because the abbreviated label looks cleaner.

Grouped rows can overlap depending on the measurement and relationship. Unique users, repeated product occurrences and separate conversion categories require careful aggregation. Do not assume every visible row can be summed into a valid total.

Parent context matters for deeper entities. A group name or keyword text can recur across campaigns. Retain the relevant campaign and group so the reviewer can identify the intended target.

When a follow-up narrows the population, state the new selection. The earlier total remains accurate for its original scope. A clear explanation avoids treating two differently scoped answers as contradictory evidence.

Keep users, sessions and events distinct

Users, sessions and events are different units. A question about visits should not automatically return users, and an event count should not be presented as a unique-person count.

Google's reporting-identity guidance explains that Analytics can combine activity across devices into a user journey. Several sessions can belong to one measured user, so those totals need not match.

The available identity and collection context can affect how user reporting is interpreted. Do not infer the business's unique customer count from an Analytics user total without the relevant definition.

For paid traffic, select the intended channel or source population and the requested unit. Paid sessions and advertising clicks are related observations, not interchangeable measurements of the same action.

An event name observed in the property is evidence that the name appears in the available data. It does not prove the implementation measures the intended business action correctly. Confirm the user's selected event for a specific comparison.

Keep these distinctions readable in the team note. A short label such as paid sessions or selected-event count can prevent a later reader from treating the result as a different unit of activity.

Read an invented team comparison

The following business and figures are invented. Hazel Outdoors has marketing and ecommerce colleagues reviewing purchase performance over two completed equal-length periods with the same source, campaign population and purchase definition.

The first period records cost 2,400, eighty primary purchases and purchase value 9,600. The second records cost 3,000, one hundred purchases and purchase value 10,500. Monetary values use the same illustrative currency.

Cost and purchase count each increase by 25%, so cost per recorded purchase remains 30. Purchase value increases by 9.38% when rounded to two decimals. Purchase ROAS falls from 4.00 to 3.50.

Value per recorded purchase falls from 120 to 105. This identifies a question about the value associated with the selected purchases, not proof of its cause. The team needs relevant confirmed business context.

Marketing can explain the account observation while ecommerce confirms whether product mix or offers changed during the period. Finance can assess the commercial implication using its own relevant margin information.

The report should therefore present the increased volume and unchanged acquisition cost alongside weaker recorded value per purchase. It should not select only the favourable number or claim that the campaign became unprofitable without the necessary business evidence.

Give the meeting a specific next question

For the invented example, a useful next question is whether the value associated with the selected purchases changed because of a confirmed offer or product mix. Ecommerce can provide that fact for the same observation period. The team should not substitute a general impression about the month.

Marketing can also verify that the conversion definition stayed consistent. If it changed, that affects the comparison's interpretation. The report should preserve the original arithmetic while explaining the population difference rather than rewrite the figures to maintain a simple narrative.

The resulting decision can be to retain the spending plan during investigation. That is an illustrative bounded choice, not a universal rule for falling ROAS. The actual team's objective and verified context determine its response.

Compare sources without combining them

A shared chart can help the team inspect compatible Ads and GA4 measurements over the same defined period. The series must retain distinct source labels, populations and metric meanings.

Different attribution is allowed in the reporting framework when the comparison is appropriately defined. Comparable metrics do not have to produce identical totals. A difference should prompt interpretation rather than an automatic claim that one source is wrong.

Do not add the two revenue series together. They can describe overlapping business activity through different measurement systems. A summed total would imply independent revenue that the comparison does not establish.

For lead or call comparisons, a corresponding GA4 event must be named or confirmed. A likely name suggested by the assistant is a candidate, not the user's selection or proof of the event's implementation.

The current named-comparison delivery gate remains separate from available catalogue definitions. Likewise, multi-table delivery is not assumed verified merely because one table works. Respect the actual enabled composition capability.

When a requested composition cannot be delivered, identify the missing selection or current app capability accurately. Do not describe an invalid or gated plan as proof that the provider lacks the requested daily data.

Choose a readable reporting grain

The grain determines how time or categories are grouped. A daily trend, weekly summary and period total answer different questions. Choose the one that supports the team's review.

The native Slack chart workflow allows twenty points per series and up to three series. For a long period without an explicit daily requirement, automatic grain can become coarser to preserve one readable chart.

If the user explicitly requests daily detail, preserve that requirement and explain any split or narrower selection needed by the chart limit. Do not silently remove early days to fit a single attractive visual.

A table can be useful for exact values or a bounded comparison. A line chart can reveal a time pattern, while a bar chart can compare categories. Use the requested supported format when it fits the evidence.

Keep labels readable and units consistent. Current-year dates do not need an unnecessary year in every label, while historical coverage still needs enough context. A weekly bucket should be understandable as its actual date range.

Ask for a visual when it serves the question. The assistant's renderer uses returned data references and does not invent arrays. A brief explanation can be better when the team only needs the meaning of one ratio.

Distinguish previews from complete evidence

A compact response can show a limited preview of rows while retaining a complete supported result behind its data reference. The visible example count is not automatically the total population count.

Ask for a count when the team needs an exact count. The reporting plan supports counts and carries completeness information. Do not count the names shown in a capped list and call it the account total.

Read coverage notes for filters and relationship datasets. A result can represent a defined subset, a complete selection or an incomplete observation. These distinctions affect what the team can conclude.

A bounded population should not become a continuous graph with invented missing values. A table or clarified request can be more honest than filling absent evidence with zeros to complete a visual.

For a selection that may lead to an edit, completeness matters even more. Editing only the preview is not equivalent to applying the user's condition to the complete intended population.

Keep the limitation near the conclusion it affects. A footnote far below a confident total is easy to miss. The reader should understand whether the evidence supports an exact claim or only useful examples.

Handle provider revisions and missing data

Google's data-freshness guidance explains that processing and attribution updates can revise Analytics results. A completed date boundary does not freeze every reported value permanently.

If a later read differs, compare the source, period, scope and definition before deciding what happened. A changed value can require an explanation of updated provider data rather than an accusation that the previous report was fabricated.

Keep the earlier observation in the review record when the distinction matters. State that the later result was updated rather than silently replacing a figure that informed a previous decision.

A failed read is not a zero result. If the requested evidence was not obtained, identify the missing area and the conclusion it prevents. Preserve successfully retrieved observations without presenting an account-wide all-clear.

Use structured errors to correct invalid plans. A metric or compatibility selection problem is different from provider unavailability. Repeating an unchanged non-retryable request does not improve the evidence.

Give the team a specific next step: confirm the selected event, resolve access or review after processing where appropriate. A useful limitation leads to a task rather than an invented explanation of the account.

Configure a saved reporting routine

The saved configuration separates source, target, metrics, range, comparison and delivery frequency. Set each deliberately. Receiving a message weekly does not by itself define a last-seven-days observation.

Daily, weekly and monthly schedules can support different team routines. Choose the schedule around the review responsibility rather than assuming that more frequent messages mean more useful analysis.

Select the intended Slack destination and confirm its account connection. A correctly generated report posted to an obsolete channel can miss the current decision-maker even when delivery succeeds.

Paused configurations should remain distinguishable from enabled ones. A saved report existing in the management list is not proof that it will currently deliver. Check its status during a handover or configuration review.

Run-now behaviour queues work for an enabled saved configuration. It does not mean the finished report has already arrived. Generation and delivery remain separate stages when explaining a delay.

A scheduled message is a starting point for the team's human review. Agree who reads it and what decision follows. Delivery alone does not guarantee that an employee has assessed the result.

Use Slack for focused follow-up questions

A reporting thread can connect the scheduled observation with the team's next question. Keep the connected account, selected period and principal metric clear as the conversation develops.

Ask for the missing evidence that affects the decision. A purchase-value movement may need the purchase count and value-per-purchase inputs. A spending movement may need a campaign contribution breakdown.

Follow-ups should preserve unchanged scope and series populations. If the user asks for the same result as a table, the metric should not silently become a different conversion population.

If the user changes the period or target, make that difference explicit. The earlier answer can remain correct for its original scope while the new request requires another observation.

Creative and explanatory work can remain conversational. The assistant can help draft a clear internal note from established evidence without calling a provider simply to change the wording.

Before a live request, resolve references such as “that campaign” when several candidates are present. A useful analytical thread still needs an exact target for a later proposal.

Reuse evidence when only the presentation changes

The reporting tools can reuse the same returned data within the request when only presentation changes. A title or format revision should not create a new measurement definition.

This helps keep totals, tables and requested visuals consistent. The renderer uses the returned data reference rather than inventing a numeric series to match the narrative.

A different period, grain or metric may require another read. Efficient reuse should not become a claim that the previous evidence answers every new question. Identify what actually changed.

Local catalogue discovery can resolve available fields and settings without a provider read. Current account values still require the relevant tools. Keep capability information separate from observed facts.

Use the returned totals for the explanation rather than repeatedly querying the same plan to obtain a differently worded answer. If a plan failed with a non-retryable error, correct the selection before another attempt.

The team's reporting routine should request the information it needs, not every available breakdown. Narrow questions reduce retrieval and interpretation work while preserving a clear path to deeper evidence when needed.

Add confirmed business context

The report does not contain every fact that matters to the business. Stock, offer timing, fulfilment limits and sales-team capacity can affect the meaning of advertising outcomes.

Obtain those facts from the responsible team and identify their timing. A promotion that began after the reported period cannot explain an earlier movement. A mid-period change may justify a narrower follow-up.

Keep confirmed facts distinct from hypotheses. A plausible product-mix explanation is not established until the relevant evidence is obtained. The assistant should not invent a business story to make the report sound complete.

Do not infer profit from revenue. Finance may have margin or adjustment information that the advertising report lacks. The report can identify a useful question without answering it from unavailable data.

Share sensitive business context only with the appropriate audience. A reporting channel can have wider membership than the people who should see detailed internal commercial information.

The team adds value by combining defined account observations with verified business facts. That combination supports a responsible decision without overstating what the platform data alone proves.

Keep the context tied to an accountable source

A useful business note identifies who confirmed the fact and which dates it concerns. The report does not need personal or confidential detail to preserve that accountability. A role and clear observation can be enough for the internal review.

If the fact remains unconfirmed, keep it as an investigation question. The team should not make a live account change based on a plausible explanation that nobody has established. A short clarification can be more valuable than another broad performance export.

Keep recommendations separate from live edits

A report can recommend a change without authorising its execution. The assistant prepares a supported edit only when the user requests it, with an exact target and operation for review.

The proposal should retain relevant parent context and visible logical steps. A campaign pause, keyword addition and group enable are different actions. Do not let one disappear inside another's details.

The authorised connected user must click Confirm and Execute. The proposal expires after fifteen minutes, and current settings are checked again before the live request. A report recipient does not automatically have confirmation authority.

Read the returned execution status accurately. Verification confirms the applicable account result. Acceptance without final independent verification is narrower, and uncertainty means some work may already have applied.

Do not blindly repeat an uncertain mutation. Inspect the target before another request, especially for creations. The report should preserve the unresolved outcome rather than state that nothing happened.

A later performance review needs its own period and definition. A verified edit does not prove causation or guarantee commercial improvement. Keep completed work and subsequent observations separate in the team note.

Maintain reporting ownership through changes

A report remains useful only when its definition and review responsibility stay current. Include reporting configuration in account, tracking and personnel handovers rather than treating it as a one-time setup.

Record the principal purpose, source, population, metric, period, comparison and channel. A new reviewer should understand what the saved message means without searching months of conversation.

Keep unresolved questions and recent verified work visible. A pending investigation of purchase value should not disappear because the reporting owner changed. An uncertain edit needs current-state checking before a repeat.

Review schedules and enabled status after the team's routine changes. Messages can continue to arrive even when nobody has accepted responsibility for reading them. Delivery does not maintain the human process automatically.

Preserve historical definitions when updating a measurement agreement. The team should be able to explain why an earlier result differs from a new population instead of treating every changed total as performance.

A usable handover states what is measured, what happened and what decision remains due. This gives the next responsible person a stable starting point while keeping the report accountable to the business objective.

Include the next scheduled review of a deliberately deferred decision. The incoming reviewer should know whether the team is waiting for provider processing, a business fact or another observation window. Silence in the channel should not be interpreted as approval, recovery or abandonment when the actual decision is still pending.

In-house reporting questions

Does a weekly schedule always report seven days?

No. Frequency and observation range are separate configuration choices. Select the intended period and comparison explicitly, then review the returned dates.

Is last thirty days the same as last month?

No. One is a completed rolling window and the other is a calendar month. The previous comparison can also have different boundaries, so retain its actual dates.

Does Revenue always mean purchase revenue?

No. Google Ads conversion value can include non-purchase primary actions, and the saved metric's label does not narrow it automatically. Select a defined purchase measurement when that is the question.

Can Ads and GA4 revenue share a chart?

Reviewed compatible series can be compared with distinct definitions and labels. They should not be added together. Specific compositions also remain subject to the current delivery gates.

Does a saved report replace human review?

No. It supplies defined evidence on a schedule. The team still needs an owner to interpret the result, obtain business context and bring the decision to the responsible person.

What makes the reporting routine reliable?

Consistent definitions, explicit dates, complete evidence and honest limitations make the figures interpretable. Clear ownership and precise follow-ups turn them into useful team decisions.