Explain Google Ads metrics in Slack: definitions, comparisons and useful next questions

Start with what the number measures

A metric becomes useful when its meaning is clear. A low cost can mean efficient traffic, limited delivery or simply a narrow reporting period. A high conversion total can describe valuable outcomes or a broad mixture of actions. A rising revenue figure can reflect more sales, a different value definition or a changed population. The number alone does not identify which interpretation applies.

Begin with a simple question: what exactly does this value count or divide? Identify the source, selected entities, reporting dates and units. For a ratio, identify its numerator and denominator. For conversions, identify the actions included. For revenue, establish whether it is purchase-only or a broader assigned value. These choices make the explanation inspectable before the discussion moves to performance judgement.

A conversational assistant can explain a general definition without reading an account. It needs tools to state actual account values. The distinction should stay clear in the reply. A general explanation of CPC is not evidence that a particular campaign's CPC rose. A measured increase can support investigation, but it does not establish the cause until the relevant evidence is obtained.

ppctor's current reporting framework uses reviewed business measurements and provider fields to preserve these distinctions. The conversation can retrieve a number, explain its components and request a supporting breakdown. This guide shows how to ask for that explanation and how to evaluate the answer, using invented examples rather than customer data or claims about particular account results.

Separate the definition, the calculation and the account explanation

Three questions often arrive together: what does the metric mean, how is it calculated and why did it change? The definition identifies the measurement. The calculation shows the arithmetic for the selected population. The explanation investigates the account behaviour behind the movement. A strong answer distinguishes them instead of treating a formula as a complete diagnosis.

For example, CPC concerns click cost. Its calculation uses cost and clicks. If CPC increases, the measured components can show whether cost grew faster than clicks. That observation does not establish whether bids, traffic mix or another factor produced the change. Additional breakdowns or settings evidence may be necessary. The assistant should not jump from the arithmetic to a cause that has not been checked.

The same structure applies to a conversion decline. Define which actions count, compare their totals across consistent periods and then inspect where the change occurred. A settings read can establish configuration, while a report can establish recorded outcomes. Neither alone proves what happened to every person on the website. Keep the level of evidence aligned with the claim.

Ask the layers separately when the conversation becomes unclear. “Define this metric” can receive a general explanation. “Show the components for our last seven completed days” needs account data. “Which campaigns contributed to the change?” needs a supporting breakdown. This sequence makes the next tool call purposeful and gives the user a chance to inspect each interpretation before moving further.

Follow a worked CPC question through its evidence

A fictional retailer, Stonebrook Stationery, asks: “What was our CPC over the last seven completed days, and explain it using cost and clicks?” The question identifies the metric, reporting period and components. It does not request a bid adjustment or an explanation of auction competition. The following query illustrates the account-level read relevant to the calculation.

SELECT metrics.cost_micros, metrics.clicks
FROM customer
WHERE segments.date DURING LAST_7_DAYS

Suppose the illustrative result is £1,500.00 in spend and 1,000 clicks. A suitable answer is: “Average CPC was £1.50: £1,500.00 divided by 1,000 clicks over the selected seven completed days.” The figures are invented for this example. A live answer must use the values actually returned by the connected account tool.

If the previous comparable period had £1,200.00 in spend and 1,000 clicks, its CPC was £1.20. The increase is £0.30, or twenty-five percent relative to £1.20. That comparison establishes a change in cost per click. It does not establish which campaign caused it or why the underlying clicks became more expensive.

The next useful request is a campaign breakdown for both periods, with consistent populations. That can identify contributions to the change. A later investigation can inspect relevant settings or other supported evidence. The worked answer stays concise because it explains exactly what the retrieved components establish and leaves the causal question for the evidence that can answer it.

Read spend and traffic together without treating either as success

Spend describes the selected advertising cost. Clicks and impressions describe traffic or exposure within their source definitions. They are useful foundations for a report, but none is automatically a business outcome. A campaign can spend more because it delivers more activity, while its purchase efficiency remains stable. Another can spend less because it stopped serving rather than because it became more efficient.

When a cost figure changes, inspect the relevant traffic components and period. The same spend spread across more clicks produces a different CPC. The same clicks spread across more impressions produces a different CTR. A campaign-type or traffic-mix shift can alter account averages. Begin with the actual components before deciding that the whole account became better or worse.

Current status and historical activity are different choices. An enabled-only selection can exclude a paused campaign that spent during the period. An all-campaign account report includes the requested historical population. If the user compares a narrow conversational result with a broader dashboard total, the scope can explain the discrepancy. The reply should name the selection so that the comparison is possible.

Use business outcomes to complete the interpretation. More clicks are valuable only in relation to the account's objective and the quality of the resulting activity. A lower CPC can still coincide with weaker purchase or lead outcomes. Ask for the agreed primary outcome population rather than turning a traffic improvement into a claim of improved profitability without the necessary evidence.

Interpret CTR through clicks and impressions

CTR describes clicks relative to impressions. Google explains this relationship in its CTR definition. For an invented example, 400 clicks from 20,000 impressions produce a two-percent CTR. The calculation describes the selected advertising activity; it does not measure the percentage of website visitors who purchased.

When CTR changes, inspect both components. More impressions with unchanged clicks reduce the ratio. More clicks with unchanged impressions increase it. The reason behind those movements needs further evidence. A change in audience, campaign mix or serving conditions can alter the account-level result, but the ratio alone does not establish which factor was responsible.

Compare suitable populations. A brand Search campaign and a broader prospecting campaign can have different roles and contexts. A single universal threshold can obscure that difference. A team can compare an asset with its own previous performance or with a deliberately chosen peer group, while preserving the same source, period and definition. The answer should state the comparison rather than imply that every high CTR is good enough.

Connect the ratio to outcomes before acting. A headline that attracts more clicks can still produce weaker purchase quality. A lower CTR can coexist with valuable traffic in a different population. Ask for primary outcomes and their cost if the decision concerns business efficiency. CTR is a useful observation, not a substitute for the complete decision.

Interpret average CPC without confusing it with a bid setting

Google defines average CPC using total click cost divided by clicks and distinguishes it from a maximum CPC bid in its average CPC guidance. A reported average describes what happened in the selected activity. It is not the same thing as a setting telling the campaign what it may bid.

That distinction matters when asking for explanations. “Why is CPC £1.50?” asks about a measured result. “What is the bid limit?” asks about configuration. The two can be related in some strategies, but the assistant should retrieve the appropriate evidence rather than answer a settings question with a performance average. A general bidding explanation should also remain separate from the account's current configuration.

Account CPC can change when traffic shifts between campaigns, even if campaign-level CPCs remain similar. A campaign with higher click costs can contribute a larger share of clicks. The account average then moves because the mix changed. A breakdown can establish that contribution, while further evidence is needed to explain why the mix changed.

Judge the result alongside primary outcomes. A higher CPC can be acceptable when the traffic produces sufficiently stronger outcomes. A lower CPC can be unhelpful when it buys weaker traffic. The CPC investigation guide explains how to move from the observed change to comparable breakdowns and relevant checks without selecting a bid edit from the average alone.

Define the conversion population before explaining its total

A conversion total is an aggregation of selected recorded actions. In the current business framework, ordinary Ads conversion measurements include primary actions and exclude secondary actions unless explicitly requested. The catalogue distinguishes all primary conversions, purchases, leads, calls, offline actions and an individually named action. Those selections answer different questions and should not share an ambiguous label.

Google's primary and secondary action guidance explains the provider roles and their reporting and bidding context. The application's business measurement makes its own explicit primary-only selection. A conversion action's category and role are configuration evidence. They do not independently establish that the website records every event correctly or that a recorded lead became a sale.

Selected primary actions are added together. If two primary actions record outcomes, both contribute to the requested total. The reporting tool does not repair duplicate tracking or merge separate actions into one real-world customer. That is the intended aggregation behaviour. If the user wants one named action, select it explicitly rather than infer that a broad total has already isolated it.

Keep call and offline measurements available without inventing a universal counterpart. Calls can be primary actions and carry assigned value. Uploaded or imported actions can overlap purchase, lead or call categories. Do not add overlapping groups into a supposedly larger total. A precise explanation names the selected population and the corresponding value definition before interpreting performance.

Interpret cost per outcome through the selected denominator

Cost per outcome divides selected spend by the selected recorded outcomes. The current catalogue provides cost per all primary conversion, cost per purchase, cost per lead and cost per call. A lower number is not automatically better across unlike populations. A £20.00 form lead and a £20.00 purchase describe different outcomes, and even two form types can have different business roles.

For an invented example, £600.00 divided by twenty primary lead conversions gives £30.00 per lead. If the user instead selects ten primary purchases, the same spend gives £60.00 per purchase. Both calculations can be correct because their denominators differ. The answer should state the outcome population rather than quote the ratio as an unqualified CPA.

A changed denominator can move the ratio without a cost change. More recorded outcomes reduce cost per outcome. Fewer outcomes increase it. That can reflect actual performance, a changed selection or reporting availability. Confirm the conversion population and comparable periods before explaining a decline as a business problem. A ratio hides those choices when its components are omitted.

When there are no recorded outcomes, the ratio has no finite value. It should not be shown as £0.00, which would imply an exceptionally cheap outcome. The answer can report the spend and state that no cost per outcome can be calculated for the selected period. This preserves the underlying evidence and points toward the next appropriate question.

Distinguish conversion value from purchase revenue

Conversion value is not necessarily purchase revenue. A business can assign values to lead, call or other primary actions. The current all-primary value measurement includes the values of the selected primary actions, including non-purchases. Purchase revenue in the Ads business catalogue selects primary actions in the Purchase category. A question about revenue needs to establish which population is intended.

Suppose an invented account records £4,000.00 in primary purchase value and £500.00 assigned to primary calls. Its all-primary conversion value is £4,500.00. Its selected primary purchase value is £4,000.00. Neither result contradicts the other. They have different definitions. Describing the broader amount simply as purchase revenue would hide the call component and misstate the requested measure.

The recorded value also depends on the selected actions and implementation. The tool can inspect metadata and retrieve values. It cannot infer that every assigned amount represents a completed paid order merely because the total is expressed in money. A business review should understand which actions carry values and why those values were chosen.

For comparisons with GA4, preserve source labels. GA4 purchase revenue and Ads primary purchase value can be compared with explicit scope, while all-primary Ads value is a broader measure. The comparison should not add the two sources together as independent revenue. A useful explanation tells the user what each value measures before discussing why the totals differ.

Name the revenue and spend behind every ROAS figure

ROAS is meaningful only with its numerator and denominator. The current purchase ROAS measurement uses primary purchase value divided by spend on the Ads side. The all-primary ROAS measurement uses all primary conversion value divided by spend. Those can differ when non-purchase primary actions carry value. The user should know which result the conversation selected.

For an invented purchase example, £6,000.00 in primary purchase value divided by £2,000.00 in spend produces a ratio of three, equivalent to 300 percent. The explanatory sentence can say that each pound of selected spend corresponds to three pounds of recorded purchase value. It should not call the difference profit, because product cost and other business expenses are not part of that ratio.

A GA4 calculation using purchase revenue from all traffic divided by selected advertising cost is blended ROAS or MER. The spend source must be selected: Google Ads spend or the advertising cost available in GA4. The all-traffic numerator should not be silently narrowed to Google Ads traffic. Naming both the traffic scope and spend source is essential to the explanation.

If the user asks for an unspecified performance view, the framework can choose a value-based or cost-per-conversion measure according to the available account context and recorded value. The reply should make the selected metric clear. A deliberate request for a specific measure is preferable when the team is comparing periods, setting thresholds or sharing a result with finance.

Explain average order value using the purchase population

Average order value relates purchase revenue to purchases. In the current Ads business catalogue, the calculation uses primary purchase value divided by primary purchases. The GA4 counterpart uses average purchase revenue. The source and selected purchase population remain important, especially when the same business compares advertising-attributed outcomes with a broader site reporting scope.

For an invented example, £9,000.00 in selected purchase value divided by 150 selected purchases gives £60.00 per purchase. That is an average across the population. It does not mean every order was worth £60.00, and it does not identify which products or customers created the distribution. A product-level or more detailed analysis requires the appropriate additional evidence.

AOV can move independently of purchase count. The account can record fewer purchases with higher values or more purchases with lower values. Revenue and purchase count together reveal that arithmetic. The cause of the change still needs investigation. Do not claim that a particular offer increased basket size merely because average value rose during the same period.

Keep definitions consistent when comparing sources or periods. Primary purchase actions and all primary actions are different populations. A refund or revenue-definition difference can affect the interpretation of a source comparison. If the selected population contains no purchases, AOV has no value for that period. The answer should state that boundary instead of displaying a zero that looks like a measured order value.

Compare consistent dates and populations before explaining a movement

A percentage change is useful only when the compared values have comparable definitions. A rolling seven-day window and a calendar month are not equivalent. Enabled campaigns and all historical campaign activity are not equivalent. Purchase-only value and all-primary value are not equivalent. Establish those choices before calculating a difference and assigning business meaning to it.

The date basis for conversions also matters. The current business framework reports daily outcomes by conversion date by default, with ad-interaction date available when explicitly requested. The answer should retain that choice. A report built around conversion date can show a different daily shape from one organised around the associated advertising interaction, even when the underlying recorded outcomes are related.

Recent periods can contain incomplete processing. A late outcome can update a previous window, and a partial current day should not be compared casually with a full day. Use completed windows when that fits the task and inspect the actual returned dates. The assistant should not substitute last month for last thirty days merely because a calendar chart is easier to label.

When a difference appears, confirm the comparison before expanding the investigation. “Use the same primary purchase actions and campaign population in both periods” can establish a cleaner baseline. Then ask for the breakdown that identifies contributions. This sequence prevents a population change from being described as a performance change and makes the subsequent explanation easier to verify.

Recalculate combined ratios from their components

Averages from different groups do not usually combine by taking their simple mean. Each group contributes a different amount of activity. A framework that explains account metrics should retain the relevant components and calculate the combined ratio from those totals. Otherwise, a small campaign can receive the same weight as a much larger one in the explanation.

Consider two invented CPC rows. Campaign A has £100.00 in cost and 100 clicks, giving £1.00 CPC. Campaign B has £900.00 in cost and 300 clicks, giving £3.00 CPC. The combined result is £1,000.00 divided by 400 clicks, or £2.50. The simple mean of £1.00 and £3.00 is £2.00, which does not describe the combined click cost.

The same principle applies to CTR, cost per outcome and revenue efficiency. Combine the compatible numerator and denominator for the selected population. Do not average daily ratios without the underlying daily weights. The current reporting framework declares ratio components and additive behaviour, and rejects combinations that cannot be safely aggregated. A missing total is preferable to invented arithmetic over overlapping rows.

Ask for the components when a combined result seems inconsistent with its rows. Check whether the total covers all matching rows or only a displayed ranking. Check whether the rows are disjoint and whether the selected measure is additive. Those questions make the explanation verifiable without exposing the technical plan in every ordinary reply.

Distinguish zero, undefined and unavailable values

Zero is a measured amount. Undefined means the requested calculation has no value under the selected components. Unavailable means the required evidence was not obtained. These are different states. Displaying them all as zero can make a failed read look like no activity or make a ratio without a denominator look like excellent efficiency.

If spend is positive and recorded purchases are zero, cost per purchase has no finite value. If spend is zero, ROAS cannot be calculated from that spend denominator. If impressions are zero, CTR has no value for that selection. The underlying components can still be reported. An explanation should identify the missing denominator rather than hide it behind a numeric placeholder.

A provider failure is different again. If the tool could not retrieve purchases, it cannot establish that the purchase count is zero. If a relationship dataset lacks item-level performance coverage, a threshold filter may have unknown candidates. The answer should preserve that uncertainty and coverage, then explain the next selection or connection needed to resolve it.

Formatting should support the distinction. Ordinary output uses readable units and no more than two decimal places, while values remain grounded in the measured data. A tiny monetary value should not be treated as a false exact zero through careless rounding. The goal is a clear explanation of the state of the evidence, not a full table of numeric values regardless of whether those values exist.

Explain source differences without pretending they are errors

Google Ads and GA4 describe related activity through different measurement systems. A comparison can be useful when the selected business quantities align and the source labels stay visible. It does not require the two totals to be identical. A difference should lead to a definition and scope review before being labelled a tracking failure.

The current catalogue supports reviewed comparisons such as advertising cost, clicks, impressions, purchase outcomes and selected conversion-event relationships. All primary Ads conversions can be compared with GA4 key events. Leads and calls need a corresponding user-selected event when compared with GA4. A phone-link event should not be described as a confirmed completed call merely because its name contains the word call.

Revenue comparison needs explicit labels. Ads all-primary value includes non-purchase primary values. GA4 purchase revenue is a purchase measure with its selected traffic scope. The sources can be shown together for the intended comparison, but their values should not be added into a combined revenue total. Both may describe related business activity rather than independent contributions.

Ask the explanation to name the components. “Compare Ads primary purchase value with all-traffic GA4 purchase revenue for the same dates” is clearer than “compare revenue.” If the goal is a ratio, choose the spend source separately. A good answer can explain different scope and attribution while preserving a requested comparison instead of refusing it simply because the measurements are not identical.

Investigate a cause instead of selecting the first plausible story

A metric explanation can become overconfident when it treats a familiar pattern as a diagnosis. CPC rises, so the answer blames competition. Conversions fall, so it blames the landing page. Revenue increases, so it credits a recent creative change. Those stories may be plausible, but a useful account answer must distinguish them from what the available evidence actually establishes.

Begin by locating the movement. Compare consistent periods and identify which campaigns or supported segments contributed. Inspect the conversion population and source. If a setting is relevant, read the current configuration. If recent changes matter, use the supported history scope and its limitations. Each step should address a specific competing explanation rather than collect unrelated data.

A breakdown can narrow the question without proving causality. If one campaign accounts for most of a CPC increase, investigate that campaign next. If several campaigns move similarly, a broader explanation may deserve attention. A measured contribution is valuable evidence. It should not be expanded into a claim about auction behaviour, customer intent or website implementation that was never checked.

General guidance still has a role. The assistant can explain common possibilities and suggest the evidence that would distinguish them. It should identify that guidance as general and state which account observations are measured. The resulting conversation gives the analyst a useful investigation path without presenting a fluent theory as a verified cause.

Choose a table or chart that helps explain the metric

A formula often needs only a short sentence and its components. A trend can benefit from a line chart. A contribution comparison can benefit from a bar chart or table. Ask for the presentation that makes the explanation easier to inspect. A visual should not appear automatically with every definition or conceal the units behind a tidy shape.

For a CPC movement, a table containing cost, clicks and CPC by period can show the arithmetic clearly. A daily CPC chart can show when the ratio moved, but it still needs the component data for a causal investigation. For a revenue comparison, distinct source labels make the two series interpretable. Do not use two identical labels that leave the reader guessing which line belongs to which system.

The current native Slack charts support up to three series and twenty points per ordinary chart. Automatic grain can choose a coarser readable interval when daily detail is not required. Explicit detailed requests preserve the requested points within output limits, with an explanation of any split or need for a narrower request. Numeric values come from provider data rather than arrays invented by the model.

Keep a short interpretation beside the visual. State the selected dates, population and definition when they affect the result. Explain whether a total covers all matching data or the displayed rows. The purpose is to make the metric inspectable, not to surround it with a technical object that an ordinary user must decode.

Work through a fictional lead-account explanation

A fictional consultancy, Westmere Planning, asks why its cost per lead increased. The team first confirms that leads mean its primary lead actions, without separating them into different lead stages. It selects comparable completed periods and retrieves spend and the selected lead count. The initial explanation can then show whether cost, outcomes or both changed.

Suppose the illustrative first period has £1,000.00 in spend and fifty primary leads, giving £20.00 per lead. The second has £1,200.00 and forty leads, giving £30.00 per lead. The increase comes from higher cost and fewer recorded leads. These invented values demonstrate the calculation; they do not establish the reason behind either component's movement.

The team asks which campaigns contributed, then checks whether the relevant action population stayed consistent. For a GA4 comparison, it names the event used for its contact form. The assistant can compare that selected event with the relevant Ads action or aggregate, keeping source labels visible. It should not choose an unfamiliar event automatically and treat the name as proof of a lead.

The investigation may lead to a configuration review or a website check. A live account edit remains separate and requires the supported proposal. The useful outcome of the conversation is a defined metric, verified components and a focused next question. It is not a guarantee that lowering bids or changing one setting will restore the earlier ratio.

Work through a fictional retailer's revenue explanation

A fictional retailer, Pineglass Living, asks why its advertising ROAS and MER differ. The team establishes the numerator and denominator of each. Purchase ROAS uses the selected primary purchase value and advertising spend. The blended measure uses all-traffic GA4 purchase revenue divided by the selected spend source. The two ratios can differ because their revenue populations differ.

Suppose the illustrative Ads purchase value is £8,000.00 and spend is £2,000.00. Purchase ROAS is four, or 400 percent. If all-traffic GA4 purchase revenue is £12,000.00 and the chosen denominator is the same Ads spend, blended ROAS or MER is six, or 600 percent. These invented values demonstrate the scope difference rather than a customer result.

The explanation should not describe the additional £4,000.00 as revenue caused by Google Ads. The all-traffic numerator includes the requested wider traffic scope. It should not add the two revenue figures together or imply that one must be incorrect because the ratios differ. A comparison is useful precisely because the definitions answer different business questions.

The team can investigate attribution, date scope and measurement implementation if it needs a deeper explanation. It can also compare purchase count and average order value to understand revenue composition. Each request adds relevant evidence. The original ratio comparison remains clear because the conversation preserves the source and population of every component.

Ask a clear follow-up in the account-linked Slack conversation

Start in the destination connected to the intended account and mention the app. Ask for the metric and its components over a clear period. “Show CPC for the last seven completed days and explain it using spend and clicks” is a useful measured request. “What does CPC mean?” is a general definition request. The difference determines whether account evidence is needed.

Continue with a focused follow-up. “Compare with the previous seven days” changes the comparison while preserving the measure. “Which campaigns contributed most?” requests a breakdown. “Use purchase-only value” changes the conversion-value population. The current reporting framework uses stored plans to retain unchanged choices, while additional detail is retrieved when the new question requires it.

Ask for a chart only when useful. A returned dataset can support requested rendering within the same request, but a total cannot reveal a daily distribution it never contained. The workflow should obtain the relevant detail rather than fabricate it. Readable output uses clear labels and ordinary asset names, not numeric IDs or unnecessary implementation fields.

If the reply seems inconsistent, identify the definition to confirm. Ask about the period, primary action population, date basis or spend source. A precise clarification helps resolve the actual mismatch. It also leaves a thread that another analyst can understand without guessing which interpretation of revenue or conversions the user meant.

Turn the explanation into a reviewable decision

A metric explanation supports a decision; it does not make the decision automatically. A higher CPC may lead to a campaign contribution check. A weaker purchase ratio may lead to a margin discussion. A missing conversion denominator may lead to a tracking investigation. The appropriate action depends on the business objective and the evidence beyond the isolated number.

Keep a short record of the conclusion. State the measured change, its population and the next step. Avoid writing a definitive cause when only a contribution was established. Avoid calling a value profitable when the calculation contains no costs beyond advertising spend. A useful handover lets another colleague distinguish the facts, interpretation and chosen investigation.

If the user requests a live edit, prepare the supported proposal with the exact target and action. The confirmation boundary remains separate from the earlier explanation. Relevant settings and authorization are checked again before execution, and supported results are verified afterwards. A metric read can identify a candidate, but it is not itself permission to change that candidate.

For continuing questions, choose a saved report or monitoring condition deliberately. A one-time explanation does not create a schedule or Rule. Review the selected definitions when the business changes its outcomes or measurement implementation. The conversation stays useful when the metric catalogue, account selection and team's business meaning remain aligned over time.

Frequently asked questions about metric explanations

Can the assistant explain a metric without reading my account?

Yes, for a general definition or explanation. Actual account figures must come from returned tool data. The reply should keep general guidance distinct from measured account observations and should not infer a campaign result from memory.

Does conversion value always mean purchase revenue?

No. All-primary conversion value can include non-purchase actions with assigned values. Purchase revenue in the Ads business catalogue selects primary Purchase actions. Choose the population you intend and keep its label visible in the result.

Are secondary conversions included by default?

No. The business conversion measurements select primary actions unless the user explicitly asks for a secondary action or secondary actions. Selected primary actions are added together; the reporting calculation does not repair duplicate tracking implementations.

Can I compare leads with a GA4 event?

Yes, with a corresponding event named or confirmed by the user. The assistant can list observed events and suggest a candidate. The confirmed selection, source labels and reporting period should remain explicit in the comparison.

Does a metric explanation justify an immediate edit?

No. Review the relevant evidence and business objective first. A supported live change requires its separate proposal and confirmation. The explanation establishes the meaning and observations needed to evaluate that decision.