breakdown
ChatGPTMetrics split by channel, campaign, ad, creative or placement. For "which channel", "which campaign performed best", "where is the money going". Each row holds attributed conversions, revenue, costs, CPA and ROAS, plus attributedSetValue: the attributed set value (acquisition value). It is not revenue and is not part of ROAS. Attributed means: the chosen model distributes revenue across the measured touchpoints before it. That does not show that a campaign caused the order, and ROAS is not profit; answer "performed best" as "received the most attributed revenue" and name the model. Costs are in the reporting currency; costsWithoutRate counts costs without an exchange rate that are missing from cost. If days is above 0, mention it. visits are measured visits entering through the row (only with consent); conversionRate divides the attributed shares on those visits, without post-view, by the visits and is therefore attributed, never measured; newCustomerShare is the share of new customers among attributed orders with a known customer type, without leads, null means no customer type is known, so no statement. ordersWithCustomerType is its denominator: the attributed orders with a known customer type. Always state the share with it (“105 of 108 attributed orders with a known customer type are new customers”), never as a share of attributedConversions: that also counts leads and orders without a customer type. renewalConversions and renewalRevenue are the part of the row that comes from subscription renewals, attributed: a renewal takes over the shares of the first order of its subscription, is included in the row totals and counts as an order by an existing customer. attributedMargin is the attributed contribution margin of the row (net revenue minus cost of goods after refunds, only purchases with a margin), marginRoas that margin per cost of the row, read as an amount per 1 unit of ad spend (1.34 means 1.34 of margin per 1 of ad spend, not “1.34 times higher”): below 1, the attributed orders leave less after cost of goods than the ads cost. It is attributed and does not say whether the orders would have happened without the ads. marginCoverage is the share of attributed purchases with a margin; if it is below 1, say so, because the costs are those of the whole row. A purchase has no margin when its net value or its cost of goods is missing, a partial refund came without a split, or an exchange rate is missing. This tool does not say which reason applies to a purchase; counts per reason across all orders come from commerce. Do not name any of these reasons as the cause unless a tool reports it, and do not say that a particular step closes the gap. null means no margin known or no costs, so no statement; never estimate it yourself. At most 200 rows, the largest by attributed conversions: limit states the cap, total the number of all rows. If truncated is true, say so and name total; a sum over zeilen is then a partial sum. All rows are in the CSV export of the "Attribution" page and in the read API with limit=all.
commerce
ChatGPTCommerce metrics as in the portal’s “Commerce” section: orders from new and returning customers, net revenue, cost of goods, contribution margin, repeat purchases, new customer cost (ad spend per attributed new customer conversion), ROAS on contribution margin and the product report. For "what is the contribution margin", "what does a new customer cost", "which products sell", "what did campaign X sell". measured is measured, from the shop’s order data; attributed is attributed using the model under model. margin.margin is net revenue minus cost of goods after refunds, margin.coverage the share of orders it is based on: if it is below 1, say so. orders.newCustomerSource says who determined “new customer”: shop = the shop, measured = our own measurement, mixed = both. orders.newCustomerShare divides orders.newCustomers by new and returning customers together; orders.unknownCustomerType counts the orders without a customer type that are left out. State the share with this denominator, never as a share of orders.total. An empty field means the shop does not send that information (no net value or no cost of goods, no contribution margin; no customer type, no new customers) or ad costs are missing; missingReason gives the reason for missing ratios. margin.withoutBreakdown counts orders without a net value or without cost of goods; the tool does not say which of the two is missing. margin.excludedByRefund counts orders whose partial refund came without a split. Both are missing from the contribution margin; name them as reasons only with these counts. Amounts are in the reporting currency (currency). unconvertedOrders counts orders without an exchange rate, missing from all amounts; costsWithoutRate counts ad costs without a rate, missing from new customer cost and ROAS on contribution margin. If either is above 0, mention it. With campaign, the campaign applies only to the product report (products): items from orders this campaign has a share in, weighted by that share and therefore attributed. Without a campaign, the product report is measured. Metrics outside products always cover all campaigns (campaignScope). Contribution margin and ROAS on contribution margin per channel or per campaign come from the breakdown tool; use it when asked. Not provided: net revenue and new customer cost as a total per campaign or per channel, customer cohorts and individual orders. If asked for these, say that the tool does not provide them; that does not mean the data is missing. Do not add up product report rows into a per-campaign total yourself. products.total is the number of rows, products.truncated is true if only the first limit rows are returned; say so then.
conversions
ChatGPTIndividual conversions with time, value, order number and attribution status. value is measured (from the event), setValue is set (a fixed value from the definition, not revenue). refunded is the refunded amount: 0 means not refunded, equal to value means fully refunded, anything in between partially. reason explains in words why a conversion is not attributed to any touchpoint. renewal is true for a subscription renewal: it takes over the attribution of the first order of its subscription and is never an order by a new customer. For "which orders", "what is unattributed", spot checks and reconciliation. For totals, use metrics; this list is capped. Server-side orders without consent are not listed individually, only counted (withoutConsentHidden); they serve revenue reconciliation only. byType counts the conversions of the period per type (purchase, lead, otherwise a custom goal), without server-side orders without consent. Purchases and leads are different sets: name them separately. On consent, this tool only provides withoutConsentHidden. If it is 0, there is no order without consent in the period. Never infer consent or its absence from the difference between two totals.
dataQuality
ChatGPTMeasurement status: events received, processed and discarded, backlog, attribution rate and open findings. Check this before interpreting numbers: it shows whether a drop is real or a measurement problem. An empty queue and no open findings only show that nothing is waiting and nothing is reported right now. They do not prove that processing ran without problems for the whole period; say “no sign of a problem” instead.
help
ChatGPTThe Metrunio help center: guides, terms and explanations about the product. For questions like "how do I set up X", "what does error code Y mean", "how does model Z calculate", "which data is stored". Returns text excerpts with their source URL, no figures from the customer’s data. For figures, use the other tools.
journeys
ChatGPTCustomer journeys: the paths individual people take across several touchpoints up to a conversion, with the sequence of touchpoints per journey. For "how do customers find us", "how many touchpoints does a sale take", "which channel starts the journey", "which click paths lead to a purchase". Per journey, touchpoints lists the touchpoints in time order (time, type, channel, source, medium, campaign, whether direct, for impressions whether viewable) and conversions lists the conversions with time, type and value, so you can see where in the journey they fall. At most 50 touchpoints per journey, namely those up to the last conversion. touchpointsOmitted counts the rest: they are missing from the list, not from the journey. If journeys are left out because of the volume, their number is in omittedForRowLimit. Can be narrowed to journeys with (or without) a viewable impression, and to the channel of the first touchpoint, such as "journeys with a viewable display impression" or "those that start with paid search". Also to journeys without a conversion and to abandoned carts ("who adds to the cart and does not buy") and to journeys with a touchpoint from a specific campaign ("which paths did the summer campaign touch").
metrics
ChatGPTKey metrics for a period: events, sessions, conversions, revenue, costs, CPA (cost per attributed conversion), CAC (per new customer, only when the shop reports the customer type) and ROAS. processing shows the processing status: events waiting (queueBacklog), the oldest pending one in seconds (oldestPendingSeconds) and the last run. Only when something is waiting there is "not processed yet" a supported explanation. measured.revenue is revenue before refunds, measured.revenueAfterRefunds measured revenue after refunds; attributed.revenue and attributed.unattributedRevenue are attributed and unattributed revenue after refunds and add up to measured.revenueAfterRefunds. Do not calculate differences yourself; quote the fields. Revenue amounts are in the account’s reporting currency. measured.conversionsWithoutRate counts conversions in another currency that have no exchange rate yet: they count as conversions, but their amounts are missing from every revenue field. If it is above 0, mention it in the answer. Ad costs are in the reporting currency too, converted at the rate of the cost day. costs.withoutRate counts costs in another currency without a rate (days, entries, currencies): they are missing from costs.total, CPA, CAC and ROAS. If days is above 0, mention it; ROAS is then too high. attributed.conversionsWithShare is the number of whole conversions that received a share in the model ("8 of 9 conversions attributed", together with measured.conversions); attributed.conversions is the sum of the shares. setValues is the set value (acquisition value): an amount the customer has defined per conversion, for example for a sign-up. It is not revenue, appears in no revenue field and is never added to it; call it "set" in your answer. Subscription renewals: measured.renewalConversions and measured.renewalRevenue (before refunds) are the payments marked as a renewal of a subscription; they are already included in measured.conversions and measured.revenue. attributed.renewalConversions and attributed.renewalRevenue are their attributed part: a renewal takes over the shares of the first order of its subscription and has no ad contact of its own. A renewal is never a new customer and does not count in CAC. For "without renewals", quote both fields (total and of which renewals) instead of subtracting yourself. costs.lateSources lists cost sources that have not delivered data for the end of the period. If one of them has lastError, the connection is broken (an expired login, for example): state the error and point to "Data connections", where it can be reconnected. Waiting for the next sync will not help in that case. A field that is null cannot be calculated; it is not 0. roas is null when there are no ad costs for the period, cpa also when no conversion is attributed, cac when the shop reports no customer type. Then state the missing reason; never write "null" or 0. ROAS is attributed revenue per unit of ad spend, not profit; say so in the answer as well. First choice for "how did we do", "how much revenue", "what did it cost". Returns totals, not a split by channel or campaign: use breakdown for that. measured.conversions counts all conversion types together: purchases, leads and custom goals. Call the number conversions, not purchases; how many of them are purchases comes from conversions (byType). These totals also count conversions from visits that are still ongoing; web analytics (web) only counts completed visits. measurement.complete is false when the period starts before measurement began (measurement.since): do not compare against such a period: a change relative to it reflects the start of measurement, not a trend.
modelComparison
ChatGPTThe same conversions under several attribution models, side by side. For "does this depend on the model", "what does first touch say", "how robust is this number". If the plan has no model comparison, only the one model it uses is listed, and note says which plan includes all of them. In that case, do not give figures for other models.
surveyAnswers
ChatGPTAnswers to the post-purchase question “How did you hear about us?” in the period, per answer with count and share. This is self-reported and is labelled “self-reported” in the text, never “reported” (that word belongs to platform figures): never revenue, never attribution, never a model, and never extrapolate it to purchases without an answer. answerRate is the share of purchases with an answer (denominator purchases); if it is small, say so explicitly. Per answer: attributed purchases with attribution shares, unattributed purchases without attribution, withoutPurchase answers without a measured purchase. Useful for questions measurement alone cannot answer: podcasts, recommendations, TV, AI assistants. enabled: false means the post-purchase question is switched off for the account; then there are no answers.
timeline
ChatGPTHow the metrics develop over time. For "development", "trend", "since when", "has it improved". Returns measured and attributed figures side by side, along with ad costs; cost is null when there is no cost data for that interval. days is the number of days in the interval, costDays the number of days with costs; if costDays is lower than days, cost is a partial sum of the known days and must be described as such. Costs are in the reporting currency; costsWithoutRate counts costs without an exchange rate that are missing from cost. Long periods are grouped into weeks or months; the granularity used is part of the answer and belongs in your statement.
web
ChatGPTWeb analytics as in the "Web analytics" area of the portal: page views, visits, visitors, time on site and bounce rate. For "how many visits", "how long do visitors stay", "how many bounce", "where do visitors come from", "mobile or desktop". view selects the answer: overview = metrics with the previous period (previous, same length right before), daily trend and a table per channel; sources = visits per channel, source, medium and campaign; technology = visits per device, browser, operating system or country (dimension). Everything is measured, only for visitors who consented to analytics and only for completed visits (30 minutes after the last activity); never extrapolate. Times are in milliseconds (avgActiveMs, medianActiveMs): time on site is the visible active time per visit. A visit is engaged with at least 10 seconds of active time, two page views or a conversion; bounceRate is the share of the others. conversionRate is the share of visits with a conversion, measured, not attributed; for attribution per channel use the breakdown tool. consentMissing counts events rejected for lack of consent; null means the counter was not running yet in the period, so no statement. technology only returns the country if it is collected (available). Not provided: individual pages, page paths, paths between pages and real time. If asked, say that this tool does not provide them and point to the "Web analytics" area in the portal; that does not mean the data is missing. total is the number of all rows, truncated is true if only the first limit rows are included; then say so. openVisits (overview only) counts visits of the period that are still ongoing or not completed yet, withConversion those of them with a conversion. They are missing from every number of this tool, but their conversions are already in metrics. If openVisits.visits is above 0, that is a supported explanation for fewer visits, conversions or less time on site here.
breakdown
Claudecommerce
Claudeconversions
ClaudedataQuality
Claudehelp
Claudejourneys
Claudemetrics
ClaudemodelComparison
ClaudesurveyAnswers
Claudetimeline
Claudeweb
Claude