#reporting

Marketing Analytics Readout: Channel Performance From Export

Turn a pasted channel export into a marketing analytics readout: what changed, reasons to verify, budget questions, and a test plan, with no invented numbers.

Act as a marketing analyst who writes weekly and monthly channel readouts for a marketing lead: what changed, how sure we can be, and what to do next, using only the numbers in the export. Inputs: - The export, pasted as CSV or a table (channel or campaign rows, spend, sessions, conversions, revenue, or whatever columns exist): [ChannelExport] - The two periods being compared and their dates: [ComparePeriods] - The business goal and the one metric that matters most (for example qualified leads, trial starts, first orders): [NorthStarMetric] - Known events in the period (launches, price changes, tracking changes, outages, holidays, budget shifts): [KnownEvents] - Attribution model or source of the numbers (GA4 default channel group, ad platform reported, CRM): [AttributionSource] - Output format: [Format] Generate: 1. A data check: columns found, periods found, rows with missing or zero values, and any sign of a tracking break (a channel at zero, conversions without sessions). If the export cannot answer the question, say what column is missing. 2. A headline summary in three sentences: the change in NorthStarMetric between ComparePeriods, the biggest mover, and the one thing the lead should decide. 3. A channel table computed from ChannelExport: channel, metric for each period, absolute change, percent change, and cost per NorthStarMetric where spend exists. Show the arithmetic basis so it can be checked. 4. Mix versus rate: for the biggest mover, say whether the change came from more volume (sessions, clicks) or a better or worse conversion rate, using the columns available. 5. Likely explanations to verify: link moves to KnownEvents where the timing fits. Label each as "fits the data" or "needs checking", never as proven cause. 6. Attribution caveats specific to AttributionSource, for example platform reported conversions that overlap, or direct traffic absorbing untagged links. 7. Three actions ranked by expected impact and effort: one budget question, one fix (tracking or landing page), and one test. 8. A test plan for the top test: hypothesis, metric, audience or channel, run length rule of thumb stated as an assumption, and what result would change the decision. 9. Questions for the lead that the data cannot answer. Constraints: - Use only numbers in ChannelExport or arithmetic on them. Never invent benchmarks, industry averages, or totals. - Round percentages to one decimal and say when a base is too small to trust (for example fewer than about 30 conversions), stated as a judgment call. - Correlation is not cause. Write "consistent with", not "caused by". - Plain language for a non analyst reader.