Attribution and conversion are not the same as profit
A fictional Swiss online retailer runs a campaign with fifty thousand impressions, one thousand clicks, and two thousand francs of advertising expenditure. The reporting platform assigns eight thousand francs of revenue to it. The shop later discovers that some purchase events fired twice after page reloads, and that the platform and shop use different attribution windows. A separate correctly measured experiment has one hundred people in each condition, with twelve purchases in the test group and eight in the control group. These small-sample observations are not a statistically calibrated proof of any population effect. The campaign also points mobile users to a slow form and includes an expired offer. Management wants to announce a guaranteed four-hundred-percent profit because the platform reports that revenue-to-spend ratio. The team must reconcile definitions, repair the customer path, and distinguish ratios from causal and financial conclusions.
1. Berechne CTR, CPC und den ausgewiesenen ROAS. Warum ist letzterer kein Gewinnbeweis?
Read the worked answer
CTR is 1,000/50,000=2%; CPC is CHF2,000/1,000=CHF2. Reported ROAS is CHF8,000/CHF2,000=4, or 400%. It uses attributed revenue, not net profit, and duplicate events, attribution scope, product costs, returns, and incrementality require review.
2. Wie ist der Testunterschied auszudrücken, und was muss vor einer Erfolgsmeldung geschehen?
Read the worked answer
Observed rates are 12% and 8%, differing by 4 percentage points, or 50% relative uplift. Sampling uncertainty remains. Reconcile event definitions and duplicates, verify windows, fix expired and slow destinations, and evaluate relevant net outcomes before claiming success.