Signal Loss

Signal Loss explained

A dashboard reports fewer completions after a website change. Before reducing spend, establish whether orders really declined, an event was renamed or a changed choice means it is transmitted less often. First align periods, time zones, counting rules and business events. Differences between systems are a reason to investigate, not conclusive evidence of a particular cause.

Privacy choices are not faults to bypass. Apple requires ATT permission for its defined tracking across other companies’ properties; alternative identifiers do not replace that permission. Web data also remains subject to the applicable conditions for device access and personal-data processing. Server-side transmission may improve a technical data path, but it does not automatically make missing observations available or their use permissible.

Models can estimate relationships that are not directly observed. Google Analytics explains that reports can combine observed and modelled key events and that modelling requires suitable data. An estimate is not a retrospectively observed individual journey. Document modelled components where disclosed, changes and comparison limits. Even more complete tracking would not turn attribution into causal evidence.

Creative Engineering starts with the decision the data should support. We take responsibility for the concept and quality. AI can organise anomalies and possible explanations; diagnosis must be checked against actual events. Combine permissible operational data with appropriate experiments or aggregate methods. The objective is a defensible decision within known limits, not a claim of complete surveillance. Assess data quality and total measurement effort together.