Manual audits are certainly possible and a systematic method to do this is described in detail in the book Successful Analytics by Brian Clifton (PhD). But as Brian points out, its painful and prone to errors. Apart from deploying a valuable analyst to a mundane job – wasting their expertise when they can be helping you grow your business – errors and data omissions are actually difficult to find. They are needles in haystacks!
The main challenge for auditors is the data volume. For example, personal data is often a very small percentage fraction of the total data collected i.e. much less than 1%. But that small fraction can be costly in terms of reputational damage to your brand, as well as get you into trouble with data protection laws, such as GDPR.
Other examples include campaign tracking parameters that go missing or become corrupted. Without these consistently in place, attribution modelling falls apart. But humans struggle to notice this due to the shear volume of campaign data constantly coming in. Ditto for e-commerce – would your team notice if 5% of transaction were being duplicated in GA? Such an error can have a huge impact on interpreting conversion rate and ROI…
All of this requires a constant scanning and verification of data – perfect for machines and not suitable for humans! Analysts are trained to find data points that have value to the business – this is when humans outperform machines.