Anomalies in data are sudden and unexpected changes – a drop in traffic, zero conversions, a sharp jump in bounce rate, or the loss of one channel. Most companies only discover it at the monthly report, by which point a whole month is lost. The solution is automatic alerts set on key metrics: a fall in conversions below a threshold, zero form submissions in a day, a significant change in traffic compared with the same day the previous week. Set alerts for things you will actually check, otherwise they become noise that everyone ignores. The most common cause of anomalies is not the market, but a technical change on the website – the deployment of a new version that removed the tracking code. That is why you should check tracking after every deployment. Also set alerts for campaigns, where a tracking outage directly affects automated bidding strategies.
See also: Data quality in analytics, Website migration without losing rankings, Measurement plan.
