A chart drops to zero overnight. Another suddenly shows twice as many enquiries. Both deserve attention, but neither immediately justifies a new marketing strategy. First establish whether the business changed, the measurement changed or only the presentation changed.
Save the unusual report with its date, period and filters. Record when the change first became visible. This snapshot helps distinguish a genuine alteration from a report that subsequently updated.
Put the comparison on the same basis
Check that both views use the same metric, dates and time zone. A partly completed day is not a fair comparison with a full previous day. A filter restricting the report to one page or channel can also explain an apparent decline.
Inspect the metric definition. Was a contact action newly marked as an important event? Are enquiries now counted by action rather than person? Did someone switch a report from visits to users? Each change can alter the number without changing how many people are interested in the service.
Allow for processing as well. Analytics documentation on data freshness explains why provisional reports may change later. Distinguish a plausible processing delay from an ongoing failure to collect data.
Look for changes on the relevant date
Review the website and campaign change log. Were pages published, forms replaced, measurement altered, links edited or consent settings adjusted? Timing alone does not prove causation, but it provides a lead that can be tested.
Test the actual journey on an appropriate device. Open the page, perform a clearly labelled test action and check its result in the responsible system. Avoid changing several settings at once. Otherwise, an apparent repair may leave the original cause unresolved.
If measured contact actions increase, compare them with genuine incoming messages. An event firing twice can manufacture growth in a report. Conversely, a working contact route can continue producing enquiries while measurement has failed. These situations require different corrections.
Make sure test activity is identifiable and excluded from business reporting. A debugging session with repeated submissions should not become the next unexplained spike.
Consider changes in the business
Was there an unusually popular post, a local event, a public holiday or an interruption to operations? Pursue explanations supported by specific evidence. One unusual day does not necessarily establish a trend. With small totals, individual events account for a large share of the result and make variation more visible.
Separate observation from interpretation in the investigation note. “The form was replaced on Tuesday, and no completions have been recorded since” is an observed sequence. “The replacement caused the failure” remains a hypothesis until testing supports it.
After a correction, document the affected period and its implications for previous reports. Do not invent missing values to produce a smooth curve. Mark gaps and explain which comparisons are limited. A visible gap is more useful for later decisions than a seemingly complete series that has no evidential basis.
