A monthly curve that appears to move may only reflect a changed definition: sample size, percentile point or collection window shifts all break comparability. Before comparing months, put the definition, denominator and collection window on the same specification card.
Archive that card with the data. When a definition genuinely changes, mark the change date and leave a visible break in the curve instead of silently joining two series. A month with no movement is recorded as-is; it is not a reason to swap definitions to hunt for growth.
This article compares no absolute values across shops and draws no conclusion from a single swing; it only asks whether each trend was measured with the same ruler.