Checkout funnel drop-off analysis loves clean weeks. Reality prefers drama: a carrier API stalls, delivery options spin, and shoppers leave. If you leave that week unmarked, next quarter’s comparison becomes folklore.
Stamp the broken days
We draw a vertical band across the affected dates and name the incident in the legend — date range, carrier, and whether the storefront showed an error or an infinite spinner. The drop-off numbers remain; the interpretation changes.
Resist the tidy delete
Deleting the week feels neat and ruins seasonality. Keep the data, widen the confidence language in the brief, and — if needed — show a second funnel with the outage week excluded, clearly titled as a sensitivity view.
Hand the stamp to operations
The best annotations come from people who filed the incident, not from analysts guessing after the fact. During intake we ask for a one-line outage log. It takes five minutes and saves an hour of speculative storytelling later.