Skip to main content

Root cause analysis

info

View this example in Motif here.

This is an investigation into a change in the volume of watch_start events that occurred on June 24.

Video​

SOL query​

// remove events before June 1
match StartEvent()
if StartEvent.ts >= '2024-06-01'
replace SEQ with MATCHED >> SUFFIX

// for each event, if it occurred after June 24 add a flag that's true, otherwise false
match split Event()
set Event.in_problem_area = Event.ts >= '2024-06-24'
combine

// sessionize
match split Session()+
if duration(Session[-1], SUFFIX[0]) > 1h

// align sequences on home_page eventually followed optionally by watch_start
match home_page >> * >> watch_start?

//match split favorites_page >> Destination()

Key steps​

  1. First we looked at the drop in `watch_start events over time, but sessionizing and doing a simple match.
match split Session()+
if duration(Session[-1], SUFFIX[0]) > 1h

match home_page >> * >> watch_start?
  1. Then we removed all events before June 1, so we had a window just before and after the drop.
match StartEvent()
if StartEvent.ts >= '2024-06-01'
replace SEQ with MATCHED >> SUFFIX
  1. Then we added a flag to every event that denoted whether or not it occurred inside the problem area.
match split Event()
set Event.in_problem_area = Event.ts >= '2024-06-24'
combine
  1. That allowed us to use the Compare feature of the barcode to see a big drop in favorites_page_click events.

intensity

  1. We switched our analysis to be on favorites_page and the event immediately following it.
//match home_page >> * >> watch_start?

match split favorites_page >> Destination()
  1. We then plotted the number of favorites_page visits grouped by the name of the Destination page.

intensity

  1. Finally, we looked at the number of favorites displayed on the favorites page and see that for all users it dropped to zero.

intensity