Define the unit and exclusive end of the window
Synthetic measure flags a case with a further customer contact within twenty-four hours of its first contact, excluding equality at the end. C-4 opens T-401 on 20 October 2026 at 09:00 and contacts again at 10:00. C-4 opens a different occurrence T-402 at 11:00. Another T-401 contact arrives on 21 October at 09:00, exactly outside its window. The measure is cases with repeats, not total messages or a customer-level flag.
| Contact | Case / time | Relationship | Repeat flag effect |
|---|---|---|---|
| P-401 | T-401 / 20 Oct 09:00 | First contact | Start T-401 window |
| P-402 | T-401 / 20 Oct 10:00 | Same case within 1 hour | T-401 has repeat |
| P-403 | T-402 / 20 Oct 11:00 | Different occurrence; first contact | Not a T-401 repeat |
| P-404 | T-401 / 21 Oct 09:00 | Exactly 24 hours | Outside measured window |
Group by case before applying the window
- Identify distinct case occurrences
- Locate each first contact
- Compare further contact instants
- Count cases with eligible repeats
Unassigned contacts need case association before they enter either numerator or denominator.
How many cases have repeats, out of how many distinct cases?
Show the worked answer
One of two cases, 1/2 = 50%. T-401 has the 10:00 repeat; T-402 has only its first contact. The exact twenty-four-hour contact adds no eligible repeat.
Keep first contact, case identity and the exact window boundary. A repeated contact does not itself explain whether the support outcome was good or poor.