The problem
New specialists had two options with a hard escalation: watch from a distance, or take it on alone. The first didn't teach much. The second went badly often enough that people stopped trying it.
How I measured it
# escalations handled by people in their first year, and how many needed a senior to take over
awk -F, 'NR>1 && $9<365 {n++; if ($10=="reassigned_senior") r++} END {print n, "handled,", r, "taken over"}' escalations.csv
# time from start date to first escalation handled alone, per person
awk -F, 'NR>1 && $11=="solo" && !(($2) in first) {first[$2]=$9} END {for (p in first) print first[p], p}' escalations.csv | sort -n
The first-year takeovers were high, and the time to a first solo escalation varied a lot from person to person.
Why it kept happening
There was no middle step. Learning an escalation means doing part of it with someone who can step in, and that step wasn't anyone's job.
The process change
Three stages for each type of escalation, tracked on a simple sheet:
- Shadow: join two escalations of that type. Take notes, ask questions after.
- Co-lead: run the next two with a senior on the call. The senior only steps in if asked, or if the customer's at risk.
- Lead: run it alone, with a senior reviewing the ticket afterwards.
A person moved up for one type of escalation at a time. Being cleared for one didn't mean being cleared for all.
What I learnt
Readiness is per type of problem, not per person. Someone can be ready to lead one kind of escalation and still need to shadow another. Tracking it that way took the guesswork out of who could take what.
What I do now
- Give people a co-lead step with a safety net before they go solo.
- Track readiness per type of problem, not as a single "senior" label.
- Review solo work after the fact, so the feedback lands while it's fresh.
stage 3 hands work over the same way → see the engagements