Sustaining a Gain After the Green Belt Leaves

Diagnosis - the drift has a plausible, identifiable mechanism: the knowledge-base index needs manual refreshing and no owner exists. That is an unassigned control, not a motivation problem. AHT rising 25 seconds and FCR falling 3 points within two weeks is consistent with a degrading search index, though it could also be regression toward the mean or loss of huddle discipline - both need checking.

Roll-out risk:

1. Volunteer bias. The 12 agents self-selected; they may be more receptive or more experienced. The 79% FCR may not transfer to 90 agents.
2. Unstable control. Rolling out a process whose support mechanism (index refresh) is unowned will replicate the decay at 7.5× the scale.
3. Huddle dilution. A 10-min huddle works at 12; at 90 it needs a structure (team-level huddles, escalation) that has not been designed.

Recommendation: do not roll out next month. First stabilise: assign an index-refresh owner with a defined frequency and a check, reinstate the huddle on the pilot team, and confirm AHT returns to ~5:20. Then roll out in waves (e.g. 20 agents), each with the same control in place, measuring against the pilot result.

A control plan that survives the Green Belt's departure needs, at minimum:

1. Ownership - a named person for each control (index refresh, huddle facilitation, data reporting), with a named deputy. 'The team' is not an owner.
2. Metric definitions and frequency - AHT and FCR defined exactly as in the pilot, reported weekly, with the chart (not just the number) reviewed.
3. Reaction plan - what happens if AHT exceeds a trigger (e.g. 5:35 for two consecutive weeks): who investigates, what they check first (index freshness, huddle attendance), and by when.
4. Documentation - the redesigned search procedure and huddle agenda written down so it does not depend on individuals.
5. Audit - a periodic check (monthly) that the controls are actually being performed, not just documented.
6. Handover evidence - a run of stable data (e.g. 6–8 weeks) at the improved level with the controls in place, not just the pilot result.

Without items 1 and 3 the project is not controlled - it is merely finished. The drift already observed is direct evidence that the control gap is real, and should be cited when requesting the handover sign-off.

Treat it as a hypothesis to test, not a claim to accept or dismiss.

• If it were motivational drift, you would expect a gradual, uniform decline across agents and metrics. The observed pattern - both metrics moving together within two weeks, coinciding with an unowned index refresh - fits a mechanism explanation at least as well.
• Test it cheaply: check index freshness dates against the AHT timeline; if AHT rose after the index went stale, the mechanism is supported. Compare individual agent trajectories - motivational drift tends to be diffuse; a mechanism tends to hit agents who rely most on search.
• Either way, the response is the same: assign the owner and the reaction plan. If it is motivational, an unowned control will not fix it either, and the huddle needs a named facilitator.

Avoid arguing about intent. Reframe: 'Whatever the cause, we have an unassigned control and a metric moving in the wrong direction. Let's fix the control and see if the metric responds - that also tells us which explanation is right.' This keeps the conversation on evidence and gets the handover signed off with a real control plan.