How the work actually moves.
The single most common cause of a visible quality drop is moving all the volume on one date. This is the staged alternative, with the rollback triggers written down in advance.
Volume moves in five steps, not one.
Each tranche has an entry condition, a rollback trigger and a named owner on both sides. You can stop at any point and the previous state is still intact.
-
Tranche 0
Shadow — no live volume
The team watches live contacts and handles historic ones. Nothing reaches a customer.
- Entry: onboarding assessment passed
- Exit: 90%+ scored on practice contacts across two consecutive days
- Rollback trigger: none — no customer impact
-
Tranche 1
One channel, off-peak
A single low-risk channel during your quietest window, with every contact reviewed before it sends.
- Entry: tranche 0 exit met
- Typical size: 5–10% of total volume
- Rollback trigger: quality below 85% on any day, or any accuracy failure
-
Tranche 2
One channel, full hours
The same channel across the full window. Review moves from every contact to a heavy sample.
- Entry: five consecutive days at or above standard
- Typical size: 20–30% of volume
- Rollback trigger: quality below 88%, or SLA missed two days running
-
Tranche 3
Add channels
Remaining channels move across one at a time, never two in the same week.
- Entry: tranche 2 stable for ten working days
- Typical size: 60–75% of volume
- Rollback trigger: any channel below standard for three days
-
Tranche 4
Full volume, steady state
Standard sampling, weekly reporting cadence, and the first optimisation cycle begins.
- Entry: all channels stable
- Your internal team is released from cover
- Rollback: retained contractually, but now an exception rather than a step
What happens when a tranche fails.
Rollback is a planned state, not a crisis. Agreeing what triggers it — before launch, in writing — is what makes it usable rather than embarrassing.
Volume returns to the previous tranche
Within one business day. Your team resumes cover at the level it was at before the step.
The cause is classified within 48 hours
Training, documentation, system access, staffing or scope. The classification determines the fix.
Re-entry needs the same evidence as first entry
Not a promise that it is fixed — the same measured exit condition, met again.
It is reported, not buried
A rollback appears in the weekly pack with its cause and its remedy, like any other finding.
Your team is never surprised by it
Both sides have named owners who agree the call. Nobody discovers a rollback from a dashboard.
Twice on the same tranche stops the plan
A second rollback at the same step means the design is wrong, not the execution. We reopen Design.
What we need from you, week by week.
Transitions slip because of the client's availability far more often than the provider's. Naming it early is the fix.
| When | From you | Typical effort |
|---|---|---|
| Discover | Access to the people doing the work today; historic volume data | Half a day, plus a few hours of interviews |
| Design | Decisions on scope boundaries; sign-off on SOPs and the scorecard | Two to three review sessions |
| Pre-launch | System access provisioned; a subject-matter expert for product training | Access within days, not weeks — this is the usual delay |
| Tranches 1–3 | A named person to agree each tranche gate | A short weekly call |
| Steady state | Attendance at the weekly review | Thirty minutes a week |
A transition you can stop is a transition you can safely start.
Every tranche has a written rollback trigger and a route back to the previous state. That is what makes it reasonable to begin before you fully trust us.
What would your first tranche be?
Tell us your channels and volumes and we will sketch the tranches, the gates and the rollback triggers — before you commit to anything.