1. Concurrent breaks should never go over 25% of available FTE on the day if possible.
2. Meetings with the exception of One to Ones should never be back to back if at all possible.
3. Start time increments to be staggered by 15 minute periods where possible.
4. Staggered finish times where possible.
5. Training to have documented needs if scheduling issues arise from it.
6. Core staff for each major unit at core times e.g. Largest Line 0900 - 1700
7. Actively monitor lateness and shift finishing times and feedback to Ops (Ops would be expected to do this as well, but the tendency will be to look at the overall % rather than the individual)
8. All Schedules are examined v shape of the day (call) profile.
9. Annual leave limits to be clearly defined and available to appropriate staff.
10. Schedule release cycle clearly defined and acknowledged by all.
11. Plan in contingency if at all possible, often referred to as ‘Back-Pocket’ by Resource Planners.
12. Clear defined process for feedback/changes on released schedules.
13. Adjust breaks, lunch, meetings, training sessions, after call work or other activities in real-time when there is a need to keep agents on the floor.
14. Test scheduling accuracy with dry-runs and be sure to include common scenarios as well as scenarios that can be considered plausible outliers.
15. Small pre-templated emails should be sent with a summary of lost or gained time after large meetings (no tone just fact)
16. Remember That the Distribution of Agent Shifts Isn’t Fair - From a scheduling perspective, we must try to treat all agents fairly and consistently, but don’t slip into the trap of believing an even distribution of shifts is fair. In addition, we shouldn’t just assume that we know what fairness looks like without directly asking agents what their preferences are.
17. Avoid Knee-Jerk reactions.
18. Remember that for non-analysts statistics rarely lead to factual conclusions - even with a full self-service setup (best in class all in this manner) people who are not experienced Analysts or similar will almost always reach the wrong conclusions on complex data.