Where schedules most often go wrong
The failure is rarely the tool
Most schedules that fail don’t fail because the WFM platform was wrong. The optimiser produced a defensible rota, the constraint rules were sensible, the configuration was reasonable. What failed was upstream of the tool: the assumptions the schedule was built on. Volume curve assumptions, shrinkage assumptions, skills assumptions, seasonal assumptions, and the assumption that the schedule designed in March will still fit the operation in October. Each of these is correctable if it’s recognised, and most operations don’t recognise them until something visibly breaks.
This series walks through the five failure modes that recur. Each is the subject of one of the next five pieces; this opening article names them so the diagnostic landscape is visible from the start.
The five failure modes
1. The flat coverage curve. The most common failure: a schedule that produces the right daily total of FTE, distributed perfectly evenly across a day that has a 3× peak. The operation runs over-staffed in troughs and under-staffed in peaks. The daily-average dashboard reports everything as on-plan. See designing for the average, not the curve.
2. The shrinkage gap. The schedule was built assuming 30% shrinkage. The actual is 36%. The six-point gap shows up as missed SL, the operation can’t explain it, and the diagnosis usually lands on “poor adherence” rather than the modelling failure. See the shrinkage assumption that’s always wrong.
3. The multi-skill illusion. The schedule treats agents as if they can all take everything; the pooling effect is built into the maths. The floor reality is that agents have nominal skills they don’t exercise, routing rules prefer primary skill, and the “pooled” team is actually two disguised single-skill teams. See the multi-skill illusion.
4. Seasonal drift. The schedule is excellent for the average week and broken for the peak weeks. December, World Cup, Easter, the marketing-campaign window, the regulatory deadline. The operations that handle this well design the steady-state schedule with peak absorption built in. See the schedule that’s right for the steady state and wrong for the peak.
5. The schedule that ages out. The rota was right at launch and silently wrong six months later. Volume patterns shifted, channels grew, the operation changed, the schedule didn’t. The dashboard says everything is on plan because the dashboard was built against the original assumptions. See the schedule that ages out.
Why this order matters
Operations that try to improve scheduling by upgrading the WFM platform first usually find the upgrade doesn’t deliver the SL or cost gains they expected. The bigger gains live in the assumptions the platform can’t see — the curve fit, the shrinkage realism, the pooling reality, the seasonal design, the drift management. A scheduling team that audits its coverage curve, validates its shrinkage assumption, tests its multi-skill pooling, builds in seasonal absorption, and runs a quarterly review typically lifts schedule efficiency by enough to defer the platform upgrade entirely. A team that upgrades the platform without addressing the assumptions usually finds the new tool produces the same flawed rota faster.
The honest exception
Some operations genuinely have a tool problem. A spreadsheet-driven scheduling process at 200+ agents; a legacy WFM platform that doesn’t support modern shift patterns; a configuration so locked-down that the planner can’t test alternative designs. In those cases the tool is the bottleneck. The diagnostic to distinguish — can you produce a measurably better schedule by changing the assumptions inside the current tool, before you upgrade? If yes, the tool isn’t the problem. If no, it may be.
The series ahead
The next five pieces each take one failure mode. Each walks through what the failure actually looks like in operations, why it’s so persistent, and what the operations that catch it early do differently. As with the forecasting series, none of the fixes is technically difficult; all of them are operationally unglamorous. The reason most operations still struggle with scheduling isn’t that the optimisation maths is hard. It’s that the disciplines around it are.
Next in the series: Designing for the average, not the curve.
Pair this with fixed schedule rotation, multi-skill scheduling, and the schedule review meeting.