Avoiding complacency in forecast accuracy and service-level delivery
The dangerous moment is right after a good run
Most forecasting and SL-delivery problems don’t start with a crisis. They start with a good quarter. Confidence in the model rises; the questioning slows down; the seasonality multipliers stop getting refreshed; the contact-mix assumptions calcify into “the way the operation works.” Three or four quarters later the conditions have shifted and the team is missing without quite knowing why. The complacency is invisible while it’s happening and obvious in retrospect.
This article walks through where complacency creeps in, the warning signs, the disciplines that prevent drift, and the leadership behaviour that keeps a successful planning team intellectually honest.
Where complacency creeps in
Model parameters. The smoothing factor, the seasonality multipliers, the day-of-week weights. Set once, lightly adjusted, rarely re-derived from current data.
Contact-mix assumptions. The implicit assumption that this year’s reason mix is last year’s. Holds for stable operations and breaks for anything going through digital deflection or product change.
AHT bands. The per-skill, per-channel AHT used in capacity calculations. Drifts up or down quietly with mix change; rarely re-baselined.
Shrinkage assumption. The compounded shrinkage figure the capacity model uses. Recalibrating annually is normal; mid-year reality often moves several points and gets missed.
Behaviour assumptions. “Most customers wait the full SL target before abandoning” or “adherence runs at 88%” — assumptions baked into the model that quietly stop being true.
Warning signs the team has stopped questioning
Four signals that complacency has set in.
1. “That’s the way the operation works.” When operational assumptions stop being framed as assumptions and become statements of fact, the questioning has stopped.
2. The model documentation hasn’t been refreshed in 18 months. If nobody can quote when the seasonality multipliers were last reviewed, they’re overdue.
3. Accuracy at the aggregate looks good; accuracy by segment hasn’t been checked recently. Aggregate accuracy can stay good while specific contact reasons or skill groups drift sharply.
4. The team can’t name the three biggest risks to next quarter’s forecast. If the answer is “nothing specific,” the team has stopped looking.
The disciplines that prevent drift
Five practices that keep accuracy honest.
Quarterly model review. Not the accuracy report — a structured review of the assumptions feeding the model. What’s changed in the operation? What in the model should change to reflect it? 90 minutes per quarter; pays back many-fold.
Accuracy by decomposition. Track WAPE not just at total but by contact reason, by channel, by day-of-week. Drift surfaces in the segments first; the aggregate hides it.
The “why is this still right?” question. Built into the planning rhythm. Once a month, pick one assumption and ask: why is this still right? Document the answer. Move to the next assumption next month.
External calibration. Compare your assumptions to industry benchmarks at least annually. Not to copy them — to notice where you’ve drifted from the wider sector.
Post-mortems on big misses. The biggest forecast miss of the quarter gets a structured post-mortem. What went wrong, which assumption failed, what to change. Often the most-valuable learning the team has.
The leadership behaviour that keeps a team honest
Three habits separate planning leaders who keep their teams sharp from those who let complacency settle.
Ask “what would have to be true for this to be wrong?” rather than “is this right?”. The inversion forces the team to surface the assumptions they’ve stopped seeing.
Reward the team that catches its own drift. Publicly. The signal that intellectual honesty is valued, not just hitting numbers, shapes culture.
Build slack into the planning rhythm. A team running flat-out has no time to question. The discipline depends on space to think.
Conclusion
Forecast accuracy and SL delivery don’t fail in big steps. They drift, quietly, in the months after a good quarter. The disciplines that prevent the drift are small, regular, and culturally enforced — quarterly model reviews, decomposition tracking, the “why is this still right?” habit, external calibration, post-mortems. Operations that build them stay accurate over years; operations that don’t cycle through accuracy and missing in quarterly waves.
Pair with forecast accuracy probably better, Poisson and natural noise, forecast with ranges, capacity planning when mix is changing, and showing planning team success.