Forecasting with events: holidays, campaigns, outages

A contact centre forecast almost always starts with history. How many calls came in on Monday mornings at ten over recent weeks? How many chats on Friday afternoons? These patterns become the expectation for next week.

That works well for ordinary weeks. But ordinary weeks are rarer than you might think. A public holiday, a billing run, a campaign or a network outage changes volume more than any day of the week.

A forecast that does not know about these is wrong exactly when it matters most.

#What counts as an event

For planning purposes, an event is anything that foreseeably changes contact volume and does not follow from the normal weekly rhythm. There are three groups.

Each group calls for a different approach.

#Calendar: known, but not uniform

Calendar events look simple. But in Germany, public and school holidays are set by each federal state. Corpus Christi is a public holiday in some states and not in others. Summer holidays start at different times depending on the state.

If you serve customers across the whole country, you will not see one clear holiday effect but a blend. For the forecast, that means recording holidays by region and weighting their effect by each region's share, not applying it across the board.

The day after counts too. After long weekends, requests that piled up come in. A good event therefore describes not just the day itself but the days around it.

#Your own activities: the knowledge sits elsewhere

The second group is the most plannable, and still the one most often missed. Marketing plans a campaign, billing sends out new invoices, sales changes tariffs. The contact centre finds out when the calls arrive.

The bottleneck is not technology but the flow of information. A shared event calendar works well. Other departments enter what they are planning, with date, audience and channel. Planning translates that into expected contacts.

#Disruptions: unpredictable, but prepared for

Nobody can forecast a network outage. But you can know how it behaves when it comes: a steep rise in the first hour, many similar requests, a slow decline.

So disruptions do not need a forecast; they need a scenario. Capacity planning works out what doubling calls for two hours would mean, and which measures help.

Typical measures are an announcement in voice self-service, a status message on the website, extra shifts and callbacks instead of hold. How to communicate in such situations is covered in Peak load and outage communication.

#Cleaning up the history

Events work in both directions. They change the future, and they also distort the past. A week with an outage in the history makes next year's forecast for the same week too high, unless someone flags it.

That is why event planning includes looking back. Past events are marked so that the forecast can exclude them or reuse them deliberately. Clean history is the foundation of every good prediction.

Weekly schedule

CoverageService level ✓

MonTueWedThuFriSatSun

Early

Late

Flex

Shift swap

Thursday late for Friday early

Approve

#Judging accuracy honestly

A forecast will never be exactly right. What matters is where and how it misses. A deviation on a quiet Tuesday afternoon costs little; a deviation on the Monday after a billing run costs service level.

So compare forecast and actual volume separately for ordinary days and event days. If event days are regularly off, what is usually missing is not a better method but a piece of information: an event nobody reported.

Granularity matters too. A daily forecast can be right while the quarter-hours around ten in the morning are clearly underestimated. For scheduling, the intervals count, not the daily total.

#From forecast to schedule

A forecast only becomes useful once it is turned into shifts. Scheduling derives from the expected volume per channel and skill how many people are needed when to hold the service level.

Events play out differently here. A known holiday goes into the plan early. An announced campaign may need preferred shifts or swaps. A disruption needs a short-notice response. Shift swaps as self-service shows how swaps help.

#Who contributes what

Forecasting with events is teamwork. Planning knows the methods, but knowledge about upcoming events is spread across the organisation.

The guide Workforce planning in the contact centre covers the basics of forecasting and scheduling. The glossary explains forecasting and workforce management.