Question 1

How does Forecasting predict how busy we'll be?

Accepted Answer

It reads up to two years of your own conversation traffic and learns your weekday baselines and seasonal patterns, then layers on the events that move volume — public holidays, school holidays, sales events like Black Friday, and your own campaigns. Every night it recomputes the next 90 days (the engine projects a full year), classifying each day Low, Normal, Busy, High, or Peak against your typical weekday.

Question 2

How much history does it need before the numbers are useful?

Accepted Answer

About 60 days of traffic is enough to start. It gets sharper the more history it has — with one to two years, it has seen each seasonal event ramp up and decay for you, so the forecasts and confidence ranges tighten. A brand-new event it has only seen once is treated as a guess, not a forecast, and labeled that way.

Question 3

Can we break the forecast down by channel, team, or topic?

Accepted Answer

Yes. Every forecast splits by channel (chat, email, voice), by team, and by topic, so you can see where the volume is actually landing rather than one blended number. That's what makes the staffing suggestions per channel meaningful.

Question 4

How are the staffing suggestions calculated?

Accepted Answer

Forecast volume divided by your team's real, observed resolution throughput over the last 90 days, per channel — which turns a volume forecast into suggested agent-days. Where a forecast day exceeds your recent typical capacity, it raises a peak alert, like "Nov 28 needs ~9 agent-days on chat; recent typical is ~6."

Question 5

Does Forecasting schedule or roster anyone?

Accepted Answer

No. It suggests staffing and flags the peaks — it never schedules a single person. The suggestion is yours to act on in whatever workforce tool you already use. Keeping a human in the loop on scheduling is deliberate.

Question 6

What happens when volume spikes for a reason you don't recognize?

Accepted Answer

Rather than hide it in a bigger number, the forecast labels the leftover lift "Unexplained." You can name that event once — a local promo, a press mention, an outage — and the model learns its uplift and ramp shape so the next occurrence is explained, not a surprise.

Question 7

Can I override the forecast when I know something it doesn't?

Accepted Answer

Yes — overrides are first-class. When you know about something the history can't (a one-off campaign, a market you're entering), you set it, and the day's math ties out around your input. Naming an unexplained event is the everyday version of this.

Question 8

Does this replace the forecast spreadsheet we keep by hand?

Accepted Answer

That's the point. The demand forecast your team has been maintaining in a spreadsheet — copying last year's numbers, guessing at the holiday bump — gets rebuilt every night from your real traffic, with every day tracing back to a cause you can see. You stop maintaining the spreadsheet.

Know how busyyou're going to be

Grounded in your history

A forecast that shows its work

Staff the peaks before they hit

The calendar your traffic already follows

Conversations

Analytics

Agent Workspace

Monitor

Expected traffic, and where it comes from.

See Atender in action