Forecasting Demand

The Forecast tab turns your historical data into a forward-looking demand prediction. This is the foundation of your staffing plan — everything downstream (requirements, schedule, cost) flows from the forecast.

How the Forecast Model Works

Turnella uses a multiplicative decomposition model — the same family used by most professional forecasting tools, well-suited to contact centre demand because it handles day-of-week patterns and intraday profiles cleanly.

The formula is:

Forecast = Base Level × Trend × Day-of-Week Index × Intraday Profile

Each component captures a different aspect of demand:

The model also generates confidence bands (P10/P90 range) showing how wide the uncertainty is around each point.

Generating a Forecast

1. Open the Forecast tab. The Generate forecast button is enabled once you have at least 4 weeks of data.
2. Choose your forecast horizon: 2, 4, or 8 weeks ahead. For most planning purposes, 4 weeks is standard.
3. Click Generate forecast. The model runs in a few seconds and displays a line chart with shaded confidence bands.
4. Review the output. Check that the day-of-week pattern looks right (Monday higher than Friday?) and the intraday profile makes sense for your operation.

What to look for

A good forecast should show:

• A realistic intraday shape — matching your typical peak hours
• A sensible day-of-week pattern — if Mondays are your busiest day, the forecast should reflect that
• Confidence bands that widen toward the end of the horizon — further-ahead intervals always have more uncertainty
• Trend that matches recent history — if volume has been growing, the forecast should trend up

If any of these look wrong, check your historical data first. An import error, outlier, or operating hours issue in the data will distort the forecast.

Manual Overrides

The model does not know about planned events: a marketing campaign, a product recall, or an office closure. Use manual overrides to adjust specific intervals.

Single interval override

Click any interval in the forecast table. An editing panel opens where you can enter an override volume (and optionally an override AHT). Overridden intervals are marked with a flag icon and are preserved if you regenerate the forecast.

Bulk date range adjustment

Click Adjust forecast to apply a percentage uplift or reduction to a date range. Examples:

• +30% for the week of a planned promotion
• −20% for the week between Christmas and New Year
• ×0 (zero out) for a bank holiday

Use overrides sparingly. If you find yourself overriding many intervals because the model is consistently wrong, investigate the underlying data quality or add more history.

Event Multipliers and Deflection

Event multipliers

For known recurring events (e.g., Black Friday, end-of-month billing cycle), configure event multipliers in Settings → Operating Calendar. Enter the date and a multiplier (e.g., 1.8 for +80% volume). The forecast applies this automatically whenever it predicts that date.

Contact deflection

If you are implementing self-service, chatbots, or IVR deflection that will reduce inbound volume over time, configure this in Settings → Deflection Config:

• Start deflection % — the reduction in month one (e.g., 0.3%)
• Monthly increase — how much deflection grows each month (e.g., +0.1% per month)

Turnella applies this as a downward multiplier on the forecast, shown as a separate line in the chart. Useful for modelling how self-service investment will change your future headcount requirement.

Forecast Accuracy Snapshots

A snapshot is a point-in-time copy of your forecast, locked before you know the actual results. After the period passes, you compare the snapshot to actual observations to measure accuracy.

Creating a snapshot

1. On the Forecast tab, click Save snapshot.
2. Give it a descriptive label: for example, "Week 24 — saved Mon 10 Jun".
3. The snapshot is locked. Re-generating the forecast does not overwrite it.

After the covered period has passed and you have imported actual observations, accuracy metrics are calculated automatically. View them in the Forecast Accuracy tab.

Understanding Forecast Confidence

Each forecast interval has a confidence level — High, Medium, or Low — based on:

• Data availability: how many historical observations exist for this day-of-week and hour
• Stability: how consistent the pattern is in the history (high variation = lower confidence)
• Horizon distance: intervals further ahead have lower confidence

Low-confidence intervals are shown with wider confidence bands. They are good candidates for manual overrides if you have operational knowledge the model cannot see.

Forecast Accuracy Metrics

WAPE is the primary metric for contact centre forecasting. It weights high-volume intervals more heavily — a 50% error at 02:00 barely moves the score, while a 10% error at your peak afternoon slot has significant impact. This is the right behaviour for operational planning.