12. Forecast Potential - a mathematical calculation that reviews the trending forecast performance and estimates future forecast potential within a banded range using upper and lower levels from the live forecast.
13. Forecast Accuracy - Forecast minus actual, divided by actual. This is the guideline by which a decision is made to change the current forecast method.
14. RAE (Relative Absolute Error) - all forecasting choices are run simultaneously and measured using RAE for the best possible result.
15. Net Staff's relationship to repeat callers - used to bump up potential abandons when understaffing starts happening.
16. Confidence Level - measured from the MAPE (Mean Absolute Percent of Error) of the recent forecast accuracy.
17. Baselines - every forecast is bounced against a baseline forecast so that you always see the amount of movement with each reforecasting trip.
18. Extrapolated weekly volume s that go down to the daily levels, and interpolated daily volumes to bring everything back up to a monthly level.
19. Finally, a hefty dose of conditional formatting rules to scream alerts when something goes wrong.
7. Annual Growth Rate - we measure the percent of change that is seen in the forecast group's data year over year.
8. Circannual Growth Rate - we measure the percent of change that is seen in the last 365 days to the 366-730 days before that, and keep that number rolling forward.
9. Seasonal Growth Rate - we measure the instant trend based on a snapshot of live recent history and utilize that for short-term forecasts.
10. YTD Growth Rate - we measure the performance between the data received so far this year against where that data was at the same time periods last year.
11. Average weekly volume - good to use in a pinch if there's an unidentified impact happening.
Multiple Growth Rates
Intense Forecast Accuracy
© Human Numbers, L.L.C. 2008-2023 | All Rights Reserved | Legal and Privacy Statements
We produce carefully hand-crafted forecasts and schedules, tailored specifically to your call center’s experiences and its agents’ preferences. We can accommodate schedule bids, special rankings, and lunch/break optimizations. Our schedules are produced two weeks out, and if you let us know about special plan coming up we can incorporate meetings, trainings, huddles, and planned vacation into your schedule runs. Human Numbers' forecasts consistently outperform those generated by WFM Software. The reason is we start with a baseline of our unique 19-point forecasting methodology using time-series analysis and we top that off with a customized layer specifically built around your own call drivers.
In our experience that WFM software does a very impressive job with the heavy lifting of drilling down a forecast into time-of-day arrival patterns. It also manages AHT forecasts for interval scheduling at an optimum rate. But that's as far as we let our forecasts go with automated WFM software, because our Excel-based forecasting models have always been able to do a better job. Even with a state-of-the-art software solution, you still need a highly-skilled workforce manager to methodically capture these types of things and adjust it in the forecast. After the event is over, that data needs to be analyzed, documenting what just happened, normalizing its effect on the trends/patterns, and archiving it for repeat offenses against the forecast in the future. If your call center is small, the ROI to purchase expensive software doesn’t always pay off.
After the math does its part, there's still room left to top it off with special manual adjustments (ex. campaigns, event-drivers, etc.) and we even reserve a special yearly cycles with average weekly volume in a 6-season calendar period in case things get really rough. We know there are very specific and complex factors that must be considered. Emergencies, bizarre moves by the Marketing department, surprise electrical outages, and other unpredictable activities can wreak havoc with your forecast. Holidays do not consistently occur on the same day of the week, nor even the same week of the year for that matter. There’s not always a mathematical precedent to rely on, and that’s where we come in.
When you let us take over this responsibility of forecasting and scheduling for you we will provide you with carefully thought-out forecasts, and we will work closely with your team to generate the best possible schedules for their lifestyles.
WFM Software has the raw data and potential power to measure every one of these points, too. But it doesn't look at them as a human does and make an experienced judgement to decide when to start using one and stop using another, which gives Human Numbers the edge.
1. Our historical data uses both raw and cleaned history. Cleaned history excludes excessive abandons and unusual outages, storms, or spikes. Each of these serve different purposes in comparison and analysis - there's no reason to limit to a single data set.
2. Percent of Week - each week of the year is different; out of the 52-week calendar year, up to 24 of them may have holiday impacts, and even the normal weeks experience their own seasonal impacts which must be considered.
3. Holidays and Cycle Days - a group may have different volumes around billing cycles or monthly events.
4. Rolling Daily Volume - needed to calculate the week of year, and also good for major changes introduced to a forecast group.
5. Abandonment rate - needed to identify how much volume needs cleaning.
6. Time of Day % and Day of Week % - every day of the week brings a different percentage, as does every day of a holiday week, and usually the week before/after the holiday. We use weighted moving averages, regular averages, and standard deviations to choose the most precise result.