Our panel of experts share their workforce optimization (WFO) best practices, highlighting key methods of improving your contact centre forecasts.
1. Assess How Your Channel Mix Is Changing
Historical data should be used to estimate the volumes of each channel – voice, chat, text and social media – so that you can accurately determine the level of skill and number of advisors you need in the contact centre at any given time.
The more channels customers use to interact with a business, the more contact centres need to plan for the right advisors to be on shift at any given time.
By looking at the data, you can determine peak periods for customer queries, channel usage and the right combination of skills needed.
By looking at the data, you can determine peak periods for customer queries, channel usage and the right combination of skills needed.
For example, on weekends, customers may prefer to contact businesses via social media and on Mondays there is a peak in telephone contacts.
You need to have advisors capable of handling these types of interactions, while ensuring balance across the other channels.
2. Engage the Wider Business in Controlling Shrinkage
Shrinkage or unproductive time in a contact centre that reduces the productive time advisors are available is one of the overlooked metrics in the industry.
Many businesses count their shrinkage by averaging it across the year, which makes results look flat week over week. This is usually a mistake on the part of the planner.
Many businesses count their shrinkage by averaging it across the year, which makes results look flat week over week. This is usually a mistake on the part of the planner, as shrinkage is both a seasonal occurrence, for uncontrollable shrink, and a strategic decision, for controllable shrink.
To ensure that these risks are controlled, contact centre planners and management need to have an open dialogue about the ongoing staffing levels.
For example, a planner forecasts overstaffing weeks in advance and alerts management to offer staff extra training or unpaid vacation to prevent it.
Conversely, if management plans to pull advisors off the phones to work on side projects, planners should be advised weeks in advance to expect to be understaffed during those times and prepare operations for overtime or other contingencies.
3. Avoid Delayed Hiring
Variance analyses are commonly used to show the difference and root cause between actual and budgeted or targeted levels of performance.
These analyses can be used as an early warning sign that there could be a permanent issue affecting a contact centres operation when volumes, handle times, attrition or shrinkage are different than planned.
Where these analyses can be useful in identifying potential issues, many executives still tend to be slow to address these warnings and delay hiring more advisors until it’s too late and the contact centre is overwhelmed and underperforming.
To avoid delays in hiring, contact centre executives need to ensure that they examine variance analyses and cross reference these findings with other data in their disposal.
Findings from these comparisons can then be used to manage the staff levels within the contact centre accordingly.
Thanks to Ric Kosiba at Genesys
4. Track Correlating Company Data
“If you just focus on the smallest details, you never get the big picture right.” – Leroy Hood
Every event within a company has the ability to impact your forecast, from marketing initiatives that could increase volume, to delays in your company’s fulfilment process.
These tracking efforts don’t need to be a huge initiative, but as events occur within your organization, something as simple as putting a flag next to significant historical data would highlight key events that could help improve your future forecast efforts.
5. Track Correlating External Data
“Insurance companies also see more activity in the aftermath of extreme weather events as people worry about ‘what if’ it happened to them.” – Amelia Xu
Tracking external data can be hugely beneficial, but you have to select the type of data that would be most impactful to your business.
In general, macro-level economic data points are a good place to start and, for many types of businesses, the weather can play a role.
No, I don’t expect you to maintain weather forecasts as part of your contact centre planning process, but when there is a major weather event it’s worth noting it in your historical data.
Tracking these external events and influences will assist you in understanding the potential impact of similar events in the future.
6. Use Multiple Forecasting Methods
“Fast is fine, but accuracy is everything.” – Wyatt Earp
If you had to select one, what’s more impactful, a great goalie or a great striker? You might say both are equally important, but the real answer is that it depends on the situation or opponent.
Establishing multiple forecasting algorithms ensures that every forecast is run through different models that validate the best path for the given situation.
The same philosophy should apply to contact centre forecasting. Establishing multiple forecasting algorithms ensures that every forecast is run through different models that validate the best path for the given situation.
The use of multiple algorithms has the potential to allow your forecasting practice to always evolve with changing patterns and behaviours without any extra effort from your team.
7. Re-forecast Often
“Continuous improvement is better than delayed perfection.” – Mark Twain
You should view your forecast as a never-ending rolling projection of volume. As such, re-forecasting as often as every 15 minutes will provide a continuously improved view of your near-term trends.
Additionally, it allows you to see the gradual shift and required corrections in your forecast over the medium and long term.
The ability to make adjustments easily and often provides your greatest opportunity to fine-tune your labour utilization, advisor schedules and shift activities.
If your forecasting solution does not allow for frequent, and ideally automatic, re-forecasting, it might be time to shop for one that does.
The ROI on improving labour utilization scales rapidly and can pay dividends for a very long time.
8. Change the Granularity of Your Forecast
“Aim small, miss small.” – Mark Baker
Forecasting at a more granular level than what you need provides an opportunity to improve how you schedule your staff.
The Law of Large Numbers tells us that utilizing larger pools of numbers than necessary will produce more accurate results.
You should consider capturing data at the finest level (1- to 15-minute intervals) and then forecasting at a slightly larger interval (15- to 30-minute intervals).
In terms of using this to improve your forecast, you should consider capturing data at the finest level (1- to 15-minute intervals) and then forecasting at a slightly larger interval (15- to 30-minute intervals).
While this may seem counter-intuitive, the law generally proves out when tested.
Changing the granularity can be a complex undertaking but the improvement to your forecast accuracy will be well worth it.
9. Keep a Close Eye on Your Forecast Accuracy
“If you can’t measure it, you can’t improve it.” – Lord Kelvin
It’s imperative that you track your forecast accuracy score, regardless of whether or not your forecast solution or process tracks it for you.
Within a few weeks, you’ll have trending data that enables you to recognize your areas of opportunity with increasingly better clarity and precision.
10. Conduct Regular Post-Mortem Reviews of the Forecast
“Sometimes you gotta just chill. You gotta chill your thinking process.” – Ghostface Killah
Setting dedicated time for the team to review the variance between your forecast and actual results places a clear priority on refinement.
Beyond that, these post-mortem meetings provide a forum for the entire group to contribute to other variables that should be factored into future forecasts.
While workforce management (WFM) systems have come a long way in reducing the effort and manual workload by live humans, these meetings allow you to step back and evaluate how well the entire system and process performed. This helps to ensure that you’re not simply relying on technology to magically handle everything.
Thanks to Patrick Russsell at 8×8
11. Evaluate the External Factors That Influence Your Forecast
It is important that the team managing your forecast has a good line of communication with other functions in your organization whose actions can have a huge impact on your contact volumes, such as Sales and Marketing.
When forecasting, it is important to understand what these functions have done in the past in order to correlate any volume anomalies to past marketing or sales events.
Also, you must understand what initiatives other departments, like Marketing and Sales, have planned for the period for which you are forecasting.
You must understand what initiatives other departments, like Marketing and Sales, have planned for the period for which you are forecasting.
For example, looking at your historical data, you see a spike last September. At first glance, this might appear to be an anomaly. However, in talking to Sales and Marketing, you’re informed that they executed a mailer campaign in September that triggered an increase in contact volume, but they don’t plan to do it again this September.
With this knowledge, you can accordingly omit this from your historical data, so that your staffing plan for this September is not skewed.
In that same conversation, they inform you that they are planning on communicating a product recall this upcoming month. You can then adjust your forecast to ensure you are staffed appropriately to accommodate the influx of contacts.
Thanks to Lauren Comer at NICE inContact
12. Clean Your Data
In order to keep a record of accurate forecasts, you need to ensure that historical data is as ‘clean’ and free of anomalies as possible.
No one can predict a freak storm and, if such an event has affected your volumes, then it needs to be removed or replaced with ‘typical’ data for that period.
For example, no one can predict a freak storm and, if such an event has affected your volumes, then it needs to be removed or replaced with ‘typical’ data for that period.
At the end of the day, if the data that goes into your system is bad, the outcome will be bad. So, you need to be cleaning your data continuously, in order to improve the high quality of your outputs.
13. Remember That Collaboration Is Key
If you’re launching a new product or your business is taking a different direction, sometimes your historical forecasts will not be as helpful. You need to intervene; you need to apply human knowledge.
The trouble is, as a forecast “expert”, this does not necessarily mean you’re an expert in your employer’s business or market trends etc.
You need to network, you need to talk to business strategists or product managers. They’re the ones who will know how the new product will perform and how many sales they can expect from their new campaign. It is not necessarily your job to know this.
Your job is to take this knowledge (expected call-in rates, response rates etc.) and layer it into applicable available historical data to create a ‘best estimate’ hybrid forecast.
With this in mind, the forecast analyst role should be collaborative, liaising with experts in other areas of the business to produce robust short-range and long-range forecasts for a scheduling and recruitment strategy.
Thanks to Alex O’Donovan at Business Systems
14. Assess the Latest Methods of Forecasting
There are a number of best practices that you can use to improve contact centre forecasts, such as:
- Measuring variance at the right intervals
- Communicating with the marketing team
- Preparing for frequent changes to your staffing plan
But let’s take a step back and consider what are the latest methods of forecasting advisors in the contact centre world? It’s good to know these, so that you are getting the bread and butter of your forecasts exactly right.
Triple exponential smoothing is the traditional method, as contact centres can use spreadsheets to split their data into three basic parts: level, seasonality and trend.
However, this technique is very time-consuming, with many contact centres now choosing to use WFM systems with sophisticated neural networks.
These AI-infused systems will automatically run through a huge number of contacts across different channels and try to match the next piece of data to the forecast.
Thanks to Anand Subramaniam at eGain
15. Understand the Impact of Multiskilled Advisors
One challenge in workforce management is understanding the impact of multiskilled employees on required lines – or the number of full-time equivalent (FTE) workers needed to meet customer service objectives.
Most of today’s forecasts rely on a statistical model called Erlang, which can be a great, cost-effective method for traditional contact centres. However, it does work on the back of two assumptions:
- That all employees share a homogeneous skill assignment
- That work items queue to a single skill profile
If your contact centre does not operate under these assumptions, your forecast accuracy will suffer, which will either affect your contact centre’s ability to respond to real-time customer demand or cause unnecessary costs.
Modern WFM systems help to avoid this problem by leveraging machine learning models that predict the unique demand of each required line.
Thanks to Paul Chance at NICE
16. Use Technology to Better Forecast for Flexible Working
Flexible working is enjoying a comeback as staff and managers recognize its benefits, for example job satisfaction, increased work–life balance and less stress for advisors.
All of these mean higher productivity, less turnover and reduced absenteeism and the ability to retain the best staff and save money at the same time.
Automated WFO technology has the power to support an increasingly virtual, mobile workforce through robust forecasting.
With complete visibility of contact centre operations, pool available resources by shift type, in different functional areas, countries and even languages.
Thanks to Peter Dempsey at Puzzel
17. Understand Where Customers Are Likely to Expect Delays
Understanding the mix of topics, emotions and outcomes for different days of the week and time of the day can make a huge difference.
For example, do people use internet banking on a weekend? Are customers (and agents) happier at the beginning of the day/week?
By analysing incoming queries for topics, sentiment and emotional intents, a bank was able to understand when customers were more likely to tolerate delays and issues. This allowed them to forecast and staff more strategically, while flexing their skills and resources accordingly.
Thanks to Dan Somers at Warwick Analytics
For more from our panel of experts, read our articles: