How to build a workforce forecasting process for customer support

Great customer support requires that you solve customer problems promptly and with empathy. But as with many things, preparation is key. Accurate workforce forecasting puts the team in a position to succeed, helping ensure that customers can connect with the right person at the right time. This article will walk you through the basics of forecasting your volume.

Why bother with workforce forecasting?

In many cases, those who have been working in support for a long time have developed a "spider sense" for what will happen on any given day, whether through hard-earned experience or spooky omniscience. For the rest of the team, though, good forecasts can help enable:

• Predictable response times: An understanding of when customers tend to call or write in allows the team to staff accordingly.
• Well-informed hiring plans: Support leaders increasingly seek data to inform their decision making, and accurate forecasts can bring clarity and show that you've done your homework.
• Team visibility: If things feel unexpectedly busy on a given day, why is that and what can be done? Forecasts help a team supplement their intuition.
• Efficiency and fairness: It’s always a tricky balance between understaffing and overstaffing. Accurate forecasts support a distribution of workload that is both efficient and fair.
  

Those last two points are especially important. Accurate workforce forecasting is good for employee morale. According to a recent SWPP survey, forecast accuracy is actually the #1 measure that affects team satisfaction.

How to build a forecast that benefits your team

Good forecasts start with an understanding of what you’re trying to achieve and the fundamental dynamics of your support volume.

Choose the right volume forecasting interval

Start by deciding on the interval and timespan that are most relevant to your goals. It’s best to match up your interval with what you’re trying to achieve:

• Headcount planning (long term): If you're planning headcount or strategizing about what your team will look like months or years from now, consider forecasting over weeks or months to get a clearer picture of long-term needs.‍
• Email or non-urgent support (medium term): For non-urgent support channels, such as email or help desk tickets, it’s best to segment volume by day for more accurate forecasting.‍
• Chat, phone, messaging, and AI-automated support (short term): When it comes to real-time support channels — like live chat, phone, messaging apps, or AI-automated support — forecasting should be done at a more granular level, such as hourly or even 15-minute intervals. This ensures you can respond swiftly to spikes in demand. However, keep in mind that you'll need substantial volume and data for these short-term forecasts to be truly effective.
  

Retrieve the right data

Most contact platforms allow you to export historical data that you can use for volume forecasting. To enumerate a few of the most popular platforms:

• For Zendesk, you can use Zendesk Explore to export a count of tickets.
• For Intercom, you can inspect the New conversations report, though there’s no way to export the contents in a structured form.
• For Kustomer, you can use the CSV Export Feature, though data goes back only as far as one month.
• For Salesforce, you can use a Custom Report Type on Cases.
• For Gladly, you can export Conversations and Contacts via CSV or API.
  

In cases where data is incomplete or difficult to export, you can always use Assembled’s out-of-the-box integrations.

Take the simplest approach of averaging the past

At its core, most forecasts use knowledge about the past to inform the future. The simplest approach is to use a headcount forecasting formula that calculates the average number of contacts over a set interval. 

For example, in order to forecast how many tickets you will receive at 9am on Monday, take the average 9am contact volume for the past several weeks. If your volume varies based on the day of the week — often weekends are lighter than weekdays —  average together only the 9am contact volumes for the past several Mondays or weekdays.

For many support teams, it’s easiest to begin running these calculations using a spreadsheet. This is the method that many workforce managers rely on. A report from Call Centre Helper reveals that more than 60% of support operations are currently leveraging spreadsheets for forecasting.

We put together an Excel template to walk you through this process from beginning to end. Click here to get your copy of the template.

Move beyond the basics and uplevel your forecast model

Now that you understand the basics of forecasting, it’s time to dive into selecting the right model for your team. Different models suit different needs, and picking the one that aligns with your data and goals can make all the difference.

N-Week Average

The N-Week Average model is simple and effective, making it a great starting point for many teams. It works by calculating the average contact volume over a set number of weeks (N), such as 8 weeks. This model is ideal for short-term forecasts (up to 4 weeks) where your contact volume has been stable and predictable.

When to use it: If you need an easy-to-explain model that works well with limited data, or if your team isn’t seeing major fluctuations in volume, this is your go-to. However, if your contact volume is trending up or down, you may want to explore other options, as this model won’t capture those shifts.

N-Week Average with Momentum

Building on the N-Week Average, this model adds a layer of sophistication by accounting for upward or downward trends. It’s perfect if your team is experiencing steady growth or declines, as it uses recent data trends to project future volume.

When to use it: If your support volume is rising (or falling) over time, the N-Week Average with Momentum model helps you forecast more accurately by adjusting for those changes. It's particularly useful when you want to account for the latest growth patterns without moving to a more complex model.

Seasonal

For teams that experience cyclical patterns, the Seasonal model is a game-changer. It looks at over a year’s worth of data to find seasonal trends, making it highly reliable for medium to long-term forecasts. Whether your team is busy during holiday seasons or sees regular patterns over the course of a year, this model will pick up on those nuances.

When to use it: If your support volume follows predictable seasonal trends—like increased traffic around holidays or specific times of the year — this model will help you plan months in advance with confidence.

Prophet

For teams with complex data, Prophet is the powerhouse forecasting model. Developed by Meta, Prophet leverages machine learning to capture both seasonal patterns and external factors, like holidays or events, that could affect your contact volume. It’s highly customizable, making it a great fit for organizations with unique or complicated support needs.

When to use it: Choose Prophet when your data shows intricate seasonal patterns or if your team faces significant external events that impact contact volume. While it offers accuracy and flexibility, Prophet’s complexity can make it harder to explain, so it’s best for teams prioritizing forecast precision over simplicity.

*This blog post was originally published on May 21, 2020. It was updated on October 9, 2024.