Estimated Wait Time (EWT) Algorithm

The Estimated Wait Time (EWT) algorithm predicts the time a visitor in the live chat queue will wait before being attended to. This document outlines the mechanics of the algorithm, its current limitations, proposed improvements, and troubleshooting steps.

How the EWT algorithm works?

The EWT algorithm calculates wait times based on the average duration of recently closed chats and the visitor's position in the queue. The formula is as follows:

  • : Maximum number of simultaneous chats an agent can handle (configurable by the admin).

  • : spot

  • Average Chat Time: The mean duration of the  (default 100) most recently closed chats.

The EWT updates dynamically as chats close, and the sample evolves. If no department is assigned, EWT defaults to a global queue calculation.

Example Calculation

Assume:

  • Maximum simultaneous chats per agent : 3

  • Average chat time: 15 minutes

EWT for contacts in the queue would be:

  • Person in 1st Spot:

  • Person in 2nd Spot:

  • Person in 3rd Spot:

  • Person in 4th Spot:

This demonstrates that EWT increases incrementally for each subsequent position in the queue while factoring in the system's ability to handle multiple chats simultaneously.

The default sample size for EWT calculation is 100 recently closed chats. However, this value is configurable, and you can adjust it in the system settings to better suit your operational needs. Consider increasing or decreasing the sample size based on your chat volume and activity patterns for more accurate estimates.

Customizable Settings

Rocket.Chat provides flexibility to adapt the algorithm to your operational needs:

  1. Sample Size (X):
    While the default sample size is 100 chats, administrators can configure this value in settings. Increasing the sample size improves accuracy but may reduce responsiveness to real-time changes.

  2. Maximum Simultaneous Chats (N):
    Administrators can define the number of simultaneous chats agents can handle, which directly impacts the calculation.

  3. Time-Based Sample:
    Instead of relying on a static sample size, you can configure the algorithm to use a time-based sample. For example:

    • Use chats from the last hour.

    • Use chats from the last 30 minutes.

    • Restrict to chats from the current day.

This ensures the algorithm reflects recent activity patterns rather than relying solely on a fixed number of past chats.

Limitations

Despite its flexibility, the EWT algorithm has some limitations:

  1. Impact of Outliers:
    Chats initiated outside operational hours (e.g., on weekends) can skew the average and inflate EWT values.

  2. Sampling Challenges:
    Fixed sample sizes may not always reflect real-time conditions, especially during sudden spikes in activity.

  3. Exclusion of Agent Activity:
    The algorithm does not factor in the number of active agents, which may lead to inaccurate predictions in cases where agents are idle or unavailable.

Troubleshooting Common EWT Issues

  1. EWT Remains High Despite Active Agents:
    Long-duration chats or small sample sizes may cause this issue. To resolve this, you can increase the sample size or use a time-based sample to make more accurate predictions.

  2. EWT Misleading During Non-Operational Hours:
    Chats initiated during weekends or outside working hours may inflate EWT. Configure the system to exclude such chats from the calculation.

  3. Department-Specific Discrepancies:
    If EWT varies significantly between departments, ensure chats are correctly assigned, and sampling settings are consistent.

    Version 7.1 brings significant enhancements to the EWT algorithm, addressing key issues to improve accuracy and reliability. Previously, chats that remained indefinitely in the queue distorted EWT calculations; this issue has now been resolved. Queue management has also been improved, with long-duration or inactive chats handled more effectively to ensure accurate estimates. Additionally, operational adjustments provide better queue management during non-operational hours, offering greater flexibility and precision in handling wait time predictions.

The EWT algorithm is a powerful tool for predicting wait times in live chat queues. Administrators can optimize their system for accurate, real-time estimates by understanding its mechanics and leveraging the available settings.