top of page

13th of June 2024

Optimal wait time for proactive chat

When instructed to, Ami can start a conversation proactively (referred to as 'invites') on the enabled webpages. This behaviour is triggered after the customer has been idle on a website for a defined period of time (also referred to as 'wait time').

The rationale behind this approach is to avoid annoying customers with intrusive pop-ups. The goal is to use this feature to assist customers effectively or to generate interest and qualified sales leads. The challenge lies in finding the right balance: triggering Ami's proactive behavior too early can diminish customer satisfaction, while waiting too long can result in losing potential customers.

This study is going to investigate the optimal wait time with regards to increasing the number of qualified leads that go through Ami.


We have conducted a series of A/B tests with three sets of delay time settings: 10 seconds, 30 seconds, and 60 seconds. The delay time was consitent across all pages to account for external factors such as different customer behaviour on different pages. This research was carried through a month long period, encompassing a sample of 10,724 chats.


We compared three different metrics for our A/B experiments:

Clicks from invites: the rate of customers opening up Ami's widget from an invite compared to all instances of customer clicking on the widget.

Chats from invites: the rate of customers chatting to Ami from an invite compared to all instances of the customer chatting to Ami.

Qualified leads: the percentage of sales conversations classed as qualified leads.

The results are displayed in the table below.




Clicks from invites






Chats from invites






Qualified leads






Our experiment has contributed to a few insights. First, we shall note that the difference between a 30 second and a 60 second delay time is negligible in terms of engagement and conversion. However, the difference between a 10-second and 30-second delay time is quite significant. A 30-second wait time resulted in 24% fewer customers opening the widget and 17% fewer customers choosing to engage with Ami. Although a longer wait time appears to yield a higher proportion of qualified leads relative to chats, the absolute number of leads generated is still lower.


The study concludes that a 10-second delay is optimal for initiating proactive AI chats, resulting in higher engagement and more sales leads. Further research is needed to explore the effects of delays shorter than 10 seconds.

In principle, we want to provide as seemless of an experience as possible, and we have to recognise that a too aggressive setup will only reduce customer satisfaction. On the other hand, a softer setup can result in lost customers, and thus, it's important to find the right balance.



Robert Flick



bottom of page