Pitfalls
Where store chat usually goes wrong.
The most expensive mistake is staffing chat for the hours the office is open rather than the hours the store is busy. Retail traffic peaks in the evening and at the weekend, and a widget that says it is live while nobody is watching converts worse than one that honestly offers an offline form with a reply time you will hit.
The second is answering the same question by hand forever. A question that arrives forty times a week is usually a page problem wearing a chat costume: the delivery estimate is buried, or the returns window is written in a way nobody can parse. Fixing the page removes the chat, permanently. The third is meeting the seasonal peak with no plan, which is what seasonal planning exists to prevent.
For measurement, the number that matters is not chat volume. Compare the order rate of sessions that included a chat against sessions that did not, over the same window and the same traffic source, and track how that gap moves as you change where the invitation fires. Measuring chat conversion sets out the honest version of that calculation, including why the raw comparison flatters chat.