What conversations started measures
The metric counts threads, not people and not sales. When somebody sends a first message to your business after engaging with a messaging ad, Meta records a conversation started and attributes it to the campaign that brought them. A thread that has been quiet for a long stretch and then restarts can be counted again, so the number is closer to “times someone opened a conversation” than to “customers”.
It also stops at the first message. Whether you replied, whether the exchange lasted, whether the person asked a serious question or typed a single word — none of that is in the figure. Meta cannot see what was said, so the metric is a doorway count and nothing more.
Why conversations started matters
It is the signal the bidding system uses. When you optimise a messaging campaign for conversations, delivery is pushed towards people likely to open a thread, and the reported count is how that optimisation is judged. Understanding what is being counted tells you what you are actually buying.
It is also the most immediate feedback a messaging campaign gives. Threads appear within minutes of the ads going live, so the metric shows quickly whether the creative and the offer make people want to ask something — long before there is any sales outcome to look at.
Where conversations started goes wrong
Reading it as a lead count is the standard error. A thread is somebody saying hello, which is a much smaller commitment than filling in a form, so the same money will always buy more conversations than leads. Comparing the two counts side by side makes messaging look cheap and makes forms look expensive, when the units are simply different.
Optimising for it without staffing the inbox is the second. The system will happily produce more threads than your team can answer, and unanswered threads are worse than no threads at all. The third is ignoring where they landed: if you enabled several apps, threads are spread across different inboxes, and a rising total can hide the fact that one destination is being left unread.
How to act on it
Use it as a delivery check, not as a business result. Watch it to see whether the campaign is reaching people and whether the creative is prompting questions, then judge the campaign on what happened afterwards.
That means keeping your own count. Log each conversation that turns into a genuine enquiry, then each that becomes a quotation, then each that becomes a customer, and compare those against the reported total. The gap between threads and real enquiries is the number that tells you whether the campaign works, and it also gives cost per conversation its meaning. Staff the inbox to match the volume you are buying, and if replies are slipping, slow the campaign down rather than celebrating a bigger figure. The setup behind WhatsApp and Messenger ads should be built around what your team can actually answer.