Meta Ads counted 17,402 welcome message views as onsite conversions across 7 accounts

Meta's insights API returns an action type called onsite_conversion.messaging_welcome_message_view. It fires when a person taps a click-to-message ad and the automated greeting the advertiser wrote appears at the top of the thread. The person has not typed anything. Across the 23 Australian Meta Ads accounts in this audit, that single action type carried 17,402 counts in 7 accounts, and every one of them sits under the same onsite_conversion. prefix Meta uses for an instant form lead.

Pooled across the seven accounts, welcome message views outnumbered first replies five to one. In three of them the ratio was 15 to 26 views for every reply. One home improvement account logged 4,290 welcome message views, 163 first replies, and 32 leads on the event it actually calls a lead.

All money in this series is AUD. This post has no spend figures in it, because the finding is about what got counted rather than what it cost. We pulled the lifetime actions array for each account, broke it out by action_type, and kept every type beginning with onsite_conversion.messaging_. Part 7 covers the whole family. This part is about its largest member.

The seven accounts

The right-hand column is the one to read. It splits the seven accounts into three groups. Accounts G and H sit at about 2.7 greeting views per reply: for every person who typed something, roughly two more opened the thread, read the greeting, and left. Accounts P, V and W sit at 15 to 26. Accounts M and A sit below one, which means replies arrived without a greeting being recorded at all.

That last group is worth a sentence, because it shows the metric is not a funnel stage every conversation passes through. The likeliest reason replies exceed greeting views in M and A is that most of their messaging ads do not have an automated greeting attached, or the threads open in Instagram or WhatsApp where the greeting is not shown or not counted. The API does not say which. What it does show is that a welcome message view is neither necessary for a reply nor predictive of one. In account A, 8 views sat next to 168 replies. In account V, 4,290 views sat next to 163.

Why this happens

When you build a click-to-message ad, Meta asks for a message template. It has a greeting line, optional quick-reply buttons, and optional frequently asked questions the person can tap. When someone taps the ad, Messenger opens with that greeting already rendered as a message from the business, as though the business had spoken first. Meta logs that render as a welcome message view.

Here is where the event sits in the sequence Meta reports, and who has to do something for each step to count.

The welcome message view is the only step in the list that requires nothing from anyone. It is the second event after a tap and it is guaranteed by the tap, provided a template is attached. In practical terms it is a click with a different name.

Meta files it under onsite_conversion. because the event happens on Meta's own surface, inside Messenger, rather than on the advertiser's website. That is all "onsite" means in Meta's action taxonomy. The prefix describes where the event happened, not whether it matters. Part 2 went through the onsite, offsite and omni prefixes; the point here is that the prefix that carries instant form leads also carries a greeting render, and nothing in the field name warns you.

In Ads Manager, welcome message views can be added as a column from the messaging engagement set. Most people never add it, so most people never see it. It matters when someone builds a report from the API instead, because the actions array returns every action type at once, and a report that sums the array, or treats every onsite_conversion.* type as a conversion, absorbs the whole 17,402. In the Google series the same failure took the form of three conversion actions on one lead. Here it is one automated event filed beside the leads.

What it does to decisions

The welcome message view is not something an ad set bids toward. In every messaging ad set configuration we know of, the optimisation goal is conversations started or something later in the sequence, so the auction is not chasing greeting renders. The damage is upstream of the auction, in the numbers people use to decide where budget goes.

Three accounts, P, V and W, hold 79.5% of the views, and in those three fewer than seven people in a hundred who saw the greeting went on to type anything. If a reporting layer in any of those accounts counted the greeting as an engagement, a conversion, or a "messaging result", the campaigns that showed the most greetings looked like the campaigns that produced the most outcomes. They were the campaigns with the most taps.

Two specific decisions get bent by this.

The first is budget comparison between messaging campaigns and everything else. Campaigns with a Messenger destination report conversations as their result, and conversation counts are already a generous measure (Part 7 goes into how many names Meta gives the same moment). Add greeting views on top and a messaging campaign can appear to produce thousands of results against a form campaign's hundreds, at a fraction of the apparent cost per result. In account V the raw comparison is 4,290 greeting views against 32 leads.

The second is the creative and template decision. If the greeting is treated as a result, a template with a long greeting that people open and abandon scores the same as one that asks a question people answer. The reply rate is the number that separates them, and in this set it ranged from 3.8 to 37.6 replies per hundred views in the accounts with meaningful volume. That spread is a creative finding, and it is invisible if the greeting render is what gets counted.

We see this most often when reporting was built once, by someone who has since moved on, and the field list was chosen by ticking everything with "conversion" in the name. Our Facebook Ads management work starts by rebuilding that field list from the actions array outward, and the messaging family is the first thing we separate out.

Where it varies

By band, this is a trades and health finding. Four of the six home improvement and trades accounts fired the event, and between them they hold 13,377 of the 17,402 views (76.9%). Two of the four health and care accounts fired it, for 4,017 (23.1%). One consumer goods account fired it eight times. No manufacturing account and no professional services account recorded a single welcome message view, which fits: click-to-message is a trades and appointments format, not a B2B one.

Within the accounts that fired it, the useful comparison is the greeting against the account's own lead event, whatever that event is (instant form, pixel lead, or a messaging lead). It changes the question from "how many greetings per reply" to "how many greetings per thing the business would call a lead".

Account V's 32 leads is a small base, just above the 30-event floor this series uses, so treat 134 as a direction rather than a precise multiple. It is still the most extreme relationship in the set: a greeting rendered 134 times for every lead the account recorded. Account W, at 16.2 greetings per lead on 193 leads, is comfortably above the floor and nearly as stark.

At the other end, G and M ran lead generation at scale alongside their messaging ads, so the greeting is a fraction of their lead count. In those accounts the welcome message view is noise. In V and W it is the largest number in the account, and if anything was ever reported as a total, it was this.

What to check in your account

1. Open Ads Manager, Columns, Customise columns, and search for "welcome". Add Welcome message views. In the same panel add New messaging connections and Messaging conversations started. Run it lifetime and by campaign.

2. Divide welcome message views by new messaging connections for each messaging campaign. Below 1 means your ads mostly have no greeting attached. Around 2 to 3 is what accounts G and H looked like. Above 10 means the greeting is rendering for many more people than are talking to you.

3. If a third-party dashboard or a Looker Studio connector feeds your reporting, find the measure it calls conversions or results. Check whether it is built by summing the actions array or by filtering it. Search the connector's field list for messaging_welcome_message_view. If it is in the sum, it is in your total.

4. From the API, request fields=actions on the account with date_preset=maximum and list the distinct action_type values. Anything starting onsite_conversion.messaging_ should be reported separately from leads and purchases, if it is reported at all.

5. For each click-to-message ad, open the message template. Read the greeting as a prospect would. If it does not ask a question a real buyer would answer in one line, the reply rate is your creative problem, and the greeting view count is hiding it.

6. Compare cost per new messaging connection, never cost per welcome message view, against cost per lead from your form and pixel campaigns. The two are still not the same thing, but at least both require a person to do something.

The principle

Meta names things by where they happened, not by what they mean. A greeting render happens on Meta's surface, so it is an onsite conversion, and it lands in the same array as a lead with the same prefix and the same shape. The only defence is to know what each action type requires of a human, and to count only the ones that require something. A tap that opens a thread is a click. The greeting that appears is your own ad talking. Neither is a result, and in some accounts they are the largest numbers in the file.