Open the metric definition before you open the report. The dev.to post asserts that most platforms create a conversation record the moment the widget loads or the bot fires its automatic greeting, with no visitor input required [14]. It names no platform and gives no sample size; the evidence offered is the author's own builds and the dashboards he opened afterwards [6]. So treat it as a hypothesis about your vendor rather than a fact about the category, and test it where the tooltip or the docs say what event is being counted. If the counted event is session creation, the headline figure is a derivative of page views on whichever pages carry the script tag.
Then run the two subtractions the author prescribes: keep only sessions where a human sent at least one message, and remove your staff, the vendor's team, and your own poking around [15]. Real equals headline minus greeting-only minus internal [1]. Both terms are usually filterable in the same interface that produced the inflated number.
The window does more work than it looks. Three months is about 90 days against a week's 7, so a single atypical busy day carries roughly a thirteenth of the weight it would carry in a weekly view [2]. That is the arithmetic behind the post's instruction to pull the last three months and not the last week [13].
Whether the distributor case transfers depends on three conditions. Your enquiries have to arrive somewhere other than the website. Your platform's conversation definition has to match the one described. And your site must not be the transaction surface itself; if the job is checkout support or docs deflection, the widget is on the correct channel and the same filtered count simply tells you how well it is doing. In the post's example none of that applies: the bot answers products, working hours and coverage area on a site last updated when the address changed [9], and the traffic is job applicants, an occasional brand manager doing due diligence, and the owner's nephew checking whether the site is still up [10].
The part I would price first is the gap on the channel that does have arrivals. Orders land in a WhatsApp inbox that goes unanswered between two and five in the afternoon because the salesman is on the road, and that inbox has no automation on it [11]. Three hours per working day of queued demand [3], against a widget whose queue may have no arrivals at all.
The useful accident here is that the instrument was built by the party selling the subscription. Every platform keeps the conversation log precisely because that log is how the renewal gets justified [12], and the author's claim is that reading it properly takes under an hour [7].
