Monday, before. A marketing team reports sessions, bounce rate, time on page and conversion rate.
Every one of those numbers describes what happened and none of them say what the visitor wanted. A high bounce rate on a page is a symptom with a dozen possible causes, and the team argues about which.
Meanwhile several hundred visitors a month want to know something specific that the site does not tell them, and there is no number anywhere in the reporting that represents that.
Monday before: the report describes behaviour
Sessions, bounces and conversions, which say what people did and are silent about what they were trying to find out.
Monday after: the report includes a question count
How many visitors asked something, what they asked, and how many of those questions the site could not answer.
Tuesday before: the site search data is ignored
Where it exists at all. Failed internal searches are the closest existing equivalent and almost nobody reports on them.
Tuesday after: the failed searches are combined with the asked questions
Producing a single list of what visitors wanted and did not get, ranked by how often each appeared.
Wednesday before: the page with a high exit rate is a mystery
Three people have three theories and the test to distinguish them takes six weeks.
Wednesday after: the page is understood
Because the visitors leaving it asked the same question first, and the question explains the exit completely.
Thursday before: content is prioritised by keyword volume
Which describes a market rather than the actual visitors, and is why content plans frequently miss the obvious.
Thursday after: content is prioritised by unanswered volume
The question asked two hundred times is written first, and the effect is measurable because the question stops arriving.
Friday before: improvement cannot be demonstrated
The numbers move for many reasons and nobody can attribute a change to a specific piece of work.
Friday after: the metric falls
The count of unanswered questions on a topic drops after that topic is addressed, which is about as clean a measurement as marketing produces.
What the metric is good for
Prioritising content, explaining exits, justifying work and demonstrating that a change had an effect.
What it is not good for
Replacing conversion measurement, since a site can answer everything and still not persuade anybody, which is a different problem.
What surprises teams most
How basic the top questions are, and how many of them the team believed were obvious from the existing pages.
What to do with the long tail
Most unanswered questions appear once or twice, and the instinct is to ignore them. Reading fifty of them reveals themes that the ranked list hides, and those themes are frequently more useful than the top three items.
How to present it to people who do not read reports
Five verbatim questions in an email each week. Nothing else in marketing measurement is this easy to circulate and this difficult to argue with.
Start counting what nobody could find
An assistant on the site produces this number continuously, including the exact wording of every question it could not answer, which is a metric most marketing teams have never had and cannot obtain any other way. The exit problem sits in bounce rate and the question nobody answered, the placement in landing page questions visitors ask anyway, and comparable measurement discipline in site visit and measurement questions. See pricing, run the numbers on the chatbot ROI calculator, and the sector view sits under marketing agencies.