Loading
Because it was written once, at launch, describing the product as it was designed rather than as it confuses people. Nobody has read the tickets since.
Customers search, find nothing that matches their words, and raise a ticket anyway. The help centre is judged a failure when nobody ever told it what to say.
Articles are written from tickets, and each keeps the link back to the ones it answers. Failed searches — what customers typed and did not find — are reported as a queue, which is the backlog for what to write next.
When an agent answers a ticket well, that answer can become an article without being rewritten from scratch.
The writing queue is evidence-led. Instead of guessing at topics, somebody works down a list of the phrases customers actually searched for and got nothing.
Article suggestions surface in the agent's reply and in the customer's search, and the suggestion is measured by whether it resolved anything.
It will not deflect a ticket the customer needs to raise. An article that hides the contact route to improve a deflection metric produces a worse customer and a worse number later.
It also does not write itself from the product. Articles come from questions people actually asked.
Nothing here is an integration. Each line is two products reading the same record from different sides.
Articles are written from tickets, and each keeps the link back to the ones it answers. Failed searches — what customers typed and did not find — are reported as a queue, which is the backlog for what to write next.
The writing queue is evidence-led. Instead of guessing at topics, somebody works down a list of the phrases customers actually searched for and got nothing.
It will not deflect a ticket the customer needs to raise. An article that hides the contact route to improve a deflection metric produces a worse customer and a worse number later.
Fourteen days, every module, no card. Or half an hour with someone who will run it on your own records and tell you where it does not help.