Sign in

Blog · Territory, capacity and quota planning · Customer service

Deflectable tickets by category: the how-to questions an article would have answered

How a customer service leader finds the tickets that a help article, a product change or an onboarding step would have prevented, from the ticket export: tickets by category and sub-category, the share that are how-to against fault, the how-to questions that recur across accounts, whether a help article exists for each and whether it was viewed before the ticket, the handling time consumed, and the deflection list ranked by hours a month.

The short answerHow-to tickets are questions, not faults, and a question asked by forty accounts in a month is a help article, a product label or an onboarding step that does not exist or is not found. From the ticket export, tickets by category and sub-category split into how-to and fault; the recurring how-to questions across accounts are grouped; for each, whether an article exists and whether the account viewed it before raising the ticket; and the handling time consumed. The deflection list ranks the questions by hours a month, and the top ten are usually a quarter of the how-to load.

A support team answers the same question about exports two hundred times a month, and each answer is a ticket, a reply and twenty minutes. The ticket export knows the question recurs, whether an article exists, and what it costs. This guide sets out the how-to split, the recurring clusters, the article check, and the deflection list.

The measures

Per ticket:

Kind: how-to or fault, from category or classification Cluster, for how-to tickets: the recurring question it belongs to Article exists for the cluster; viewed by the account before the ticket

Per cluster, per month:

Tickets, accounts, handling hours Article state: none; exists, viewed; exists, not viewed

The rows you need

  • Ticket export: ticket, account, category, sub-category, subject, first message, handling time.
  • Help centre: article, cluster mapping, views by account and date where logged.

Account identifiers only.

The assertion

tickets = how-to + fault + other, per period

Every ticket in one kind. Unclassified tickets are counted and clustered where the wording allows.

A worked deflection list

How-to tickets: 1,840 a month, 610 handling hours.

Cluster Tickets Accounts Hours/month Article Viewed before ticket Fix
How to schedule an export 210 140 70 None Write it
Where to change the billing contact 160 150 27 Exists 8% Placement: link from billing page
How to add a team member 140 120 23 Exists 61% Article unclear: rewrite
Why the report totals differ 90 40 60 None Product: show the definition in the report
Password reset for SSO accounts 85 80 14 Exists 12% Placement

Top five clusters: a third of how-to hours. Two need an article, two need the article found, one needs the product to say something.

Three fixes

Article state Fix Owner
None Write it Support content
Exists, not viewed Naming, search, placement Support content, product
Exists, viewed, ticket anyway Rewrite, or the product Support content, product

Tracking

Per cluster, tickets per month after the fix. A cluster that does not fall was not fixed by what was done, and the row says so.

Where it goes wrong

How-to and fault blended. The deflection list is full of bugs.

Clusters unnamed. The same question counted as two hundred different tickets.

Article existence unchecked. Articles written that already exist.

Ranked by count. The two-minute question over the hour one.

Every month, the clusters and the list

Mapped once, the ticket export and the help centre produce the how-to split, the clusters, the article state and the deflection list ranked by hours every month. Covirage builds this from the exports as they are. The customer service page describes the setup, and the effort per account guide covers the accounts the how-to hours concentrate in.

Questions people ask

How are recurring questions grouped?

By sub-category where the desk has one, and by the ticket's subject and first message clustered on wording where it does not. The grouping proposes; a support lead confirms the clusters and names them. Once named, the cluster is a row that can be tracked month to month.

What does 'article existed and was not viewed' mean?

That the article is there and the customer did not find it, which is a search, naming or placement problem rather than a writing one. Where the help centre logs views by account, the join says which. It is a different fix from no article at all.

Why hours rather than ticket count?

Because a question asked eighty times that takes two minutes is less than one asked twenty times that takes an hour. Handling time per cluster is the cost, and the deflection list is ranked on it.