## TL;DR
Your FAQ already exists, scattered across resolved tickets: cluster recent tickets by topic, take the top 20 by volume, and write each answer from an actual resolved thread, verified by the agent who solved it. Publish where customers already look and link the articles from your macros. Refresh quarterly from new ticket data, because the top questions drift with every release.

## The query

```text
how to build a support FAQ from ticket history
```

## Use this when

- The help center doesnt answer what customers actually ask
- You are building an FAQ from scratch
- Deflection is flat and tickets repeat
- You need evidence for which articles to write first

## Not for

- Help-center navigation or information architecture
- Article writing style guides
- Video or interactive tutorials
- Marketing FAQ pages

## Steps

### 1. Export tickets and cluster by topic

Pull three to six months of tickets and group them by what the customer was really asking, not by your internal categories. Tag-based clustering gets you most of the way; read a sample from each cluster to confirm the grouping is real.

Expected output: ticket clusters with a plain-language question per cluster.

### 2. Rank by volume and pick the top 20

Sort clusters by ticket count and take the top 20. These are your FAQ entries, in priority order. Resist adding pet topics with no volume behind them; the FAQ earns trust by answering the questions people actually have.

Expected output: a ranked list of the 20 highest-volume customer questions.

### 3. Write each answer from a resolved thread

For each question, find a ticket where an agent answered it well and adapt that reply into the article. Have the agent (or topic owner) verify it. Real resolved answers beat written-from-scratch docs because they already survived contact with a confused customer.

Expected output: 20 articles drafted from real resolutions, each verified.

### 4. Publish where customers already look

Put the FAQ at the top of the help center, surface relevant articles in the contact form before submission, and link each article from the macro for that ticket type. An FAQ nobody encounters deflects nothing, placement is half the work.

Expected output: the FAQ live in the help center, the contact flow, and the macro library.

### 5. Refresh quarterly from new data

Re-run the clustering every quarter: new features create new questions, fixes retire old ones. Archive articles whose ticket volume died and draft new ones for emerging clusters. Put the refresh on the calendar or it wont happen.

Expected output: a quarterly refresh cycle with archived and added articles.

## Variant phrasings

### turn support tickets into faq articles

Steps 1 through 3. Cluster, rank, adapt resolved answers.

### how to decide what goes in a support faq

Step 2. Volume-ranked questions, nothing else.

### faq from helpdesk data

Steps 1 and 5. Cluster the data, refresh on a cycle.

## Why it happens

Help centers usually get written from the product's structure (features, settings, plans) while customers ask from their situation ("my invoice is wrong"). Ticket history is the only honest record of the customer's vocabulary and the questions' real frequency. Building the FAQ from that record aligns the answers with the asking, which is the entire mechanism of deflection.

## Edge cases

- Questions with account-specific answers: some top questions cant be answered generically ("why was I charged X"). Mark them as contact-us with the info to include, dont fake a generic answer.
- Answers that change with releases: date-stamp articles and tie reviews to the release cycle, or the FAQ quietly goes stale.
- One question, many variants: customers phrase the same question ten ways. Put the variants in the article's first paragraph so search finds it.
- Low-volume but high-pain questions: a rare question that causes churn or escalation deserves an article even outside the top 20. Note the exceptions explicitly.

## Provenance

Resolved from the public thread: https://vectle.com/posts/pst_CoKhz17ARYP7cTH9jmqMIA