How to reduce support tickets: a four-week method, not a list of tips
By The Relay team · Published · 4 min read
The short answer
Tag every ticket for two weeks, rank the tags by volume, write one clear help article for each of the top five, and publish them where customers ask (the widget and the help center). After four weeks, compare tag volumes. Repeat with the next five. The loop works because it starts from your actual tickets rather than generic best practice.
The loop
- Weeks 1 and 2: tag everything. Give each conversation one primary tag as you resolve it.
- End of week 2: rank the tags by number of conversations.
- Week 3: write one article for each of the top five tags. One question per article.
- Week 3: make them findable. Publish them in the help center so the widget search and the bot can use them.
- Week 4 and after: measure. Compare each tag's weekly count with the first two weeks. Keep what works; rewrite what doesn't.
- Repeat with the next five tags.
Why most advice on this doesn't work
Lists of tips ("build a knowledge base", "add a chatbot", "be proactive") skip the one step that makes any of them pay off: knowing which questions are repeating in your product. Without it, you write articles nobody searches for and the same three questions keep arriving.
Be wary of headline figures such as "a knowledge base cuts tickets by 40%". Vendor blogs repeat them, and we have not found a primary source that applies to a small SaaS. Measure your own numbers instead; the loop below produces them.
Step 1: choose tags you can keep using
Ten or fewer tags, named after what the customer wanted, not what team handles it. A starting set to adapt (these are examples, not data):
| Tag | Covers |
|---|---|
| login | Passwords, sign-in links, locked accounts |
| billing | Invoices, cards, plan changes, refunds |
| how-to | How do I do X in the product |
| bug | Something is broken |
| feature-request | Can it do X? Will it ever? |
| integration | Connecting to other tools |
| account | Changing email, deleting data, adding teammates |
In Relay, you tag conversations in the inbox. For this loop, pick one primary tag per conversation and stick to it.
Step 2: rank, then write
Count conversations per tag. The top five usually account for most of the volume, and within each, a handful of specific questions repeat. Write the article for the specific question, not the tag: not "billing" but "How do I download an invoice?". The writing rules for bot-friendly articles apply to people too: customer words in the title, the answer in the first paragraph.
Step 3: put the articles where the question happens
- The widget. Customers can search the help center from the widget's home tab, and the bot answers from the same articles.
- The public help center. Articles are findable by search engines as well as by customers.
- Your saved replies. Link the article in the reply you use for that question, so humans and the bot point to the same source.
Step 4: measure what changed
- Weekly count per tag, compared with the baseline. This is the number that matters.
- Bot-resolved conversations: how many ended with "That helped". Relay's analytics, on the Team plan, show the bot resolution rate and the busiest tags.
- Handoffs the bot makes on questions you now have an article for. If they continue, the article is badly titled or buried.
Common mistakes
- Writing one long article for a whole tag.
- Naming articles after your features instead of your customers' words.
- Never retiring stale articles, so a bot or a customer finds the old answer.
- Counting only the tickets you reduced, not the ones that were never opened (compare against total conversations).
- Skipping the human path. Customers who can't get a person after an article didn't help will write in angrier.
Questions
Common questions
- How can I reduce support tickets in a SaaS product?
- Find out which questions repeat by tagging tickets for two weeks, write one clear article for each of the top ones, make them findable in the widget and help center, and compare tag volumes a few weeks later. Then repeat with the next group.
- What is ticket deflection?
- Ticket deflection is a customer getting an answer from self-service (an article or a bot) without opening a ticket with a person. How it's measured varies by tool, so define it for yourself: for example, conversations the bot resolved with "That helped", plus article views that didn't lead to a message.
- Does a help center reduce support tickets?
- It can, for questions that repeat and have a clear answer. Whether it does for you depends on whether the articles match the questions customers ask, which is why the loop starts from your own tags.
- Will an AI chatbot reduce my tickets?
- It can answer repeat questions from your articles, and hand the rest to you. Its answers are only as good as the articles behind them, so start with the articles.
Sources and how we wrote this
Relay is made by the team that wrote this post. Figures for other tools come from their own public pages, listed under Sources, and are list prices in US dollars that can change. If something here is out of date, tell us on the feedback board and we'll correct it.