You can review autoreplies metrics to understand how your article recommendations for email and web form are performing. Based on the metrics, you can make changes to your help center content to improve your suggestion rate, click-through rate, and resolution rate, and to decrease your rejection rate.
This article contains the following sections:
Improving your suggestion rate for article recommendations
The suggest rate is the percentage of questions that autoreply capability was active for and has sent suggestions for. If the suggest rate is low, or lower than expected, it can indicate that article labels are misconfigured or that you receive a larger-than-usual number of questions from unsupported languages.
-
Check your autoreply trigger notification body (email autoreplies only). Your autoreply triggers require the placeholders
{{answer_bot.article_list}}
or{{answer_bot.article_body}}
. If these placeholders are not present in the body of the email notification the trigger will still fire, but no articles will be suggested. This can cause the suggest rate to reflect inaccurately. Learn more about configuring autoreply triggers. - Check your article labels. If you're using labels, this might be the cause of a lower suggest rate. Labels restrict which articles are searched. If relevant articles aren't in the restricted set, then no articles are suggested. This can cause the suggest rate to decrease. Learn more about using labels with autoreplies.
Improving your click-through rate for article recommendations
Article recommendations sent in email already have a disadvantage as it requires users to first open the email and read the body, before they are likely to click through to any articles. Click-through is the percentage of answers clicked by end users from the total articles suggested.
- Make the subject line more appealing and accurate. The email isn't just an acknowledgement email, it's got valuable content within it that may answer the user's question. Be clear about this in the subject line.
-
Don't hide the content. It's important that you craft the body of your message with
purpose. Use conditional logic to craft a message to your users and
make sure the
{{answer_bot.article_list}}
or{{answer_bot.article_body}}
placeholders are positioned at the best place in the email body.
Improving your resolution rate for article recommendations
- Retain as much question context as possible. If you can retain as much similar phrasing of the original question within the article, the contextual match will be better. For example, the question "I can't log in, so how can I reset my password?" should result in an article with the title "How do I reset my password so I can log in?"
- Keep the article focused on a single problem. Each article should focus on one problem and one solution only. If you have long FAQ articles, consider breaking the FAQs into separate articles and grouping them together in an FAQ section. Also, remember that there is greater emphasis on the first paragraph or two of the article (approximately the first 75 words in English), so be sure to include as much contextually-relevant information into the top of the article as possible.
Decreasing your rejection rate for article recommendations
- Write articles with a clear title, concise introduction, and narrow focus. Article titles should be phrased as a question or a simple, active phrase. The intro should include keywords and context in the first 75 words, and the article should be focused on a single, specific topic.
- Use article labels to filter results. Labels can help reduce "noise" in your help center by focusing retrieval results on the articles you want to be considered. Labels can also help target customer segments by showing each segment only the relevant articles.
For more information, see Optimizing your help center content.