Skip to the indexed answers
Digital Marketing Index

usdigitalmarketingindex.com SEO

How can customer support transcripts reveal SEO keyword opportunities?

Support transcripts show the words customers choose when they describe a problem, before anyone on your team has renamed it. Those words are query candidates. The method is to pull the customer’s own question from each conversation, strip everything that identifies a person, group the questions by the problem behind them, and then check each group against search data before writing anything. A transcript proves that one customer asked. It does not prove that people search for it, which is why the last step exists.

Extracting customer questions

Read the customer’s first message, not the agent’s summary. The summary uses your product vocabulary. The first message uses theirs.

From each conversation, copy four things into a sheet:

  1. The question as asked, shortened to the part that would make sense typed into a search box.
  2. The nouns the customer used for your product, feature, or part.
  3. Any error text or symptom they quoted.
  4. The task they were trying to finish when they got stuck.

Skip conversations about account-specific matters such as a single refund or a delivery date. They describe one customer’s situation and have no search equivalent.

Cleaning sensitive information

Do this before the text leaves the support system. Whether transcripts may be reused for research depends on what customers were told when they contacted you and on the law that applies to your organization, so confirm with the person responsible for privacy before exporting anything.

Remove or replace the following in every row:

  • Names, email addresses, phone numbers, and postal addresses
  • Order, account, invoice, and device serial numbers
  • Health, financial, or legal details the customer volunteered
  • Anything that would let a colleague work out who wrote the message

Keep the sheet internal. The output of this work is a list of topics. Customer sentences are not copy to publish.

The rows below are made up for this guide to show what a cleaned transformation looks like. They are not drawn from real tickets.

Redacted ticket excerpt Problem behind it Query candidates
"my [device] keeps beeping three times after I change the filter" Alert after maintenance three beeps after filter change; reset filter alert
"can I use this on a well or does it have to be city water" Compatibility with a water source works with well water; city water required
"the app says paired but nothing shows up" Pairing succeeds, no data paired but not showing data; app not syncing

Grouping language by problem

Sort the sheet by the problem column and merge rows that describe the same underlying need in different words. Each group becomes one candidate topic with several phrasings attached.

Within a group, note which phrasing appears in the largest number of separate conversations. That phrasing is the leading candidate for a heading. Note also where customers use a word your site never uses. If your pages say cartridge and customers say filter, the gap is a finding in its own right.

A group supported by a single conversation stays on the list as a lead. It is not yet a topic.

Validating discovered demand

A support queue over-represents existing customers with problems. Search data covers a wider population, so compare the two.

  • In the Search Console performance report, filter queries by the distinctive words of each group. Impressions for those words show that searches containing them already reach your site.
  • An empty result is weak evidence. Search Console’s documentation says its query table “omits some queries that are searched a very small number of times” (Search Console’s notes on performance data discrepancies), so a rare question can exist without appearing there.
  • Search the leading phrasing yourself, signed out, and read what kind of page answers it. If the results are other companies’ help pages and forum threads about the same symptom, the problem is searched.
  • Check whether an existing page of yours already answers the question under different wording. Adding the customer’s phrasing to that page may serve the need better than a new page.

Record the outcome per group with the date: confirmed in Search Console, confirmed by results, or unconfirmed. Suggestions from the search box are a second source of real question wording, gathered by a different method described in the guide to using Google Autocomplete to find question keywords.

A group is ready to brief when three things are true: more than one conversation supports it, at least one validation check confirmed it, and no existing page already owns the answer.

Link trail

Reader comments on this answer

Add your comment to this answer

Your email address will not be published. Required fields are marked *