Uncover crucial insights by analysing your support queries

Folashade says: “Analysing support queries is crucial for identifying customers’ pain points, uncovering their search intent, and discovering patterns and emerging trends in 2025.

You can also use topic modelling to enhance your SEO strategies.”

What are the best support queries to analyse?

“There are a lot of different ones, obviously. You can get them from contact forms, calls, emails, etc. It all depends on the platform that you use to gather these support queries.

Take them all together into one document or data frame and start to look into them.

You aren’t necessarily looking for a specific query, you’re analysing as much as you can to see what’s there and trying to find repeated words and phrases to determine common pain points.

Start by getting everything that you have, look at the frequency of the queries, and start to make sense out of it.”

Do you use AI or automation to help make sense of it or is that a human job?

“I am looking at this from the perspective of a large business, so I prefer to use machine learning to do this kind of analysis.

Identifying customers’ pain points is one of the first steps. You look at identifying their pain points, uncover their search intent, and then look for patterns and trends in the queries. You don’t necessarily have to have something that you’re looking for in mind. You can, but it’s always good to look at everything.

If you’re a large business, I would advise that you do this at least once every 6 months or once a quarter. It depends on what you prefer.”

Why is it so important to identify pain points first?

“By looking at pain points, you’re trying to address the issue with your content. Yes, it’s good to know the positive side or what customers find interesting about your product or service, but looking out for the pain points will help you to find gaps. It will help you find ways to help the customers and provide valuable content to them.

Through this analysis, you will be able to see where they’re not finding the content that they’re looking for, so they are having to reach out instead.

You want to address that complaint, so they don’t have to contact customer service or fill in a contact form. The reality is that not every customer or user is going to do that. They will probably just leave your website and go somewhere else if they don’t see that information.

You’re trying to get them to convert by giving them the information that they need. If they have to contact you, then take that into consideration when you’re looking at the next piece of content that you are going to create.”

How do you identify intent and what purpose does it serve?

“Once you have built that large dataset of support queries, it could have a thousand rows or more – so I prefer to use machine learning for this. To identify the pain points, you start looking at the sentiment, and you carry out a sentiment analysis. That is going to break down those queries into positive, negative, and neutral.

Because you are looking at the pain points, you are going to look at the negatives. Obviously, with machine learning, there are a lot of things in the back end like tokenization, lemmatizations, terms, etc., that go into putting those queries into different categories. You take that negative feedback, look at it, and start to identify the search intent.

You can do that through something called ‘parts of speech’ analysis. It breaks down the sentence into single words and starts to look at the intent behind each of those words. This is where lemmatization also comes in. Lemmatization looks at the word and takes it back to the root – or lemma.

It all depends on context. If someone says, ‘I want to book a flight’ the lemmatization process would look at the context of what that person is trying to say. Obviously, that person could have been wanting to book something else, but they specified that they wanted to book a flight.

Machine learning does all of these things for you, and it starts to put them into different parts of speech like nouns, verbs, adverbs, etc. Then, you start to see meaning in these things, and you start to see the keywords. Interestingly, you sometimes find out that the words your customers use around your service could be different from what you expected.

When you break down these words and do this kind of analysis, you understand what the customers are complaining about – the main lemmas, and the main keywords that they use – and you start to make sense of it.

There are also several other things you can do. You can combine different parts of speech to start to understand the intent. You can look at the verbs and the nouns and combine them to understand what intention your customers have.

If they use the phrase ‘connect flight’, are they complaining about their connecting flight? Are they looking at something else? The phrase could be ‘wants refund’ or ‘prefer blue’. When you look at things from the perspective of the verb and the noun, and bring them together, you start to understand the intent of the query.

Another way you could look at it is the adverbs and the adjectives. If a customer uses the phrase ‘easily understand’, then you know that you need to look for what it is that they are trying to understand. These words show their intent and what they are really driving at. You can take that phrase, go back to the data frame or the documents that you have, and filter for that.

If you filter for the phrase ‘prefer blue’, you can start to figure out what they are really saying around that word. If you still have a large data frame after you have filtered for ‘prefer blue’, for instance, you can do text summarisation to break that down and get the important points from it.

That’s what I mean by understanding their search intent: digging into the data and trying to understand what their intention was for reaching out to you.”

Can you also get value out of the positive feedback that you get?

“You definitely can. You want to show customers that you have strengths, so that’s another thing that you could also look at. You can do that side by side, and also look at the positives, then start to put that on your website and create content around that.

That’s another interesting path, and it’s a way to find case studies as well.”

Is there a specific machine learning software that you use for this?

“You can do that on any software, and you could do it with Python or R.

The main thing is to understand what you’re trying to achieve and find out how to do that. If you’re trying to do sentiment analysis, look up how to do that with R or with Python. It all depends on what you prefer reading.”

How do you discover patterns through syntax and topic modelling?

“Looking at syntax is about combining those parts of speech – the verbs and the nouns, the adverbs and the adjectives. If you look at it, there are different ways you could combine these parts of speech to give you what you want.

Topic modelling is where you start to sentimentally combine these keywords, or lemmas, that you have found by analysing the parts of speech. You will have single words that came out from that analysis, and you can look at them individually, but you can also combine them with syntax.

If you have over 1,000 lemmas, how can you start to make sense of it? Some of these words already have semantic relationships. When you look at topic modelling, you start to think about how machine learning can combine the words that have semantic similarities. It looks at the frequency of the word, but it also looks at the semantic relationship to identify how it can cluster those words.

When it clusters them into words that share semantic meaning, you will be able to see that clearly. You can look at them and start to create themes out of them. That could be the topic exploration because it’s a little manual, but you already have the themes.

If it is your business, you will understand those themes. If you work in the food industry and you find out that the words that were semantically categorised together were ‘rice’, ‘chicken’, ‘beef’, etc., you can manually categorise that by looking at that semantic output to say that the theme is ‘meals’.

You still have to put a bit of human input into that, but machine learning can help you do the initial semantic categorisation, based on their meanings and occurrences.”

How do you uncover gaps and opportunities that your customers haven’t spoken about?

“You can use data from other places as well. However, if you’ve got a lot of data, you can make sense of it from the data that you have already.

To do a gap analysis, though, you could look at data from other places.”

How can you use query patterns to personalise content?

“Once you have found the themes, you understand the query patterns – the trends in the queries that people are asking for. Then, you can start to personalise your content to address these queries.

Create content that actually addresses their concerns. You’re answering queries that are personalised for your customers, to help them understand, and for you to create value for them. The main purpose of SEO is giving the user relevant, valuable content that is very helpful.”

Is this something that you can just do once and then you have enough data or is it something that you need to do on an ongoing basis?

“This is something that I would advise that you do at least once per quarter or once every 6 months, depending on what works for you.

It’s an important thing to look at because, in the long run, it helps your brand visibility, it gives you notoriety in your niche, and it helps you to provide relevant answers to the questions that your users have.

It helps your organic visibility in general because this is the kind of helpful content that search engines want you to provide.”

If an SEO is struggling for time, what should they stop doing right now so they can spend more time doing what you suggest in 2025?

“Look at the data that you have, like customer feedback and support queries, and start answering those questions. You could create themes out of it or answer most of them with your content by creating FAQs, blog posts, video posts, etc.

Answer the questions that customers are concerned about when you put out content. There are some questions that can require extra effort, so you can push those to the teams that can work on them. However, you can focus on those that you can easily address.

If customers are complaining about refunds, the questions could be, ‘How long does it take for me to get my refund?’ or ‘Do I get a full refund?’ Your content can answer those questions. Use your content to create valuable answers to what customers are actually looking for.”

Folashade Uba is an SEO Analyst at Maersk, and you can find her over on LinkedIn.