Webinar - Predict User Search Intent to boost campaign ROI

Join our CTO Alexandra Anghel to learn how to predict user search intent for your SEO or PPC campaigns at scale, with zero manual work.

Overview

Know WHY customers want something, so you can deliver the HOW.

Learn how we worked with one of our customers in the ad-tech space to predict users intent for a volume of around 5 billion search queries spread around 17 languages.

Leverage the same AI tool that helped them boost ROI for their campaigns and increase your own revenue and SEO or PPC campaign performance.

Here's what you'll learn in this webinar:

✓ How to use Google Ads and/or Google Search Console to create a CSV file
✓ How to use MorphL (FREE) to extract user intent for all the queries in your CSV file
✓ How to filter the queries and get the most relevant insights to improve your SEO and PPC campaign content

Get the checklist

Save tens or hundreds of hours of manual work. Instead of manually assigning user intent to keywords in SEO or PPC campaigns, leverage the free MorphL tool to extract it. Find out how it works and how to use it.

Download for free

Webinar takeaways

icon
The main types of User Search Intent

User Search Intent states the user’s goals or intentions when using a search engine, and includes information on the links they access and their activity on the target website. To optimize the results provided, we have split the User Search Intent into four different categories: Informational or Awareness – this category is related to finding information on a topic; Navigational – also called visit-in-person, meaning finding a place nearby or other types of local information; Transactional – accomplishing a goal or engaging in an activity; Consideration – this category is in-between informational and transactional intent.

icon
How Machine Learning can help reveal User Search Intent

For a small volume of data we can label the queries manually, but this approach doesn’t work for larger data sets. AI can help with attaching intent to large volumes of queries when we don’t have an exact match for a known word.

icon
The free method to generate search intent for your keywords

At MorphL, we created a practical AI application that uses Machine Learning to help you save hundreds of hours of manual work. The process is simple: you upload into our demo your CSV file – that you can easily export from Google Search Console or Google Ads – and The ML model generates predictions and includes your queries in the four categories. You can then download the new CSV file that was sent to your email, and filter the predictions to find the information you need.

icon
How a CEO or a PPC specialist can use this information in real life

This data can be used to segment Ad campaigns based on user intent and improve the click-through rate and ROI. It can help you personalize landing pages for users based on the Ad campaign that drives them to the page, or create specific types of content to increase conversion rate. The downside of MorphL’s AI model is that, currently, it only works for English queries.

Read the transcript

Ciprian Borodescu: Hello, everybody. This is Ciprian Borodescu. I'm co-founder and CEO at MorphL, and it's my pleasure to welcome you to our second webinar - Using AI to Predict User Search Intent for SEO or PPC campaigns. Now, for those of you who don't know, MorphL is a platform that uses AI to predict user behaviors in digital products and services with the end goal of increasing product metrics, and business KPIs.

Ciprian Borodescu: Today, I'm here with Alexandra Anghel - CTO at MorphL. Thank you so much for being here! I know that you have a lot of things to tell us, so I suggest we dive right in.

Alexandra Anghel: Thank you for having me. It's my pleasure to talk about our user search intent model and some use cases where it can be applied. I'm not going to go into very much technical details, and I'm trying to keep the information as practical as possible.

Alexandra Anghel: Our challenge was to predict user search intent for millions of English search queries. Naturally, a user search intent is tied to the user's activity in the search engine and the links that they click, and their activity on the target website. But we didn't have access to that kind of data. Our goal was to extract insights as rich and as actionable as possible from the queries themselves.

Alexandra Anghel: If someone searches iPhone 8, what do they want to do? Are they looking for product recommendations, specifications, use? Do they want to see some pictures? The fact is that we don't know for sure, but we can make an assumption based on several intent categories.

Alexandra Anghel: For this use case, we decided to split user search intent into three categories, with a fourth, “Consideration” being added at a later stage. These types of intent correspond to the layers in the marketing funnel, and we have: "Informational or Awareness", which is related to finding information about a topic. We have "Navigational" queries, which are also called, "visit-in-person", and are related to finding a place nearby or other types of local information. The "Transactional" queries are related to accomplishing a goal or engaging in an activity. And finally, "Consideration" queries are in-between informational and transactional intent.

Alexandra Anghel: What's challenging is that the intent behind these queries is not always clear, and inherently, some of them will have a multi-intent. For example, if someone searches for the word, "hotels", the intent depends on the context. If they're using their mobile device, their intent can be navigational - which is finding a hotel nearby - or related to consideration intent, which is making an online reservation. It could also be a transactional intent, although this generic search suggests that the user might not be ready, yet, to make a booking.

Alexandra Anghel: Of course, if we're talking about small data volumes, where we have a few hundred queries, we can manually label them or use a dictionary search. For example, if a query contains the word "buy", we can say that it's a transactional query. But, this approach doesn't work for tens of thousands of queries, not to mention millions, which is the data that we have available. This is where Machine Learning can help with attaching intent to large volumes of queries when we don't have an exact match for a known word. For example, a query that contains a variation of the word "buy" or a word that's close to it, such as "price" or "purchase", will still be labeled as transactional.

Alexandra Anghel: Now, I'm going to show you how you can use this model on your own query. I have here a CSV file with the English queries - and please keep in mind that if you are exporting the queries from Google Search Console or Google Ads, you'll have to clean up the exported file and keep only the queries with one query per vote, like you see here. Next, you can upload the CSV file in the form located on our user search intent demo page. It will take a few minutes for the model to generate predictions, and after they are ready, you will receive an email with the predictions as an attached CSV file. The email also contains a link to a complementary Google Sheet that you can use for analyzing the predictions. You'd have to download both the complementary Google Sheet and the CSV file with the predictions and upload them to a Google Drive folder, like you see here.

Alexandra Anghel: The CSV file also needs to be converted into a Google Sheet. Once they are uploaded on Google Drive, you can import the CSV file with the predicted user search intent in the Raw Predictions tab from the complementary Google Sheet. You'll receive a notification for allowing access and after you confirm, the queries will be imported in the Raw Predictions tab. As an alternative, you can also copy and paste the queries with the prediction in the same tab.

Alexandra Anghel: Next, you can use the Prediction Filters tab to extract queries that have a certain confidence score. For example, in this case, we are filtering queries with a probability higher than 75% of being transactional. And finally, in the Filtered Predictions tab, we have the queries that were given.

Alexandra Anghel: Going back to the Prediction Filters, we can apply the same process for filtering navigational queries with an 80% confidence score. So, we can change the score to 0.8 for navigational and set transactional to 0. Since queries are multi-intent, we are not limited to filter one intent particularly, at the time. We have the possibility of setting the filters to 0.5 for all intent categories, and in this way, we can filter queries with multi-intent. Going back to our navigational example, if you go to the Filtered Predictions tab, you'll see the navigational query.

Ciprian Borodescu: Okay, so now we have the predictions and we can also filter those predictions and extract a subset of queries that are transactional or informational, and so on. But, what's the end goal as a CEO or a PPC specialist? How can I use this in real life?

Alexandra Anghel: There are very practical examples of how intent categories can be used in your life. You can segment Ad campaigns based on user intent and improve your Click-through Rate and ROI for them. You can personalize landing pages for users, based on the Ad campaign that drives them to the page, or create specific types of content that address distinct user intent to increase conversion rate. And one more, you can modify or update a page or a resource to attract qualified traffic instead of visitors that don't convert because they're not ready to commit.

Ciprian Borodescu: Excellent!

Alexandra Anghel: To give you a specific example: one of our customers, an AdTech company, has used our intent predictions to personalize the ad copy depending on the intent category, and this led to a 7% increase in the Click-through Rate and a 28% increase for their ROI.

Ciprian Borodescu: Awesome! Now, Alexandra, can you tell us a little bit about the limitations of this user search intent model?

Alexandra Anghel: One of the limitations is that the model works with only one language at the time, and right now we can generate predictions for English queries, but we're experimenting with other languages, such as Spanish, French, German, and Italian. So, if any of you listening out there are interested in a specific language and have a big data set to work with, please reach out to me or to Ciprian.

Ciprian Borodescu: Alright! This was super interesting and practical. Thank you for taking the time to explain how the user search intent model works.

Alexandra Anghel: Thank you for having me and I'm looking forward to hearing your thoughts on the predictions we're generating for your queries. And I'm more than happy to hear any suggestions or feedback you might have.

Ciprian Borodescu: Thank you all for watching, and be sure to follow this webinar series. We're going to post a new webinar every month. The next one is scheduled for July 24th, so stay tuned for the next videos, and make sure to subscribe or leave a comment if you have any questions. See you next time!

Supported by: