Tips and Tricks for Understanding User Search Intent

Author: Charles Meaden
Tips and Tricks To Really Understand
Charles Meaden
Digital Nation

View Slide

No More Than 30 Seconds On Me
• Started Digital Nation 26 years ago
• Specialise in search (paid and
organic) and analytics
• Reside in the strangely named The
Mumbles (Swansea, Wales)
• If you’re looking to move out from the
city, you get views like thse

View Slide

What’s The User Intent?
• The phrases users type into
search boxes are a goldmine of
useful data
• The combination of words lets
us know precisely what the user
was looking for
• Also, where they are on the
journey
– Best lightweight hiking boots
• Informational query
– Roclite 345 price
• Transactional query

View Slide

Quickly Establish What People Are Looking For
• The most searched term on UK councils sites is…
• Recycling
– People need to know what weeks and what recycling to put out

View Slide

Top Level Search Intent
• The most common are
– Navigational
– Informational
– Commercial
– Transactional
• Google has a slightly different model
– Know” queries:
– “Do” queries:
– “Website” queries:
– “Visit-in-person” queries
• My challenge with them is that they are way too broad

View Slide

I hate word clouds…
• Single words that is

View Slide

2 Or 3 Words Provide Context
• In this case adding
– Gender
– Brand
– Colour
• Tells us far more about what
someone is looking for

View Slide

There Is A Lot Of Data To Be Crunched
• For two major UK retailers we’ve extracted over 750,000
unique search terms via the Google Search Console API
• For our GoSimpleTax client with a lower search volume per
month, you’re still looking at
– 82,000 unique search terms
– 7,800 different words
– 26,000 question phrases
• What, where, how

View Slide

Google Search Console to Big Query Export
• Google slipped this
out in May 2023
• Does a daily export
to Big Query
• Tested it across 3
different sized
clients
• You can do some
really clever stuff
like generate
ngrams
• If you are at
involved with search
you should enable
this

View Slide

Tip 1: Every Data Set Is Slightly Different
• We’ve worked on search query intent projects for a wide range of customers
• The phrases people use to find them and enter in their internal search engines are slightly
different
• Take the time to build an adaptable model which you can use time and time again

View Slide

Tip 2: Eyeball The Data
• Before running any automated process
over the top, use your eyes
• You’ll get a feel for the data
• You spot patterns and anomalies in the
data straight away

View Slide

Tip 3: Find The Most Used Words
• This will allow you to spot quickly any
potential issues
• In this example we’ve got
– Plurals
– Apostrophes
– Synonyms
• You need to decide for each project
which ones to change
• The list was generated using the
Hermetic Word (and Phrase) Frequency
Counter Advanced Version

View Slide

Tip 4: Clean Up Your Data
• Assume your original data isn’t perfect
• Lowercase all your search terms
– Thank you GA4 for not lowercasing search terms
• Remove unnecessary spaces
• Remove apostrophes and any unnecessary special
characters
• Correct common spelling mistakes
• My favourite tool for doing this is Analytics Edge
– Excel plugin for Windows
– Mac and standalone versions in Beta
• Allows me to create macros that automate common text
cleaning tasks

View Slide

Tip 5: Depluarlise Your Words – Part 1
• What we are looking for is the intent
• Tracksuit and tracksuit is the same word
• Women, womens and women’s is the
same word
• This regular expression will do the trick
• \b(\w+)(s)\b
• However there is a gotcha…

View Slide

Tip 5: Depulararise Your Words – Part 2
• Taking off the s works for most words
• But not if they are a name
– Brands
– Citys
• Eyeballing the data will help you spot
these
– Reusable queries are even better
• Have a routine that adds the S back to
these

View Slide

Tip 6 – Make A Decision On Synonyms
• Some phrases are treated the same by search
engines
– Does your internal search do the same?
• Kids and Childrens
• Ladies, Ladys and women
• There is no hard and fast rule here, apart from
common sense

View Slide

Tip 7: Two Word Combinations
• Some words were meant to go together
• Splitting these up makes analysing them harder
– HMS Daring
– Ralph Lauren
– San Francisco
– Trailfly 270
• Add a hyphen between these words
– hms-daring
– ralph-lauren
– san-Francisco
• These will now be treated a single words

View Slide

Tip 8: Word Order Doesn’t Always Matter – Part 1
• What’s the difference in intent between these queries
– womens french-connection jeans
– french-connection womens jeans
– jeans womens french-connection
• None what so ever - they’re all looking for the same thing
• Sorting the phrases into alphabetical order and deduping can massively reduce the number of
phrases you need to work with
• On a recent project, we cut the number of phrases we needed to analyse by 50%
• This is how we do it
– We do it in Analytics Edge using a macro, but the process could be easily adapted

View Slide

Tip 8: Word Order Doesn’t Always Matter – Part 2
1. Load your queries into a table
2. Take each query in turn such as ‘womens french-
connection jeans’
3. Split this into separate words each on a separate row
4. Sort alphabetically
5. Merge the rows into one row and separate by a space
6. You’ll now have a query that looks like this ‘french-
connection jeans womens’
7. Repeat across all queries
8. Deduplicate the data and total any numerical columns

View Slide

Tip 9: Look For Common Word Combinations (Ngrams)
• It’s easy to spot patterns if it’s just 100
phrases
• Harder at 1000,
• Impossible at 10,000 or more
• Couple of handy tools
• Hermetic Word Frequency Counter
– Best £30 you’ll spend I promise
• Analytics Edge
• Big Query ML.Ngrams function
– Calculate 2,3 and 4 word combinations
over millions of rows in seconds

View Slide

Tip 10: Build Up Libraries of Common Phrases
• Every project will be similar, but different
• We build up libraries of common terms
– A general level
– Client and Industry specific
• Some examples
– Question words such as what, why, how
– Colours
– Industry specific terms

View Slide

What Next – A Couple of Ideas
• Clustering
– Look for common patterns using machine learning
– https://www.keywordinsights.ai/
– Follow Lee Foot on Twitter @LeeFoot as he has done some amazing stuff using Python
• Missing Content on Your Sites
– Combine the data with data from website crawls and SERP scraping tools such as ValueSERP to find
missing content gaps
• Product Discovery
– What products and services are people looking for that you don’t have

View Slide

Thank You
• Find me on Twitter and Linkedin
• https://twitter.com/charlesmeaden
• https://www.linkedin.com/in/charlesmeaden/

View Slide

Tips and Tricks for Understanding User Search Intent

Tips and Tricks for Understanding User Search Intent

Charles Meaden

Other Decks in Marketing & SEO

Featured

Transcript