'Future Evolution of the Internet' delivered by Geoff Huston at Everything Op...
BrightonSEO - How to do ecommerce keyword research at huge scale
1. Tom Smits & Patrick
Reinhart
Bol.com & Conductor
How To Do eCommerce Keyword
Research At Huge Scale
@stimsmot | @askreinhart
2.
3. Let’s Talk About What We’ll Be Covering Today
• The Challenges Of Large Scale Keyword
Research
• Building A Keyword Calculator
• How To Cut Through The Noise & Determine
Where To Focus
• The PIE Methodology – What It Is & How To
Apply It To Your Research
• How to this in real life at a large e-commerce
retailer bol.com from the Netherlands
• Showing Results
5. The Sheer Amount Of Combinations That Exist
Categories
Sub-Categories
Modifiers
- Colors
- Styles
- Sizes
- Prices
Potentially Millions
Of Combinations!!
6. Let’s Say You Sell Jeans
For SKU’s That Are Just Jeans You Have:
Gender Color Style Size Price
Men’s Denim Skinny 30 $25
Women’s Black Boot 31 $50
Girls White Distressed 32 $75
Boys Green Relaxed 2T $100
Toddler Red Low Hip 12 Months $125
933,120
Possible Combinations!!
(for just 1 category)
8. Check For
Product
Modifiers
Break Out
Specific
Product Lines
& Their
Modifiers
Apply The #
of Modifiers
For Each
Product Line
Only Use
Modifier
Combos That
Make Sense
Determine how
many ways you can
modify a product .
i.e. Color, Size,
Style, Etc.
Add Up # Of
Multiplier
Expressions
To Determine
Total
Universe
Keep it clear and
organized by taking
the modifiers
discovered in the
previous step and
placing them into
individual rows within
their respective
categories.
Now it’s time to roll
up your sleeves and
go through each
category to
determine the
amount of modifiers
each has and enter it
into the calculator.
When you are doing
your math, you want
to make sure that the
combinations you
put together make
sense. This is the
first step in focusing
your universe.
Add up the totals
and create a
dashboard so that
you can see what
your total universe
looks like.
Let’s Build A Calculator
9. Here’s How It Can (should) Look
Main Category
Sub-Category
Modifiers
Total # Of
Permutations
10. Applying Rules To Your Set
Make sure your
combinations
make sense!
All of these
combinations map
back to a users
intent.
11.
12. Combine The Totals To Get Your Sites Universe
Pro Tip: Share this
information with
others to show
opportunity & inform
using search data!
14. What Does The Salty Octopus Do?
The salty is a cross-reference report that:
• Validates the # of keywords to focus on within your
keyword universe.
• Cross-references current performance with your top
organic competitors for each line of retail.
• It sorts the keywords by # of occurrences to show highest
relevance to help us internally prioritize current sets and
then search volume.
• We omit certain keywords in this set including:
• brands
• .com’s
• careers
• etc, you get it.
15. How Does That Cut Down On The Noise?
xxxx
Pro Tip: You can
also use this
report to identify
content gaps!
16. This sounds like a
lot of work!
Why are we
doing all of this?
Which way
do we go?
First star
on the left.
Thanks!
Wait..wha?
Gotta go!
18. What Is PIE?
PIE is a methodology that aids in the prioritization of your keyword set.
Protect
Expan
d
Page 1 Keywords
Assess the current state
of Organic Performance
and develop a plan to
protect it.
Improve
Page 2 Keywords
Identify low hanging fruit
and quick wins for
incremental increases in
traffic and revenue.
Page 3 and Beyond
Explore the untapped
market potential and
establish a plan to
capture it.
19. 19
• Webshop only retailer
• Strongest retail brand of the Netherlands.
• Most loved brand of the Netherlands
• >8 million active customers
• >16 million products; non food
• > 4.000 pick up points
• >19.000 sellers
• > 65 million visits per month
• >1300 colleagues
Use of keywords
at large scale
20. Managing Landing pages for Keywords at Large Scale
1 Crawling and Indexing 2 Optimizing titles
24. General rules to lower combinations:
1. Max 4 levels (4 clicks) deep via robots.txt
‘Black Giant Mountainbikes for Men’
’26 inch ladies mountainbikes with carbon frame
1 2 3 4
1 2 3 4
2. Never index certain facets like:
Delivery options i.e. ‘Tomorrow at home’
Price ranges i.e. ’10 to 100 euro’, ’11 to 100 euro’
3. Never index combinations
of the same facet:
i.e. ‘black and red mountainbikes’
i.e. ‘JBL and Sony speakers’
i.e. ‘apple and samsung smartphones’
6000 categories
2.500 facets
100.000s of facet values
How to lower
the
combinations?
25. How to lower the
combinations?
Facet 1
Facet 2
Facet 3
Facet 4
Facet 5
Facet 6
Facet 7
Facet 8
Facet 9
Facet 10
Facet 11
Facet 12
Facet 13
Facet 14
Facet 15
Facet 16
Facet 17
4. Only index the values for the
most popular facets.’
Facets Automatically
Ordered by popularity
26. 26
2500 facets – 100.000s of facet values
Facet 1
Facet value
Facet value
Facet value
Facet 2
Facet value
Facet value
Facet value
Facet 3
Facet value
Facet value
Facet value
Facet 4
Facet value
Facet value
Facet value
With Search
Volume
Without Search
Volume
Length
Wheel size
Target audience
2500 facets – 100.000s of facet values
27. 27
5. Only Index facets with search volume
Facet 1
Facet value
Facet value
Facet value
Facet 2
Facet value
Facet value
Facet value
Facet 3
Facet value
Facet value
Facet value
Facet 4
Facet value
Facet value
Facet value
With Search
Volume
Without Search
Volume
Length
Wheel size
Target audience
The largest challenge with keyword research at scale is the sheer size of the universe and the number of combinations that exist for your product set based on the # of Categories, Sub-categories, and the modifiers that exist within the products set such as colors, styles, sizes, price, etc.
The list swells or contracts depending on your product set, but either way you are looking at potentially millions of different combinations and permutations.
For example, lets say you sell jeans.
You have gender, color, style, size, price and then let’s say you have 5 different modifiers within each. If you look at the math, this can equate to over 933,120 combinations, and this is a relatively small set of modifiers!
Note: 933,120 possible combinations when we decide the same selections in different orders each count as different permutations.
Once you figure out what your universe looks like, you have to ensure you have the ability to cut through the garbage within the universe you discover and focus on what’s important and not try to boil the ocean.
Your entire campaign and strategy is based on the research you conduct, so not doing it properly can result in serious repercussions and lost time down the line.
Remember, trash in, trash out.
Don’t be this guy.
We build a keyword calculator that helps us map your SKU's to the different ways we know people search.
Some of you may say that step 3 is extremely manual, well, that’s because it is.
That’s not to say there isn’t a way to do it, I am sure there is, but my team and I are working on behalf of our clients and typically don’t have access to their databases.
We have tried to do it with different on page elements, i.e. faceted search filters, running crawlers through and trying to tie back different combinations, but at the end of the day the name of the game is accuracy and none of those exercises made us less than confident that we were covering the totality of the site.
As you can see, we are able to clearly see the total # of permutations for any given category and sub-category based on the amount of modifiers that exist for each.
A site of this size is going to return millions of possible combinations back, so how can we start to focus those permutations?
To cut through a lot of the noise make sure your permutations make sense!
I see a lot of bloat in large-scale projects based on multiplier combinations not being controlled, which is a very simple task and helps your math sound human.
Take a look at the different combinations that could exist and create rules that will return keyword combinations that map back to the users intent.
There is no need to create / research broken combinations, the closer we can get them to the searchers intent, the better.
Something that a lot of folks forget when they are doing research at this scale is user intent.
Just because your site is huge, doesn’t mean you HAVE to track millions and millions of keywords in every combination possible.
We do this research because it helps us figure out what our customers are thinking when they are going through their buying process.
If we map keywords back to their intent, we can then use Google as a vessel to give them what they care about, be that a product, a piece of content, an answer to their question, or a solution to their problem.
We’re not trying to create the biggest list, we’re trying to sort through that list to find out what our customers care about. That’s the point of these exercises.
Ok back to the calculator!
We run what we call a salty octopus – everyone laughs or looks at me weird, either way, win.
Besides being an irritable aquatic creature, the salty is a cross reference report that (read off slide)
The thinking here is if everyone is ranking highly, they are probably important to your business.
This is also the part of the process where we omit a lot of keywords that we know we don’t want to target such as (read from slide)
As you can see, the report cross-references your site along with your competitors (don’t use too many, 5 max so you don’t get lost in the data) so you can see where there are commonalities.
As was previously stated, the thinking here is if everyone is ranking highly, those terms are probably important to your business. This helps cut down on the noise.
As a bonus, you can also use this report to look for content gaps as well by looking for keywords they rank for and you don’t.
Why go through all of this?
The goal of these exercises is to differentiate between what we can do vs. what we should do.
The calculator tells us what the universe is and what all of the possibilities are,
While the salty tells us where we should focus so we actually get somewhere.
For the keyword set specifically, we apply what we call PIE Methodology.
PIE is a methodology that aids in the prioritization of your focus keyword set.
Protect - keywords ranking on page 1 of Google aka brand has visibility already so we don't want to lose that. these are typically the traffic driving keywords.
Improve - your opportunistic keywords, these are on page 2 of Google and with attention and strategy can move up to protect.
Expand - these keywords are beyond page 2 and in more of a developmental state, and require more effort / time and content creation. this is where the long-term opportunities exist.
Use Stephan’s sand castle analogy.
Depending on what your strategy, we can apply this to the entirety of the site, or just one line of retail, it allows for flexibility. That helps us categorize those keywords, then allows you to provide valuable and actionable recommendations that can lead to realistic, obtainable results.
Protect – Improve – Expand. PIE.
Approximately 300.000
We managed to increase inversion and CTR which for us shows this is an effective strategy and the concept works.
We managed to increase inversion and CTR which for us shows this is an effective strategy and the concept works.