In this session, Billy will explain why marketers should approach YouTube with the same amount of rigour and data-mindedness that they do with Google. He'll also explore the different data types that everyone should explore when it comes to YouTube, going far beyond simply keywords.
Join this session to:
- Understand the data and insights available from YouTube
- Get to grips with the key considerations for optimising YouTube content - from titles and descriptions to product links and playlists
- Learn how to use data to build a strategy for new content, and prioritise video content ideas
- Uncover the key metrics to determine success on YouTube
7. Source: Backlinko
Low Impact Medium Impact High Impact
● Keyword-Rich Tags
● Keyword-Optimized
Titles
● Keyword-Optimized
Descriptions
● Subscriber Count
● Number of Likes
● Subscribers
Generated
● Video Comments
● Length, i.e. longer
ranks over shorter
● Video Shares
● View Count
● Filmed in HD
This is how what we know about how YouTube
weights all of the different ranking signals
9. However, you can infer things from Google
properties and products
Google Video AI has
machine learning models that
automatically recognize
objects, places, and actions
in video.
Google Video AI
The Google Cloud Speech
tool is a machine-learning
tool that specialises in
recognising speech in all
languages.
Google Cloud Speech
10.
11. All of this is data that we should
be collecting and optimising.
12. As we see it, there are 3 types of data needed
to increase YouTube visibility
Data Type What it means How we get it Deliverables
1 Top-level keyword data Topic +
thematically
relevant data
YouTube keyword
research
Idea, title +
summary
2 Granular keyword data LSI + long-tail
keywords
Topic research +
LSI generators
Script + video
description
3 Visual data Objects,
backdrops,
actions, animation
style, visuals
Video research Storyboard
14. We use traditional KWR to get this data
Keyword
Research
Categorise
our queries
Ideate!
Overlay other
metrics
We do YouTube
specific keyword
research
We then group
queries together into
larger categories
We then add other
performance metrics
for analysis
Lastly, we ideate
based on the data
we have collected
15. But we use other metrics to better quantify
the opportunity
This is the available
search volume for all
queries that fall into these
categories
1.
Search volume
Engagement is based on
the total number of video
views that videos in this
category achieve
2.
Engagement
Difficulty tells us how hard
it will be for us to rank for
videos that fall into this
category
3.
Difficulty
16. To scale this easily, we use our machine
learning ideation tool, Solomon
Search Volume
Ahrefs
Engagement
YouTube
Difficulty
Custom Script
18. Create overall
structure for video,
including script +
storyboard
Create video idea, title
and summary
All identical and
thematically-related
queries are grouped
together
It then becomes quite easy to drill down into
categories and ideate from the data
19. Hook + Explainer – Show Information – Brand Name
And integrate that data into video titles in a
structured format
21. We should treat the script like it’s on-page
copy
Regular keywords
identified in the first step
that need to be spoken
about
1.
Keywords
Latent semantic
indexing keywords that
are related to the topic
chosen for the video
2.
LSI Keywords
22. We don’t always need to reinvent the wheel,
there are existing tools that can be used
23. It’s important to create structure for
descriptions
1. Intro sentences - this should be 2-3
attention-grabbing sentences
2. Detailed video description - 200 words
to explain the video further
3. CTA - any relevant call-to-actions,
including further reading, resources etc
4. Links - links to social media profiles etc
25. Research videos for that topic and see what
common themes you can aggregate together
Text
explainer
of product
benefits
The product is
in shot
Hands indicate
the product is
being used
Buildings show
an urban
location
26. Google Video AI has
machine learning models that
automatically recognize
objects, places, and actions
in video.
The Google Cloud Speech
tool is a machine-learning
tool that specialises in
recognising speech in all
languages.
But using Google products, we can start to do
our own, more in-depth analysis at scale
Google Video AIGoogle Cloud Speech
27.
28. Video
Rich
Annotations
Features
Logo
Scene Changes
Camera
Phone
Buildings
Hands
Text
Number of Scene Changes: 6
Logo Screentime: 3 seconds | 7%
Logo first shown at: 00:38
Camera Screentime: 34 seconds | 83%
Phone Screentime: 3 seconds | 7%
Buildings Screentime: 4 seconds | 10%
Hands Screentime: 21 seconds | 51%
Text Screentime: 32 seconds | 79%
The video has the following attributes:
● Explainer/Product information video (due to high screentime of both the ‘hands’ object and text)
● Product shown prominently (due to the high screentime of the ‘camera’ object)
● Closing with Brand Logo (video ends with the shining logo sequence)
● Product detail view throughout the video
Data
Aggregation
29. Video AI pulls
features from top
performing videos
Detail goes into the
brief for creative
direction
Features analysed at
scale and
recommendations created
We then feed this data into our creative briefs
and storyboards
31. Treat YouTube with the same data-focused
methodology that you would Google
Data Type What it means How we get it Deliverables
Top-level keyword
data
Topic + thematically
relevant data
YouTube keyword
research
Idea, title +
summary
Granular keyword
data
LSI + long-tail
keywords
Topic research + LSI
generators
Script + video
description
Visual data Objects, backdrops,
actions, animation
style, visuals
Video research Storyboard
32. Thanks so much for
listening!
billy.leonard@croud.com if
you have questions :)