12. More conversions at a
similar cost per action
with Performance Max
Source: Google data, Global, Ads, July - September 2021; Google data, Global, Ads, November - December 2022
15. Google Analytics 4
audience builder
Easily build and apply
Analytics audiences
directly as you build
campaigns in Google Ads
16. New customer
lifecycle goals in
Performance Max
Maximize revenue and profitability at every step
of the customer lifecycle, from acquisition to
retention with Performance Max.
17. Confidential + Proprietary
AI learns from
you.
And you learn
from AI.
You
Google AI
=
your business
multiplier
+
An AI-first approach
starts with you
18. The machine learning v
deep learning advantage
(Ian Kahn, Media Director & Automation Lead, Tug)
20. “Teddy Bears working on new AI research on
the moon in the 1980s”
Dall-E AI Art Generator 2022
Deep learning allows AI systems to “teach
themselves” how to best achieve an outcome.
Accounts using a modern search account
structure enable deep learning based on a
myriad of real-time user signals that go
beyond what can be set or even interpreted
by a human user.
AI Art and Smart Bidding: What They Have in Common
21. The following focus areas are KPIs for well-optimized machine learning friendly PPC accounts.
Goals:
3,000
Impressions per
ad group per
week
10 Conversions
per Campaign
per week
Use Broad Match
keywords & Dynamic
Search Ads
Remove KWs with <10
Impressions/week
Avoid low-
traffic landing
pages
Use 1 landing
page per ad
group
Remove manual
bid adjustments
Machine Learning Friendliness in Practice
22. Benefits:
Maximized smart bidding
performance through
deep learning
Improved scalability
and sustainable
growth
Optimizations have
quicker and greater
effects
Access real-time
user behavior
signals
Better insights
through statistically
significant data
The following focus areas are KPIs for well-optimized machine learning friendly PPC accounts.
Machine Learning Friendliness in Practice
23. 1.8x
ROAS
+14%
Revenue
+84%
Value
Modern Search Adapts Swiftly to Supply & Demand to Maximize ROAS for Chamäleon Theater
January 2023
Result: Using fully-optimized modern search with a ROAS target for
Chamäleon theater, the account automatically recognized a supply crunch due
to sold out shows and restricted daily spend to maximize ROAS. The result
was 90% of all available tickets sold for Chamäleon Theater in January with
paid search revenue exceeding forecast targets.
Daily budgets automatically restricted to avoid waste
Increase compared to forecasted value per conversion
Above target revenue benchmark
-38%
Costs
Above Target ROAS
24. AI in organic search
(Monet Blake, Senior SEO Director, Tug)
25. When You Think of AI in Search Today,
What Comes to Mind?
31. Closing Data Gaps For a Cookieless Future
Consent Mode
● Overcome data gaps when users
opt out of cookies.
● Utilize cookieless pings and 1PD
with machine learning to collect
aggregate data.
GA4
● Modelled user events
● Event based tracking
● Privacy and compliance built in
SST - Server Side Tracking
● Higher data quality as ad blockers and
intelligent tracking protection (ITP) have no
impact.
● Unaffected by browser types and/or version.
● Higher level of security
● Faster website due to reduced loading time
32. Utilizing Server Side Tracking to Maximize Attribution and Improve Performance
Outcome: Chamäleon Theater implemented Server Side tracking to
improve data collection, comply with user privacy, and improve digital
marketing performance. This additional data improved attribution and
smart bidding’s ability to optimize ad campaigns.
Increase in attributed conversions in Google Analytics
compared to webshop sales (from 60% to 85+%)
Decrease in attributable cost per conversion from Google ads
-50%
CPA
+30%
Increase
Increase in average monthly revenue from Chamäleon
Website
116%
Revenue
33. Next generation intent mining
& sentiment analysis
(Monet Blake, Senior SEO Director, Tug)
38. The Evolution of Intent & Sentiment Analysis
in Our Strategy
Keyword Research
Competitor Analysis
AI (NLP) to Categorise Large
Keyword Sets
Topic Identification
Keyword Research
Competitor Analysis
AI (NLP) to Categorise Large
Keyword Sets
Enhanced Topic Identification
NLP to Identify Competitor
Intent Coverage
NLP to Identify Keyword Intent
Old Strategic Approach
New Strategic Approach
When it comes to machine learning it’s important to understand the difference between machine learning and deep learning. Traditional machine learning takes a set of parameters and uses those parameters to learn to reach a conclusive result. Deep learning however, goes one step further and defines its own set of parameters. This makes is much more powerful and in this example, not only is able to determine what is or isn’t a cat based on an image, but can begin to differentiate between breeds, all on its own! The caveat here: deep learning requires a LOT of data to work well. This is what we are trying to enable with the Hagakure method.
A great practical example of deep learning is AI art. This image, created by the simple prompt “Teddy bears working on new AI research on the moon in the 1980s” is able to create a stunningly accurate image. This is because the DALL-E Art generator has been able to analyze millions and millions of images and learn from its users’ prompts, going above and beyond a set of human-defined parameters. At Tug, we use the same approach to optimizing our ads for deep learning.
Finally, only use one landing page per ad group to increase efficiency of your bidding strategy.
Finally, through the Hagakure method we get new clean consolidated data which allows us to generate much more valuable statistically significant insights than ever before.
Explain Bulllets
When we think of AI in search today, I think most of our minds might instantly jump to generative AI chatbots (like ChatGPT, or more recently Bing Chat or Bard) or advanced machine learning models.
However, AI's integration into the fabric of search goes back further than many realise.
In fact modern search has long been powered by artificial intelligence. This isn't just about generating responses but understanding and predicting user intent, providing more tailored and personalized search outcomes.
And as Kelsey mentioned, For AI in search to excel in delivering precise search results, it fundamentally relies on the quality of the content it has access to.
In fact, since 2019 we have been experiencing the results of AI, specifically natural language processing.
If we take this example here, we can see how the use of BERT, Google’s NLP model has been able to apply more contextual nuances to the original search term to understand that “adult” is being matched out of context, and pick out a more helpful result.
This is because BERT models can process words in relation to all the other words in a sentence, rather than one-by-one in order. So by looking at the words that come before and after it - which is rather useful for understanding the intent behind search queries (but more on that later).
(Bidirectional Encoder Representations from Transformers)
What you’re seeing on screen is Google’s response to generative AI-powered search that’s currently being tested in Search Labs.
This new way of searching will focus on conversational results as the results will be presented narrative text and the user can interact by asking follow up questions.
It also won’t replace traditional search results but appears above them. A user can still scroll down to view the SERP we’ve all been used to.
While it’s still in the experimental phase in Google Labs, there are murmurs that this will be released at the end of this year or early 2024. So there’s really no time like the present when thinking about how you can adapt your SEO strategy to get ahead
Earlier this year Google launched its beta version chat style search feature known as Search Generative Experience or SGE.
So how will this impact SEO?
Well, as we examine this example, we can see how the results have built out an almost seamless buying guide for this user looking for electric bikes. The inclusion of various intents allows for more holistic purchasing experience as answers to their considerations are already supplied
Another benefit people appreciate is that this experience is integrated into Search. They like that they can easily scroll and access a broad range of sources on the web, in addition to what they see in the AI-powered snapshot.
8.5 billion searches, 15% of which are brand new searches - so a plethora of opportunities - https://www.oberlo.com/blog/google-search-statistics#:~:text=But%20how%20many%20is%20that,Internet%20Live%20Stats%2C%202022).
If this goes ahead 6/7 topics emerging - AI 3rd biggest shift in the internet
Explain Bulllets
As we’ve just discuss, using AI to determine
If we use property in paris as an example it would traditionally be grouped under transactional as we would assume the main intent is to purchase a home or property.
But now, with AI based intent mining and Natural language processing we can uncover new topics & a deeper understanding of the intent when someone searches for properties in Paris. Serps have developed as a result and have expanded our opportunities to reach new audiences. So our strategies need to evolve too
So when what are all the things you think people might be looking to know when searching for "Property in Paris"
what do you think the intents behind someone searching for that is
However it is however it is extremely manual, as you can see in the video it takes a few prompts to do this and is
When discussing the capabilities of our tool, it's essential to highlight its progressive approach to understanding user intent. Traditionally, tools operated on a straightforward one-to-one mapping: one keyword equaled one specific intent.
In our tool's framework, while keywords are still a foundational input, they no longer dictate intent in isolation. Instead, we've moved beyond this simplistic model. And here's where it gets even more interesting: We utilise technologies similar to those employed by giants like Google. This allows us to delve deeper, not just into intent, but also sentiment. By harnessing this advanced technology, we can better comprehend the nuanced meanings and emotions behind search queries.
The result? A richer understanding that empowers us to craft more effective and tailored content strategies, ensuring our clients stand out in today's competitive search landscape.
Mindset shift - less keywords more topics
Moderated and unmoderated model
Ie. Like Google we can input then train the model and ‘educate’ it on what you mean