Machine Learning: Empowering. Marketers' Ability To Determine. Intent & Inform Targeting ... functionalities to auto
DECEMBER 2017 Predictions Magazine
Machine Learning: Empowering Marketers’ Ability To Determine Intent & Inform Targeting
ANGEL A KNIBB HEAD OF PAID SEARCH
EWELINA LEDO
SHIBLU AHMED
PAID MEDIA PERFORMANCE PAID MEDIA PERFORMANCE
KIT BIENIAS PAID MEDIA
DIRECTOR
DIRECTOR
MANAGER
IAN OH
DAN WALKER
JONATHAN KELTERER
NATASHA HOLE
SENIOR PAID MEDIA
SENIOR PAID MEDIA
PAID MEDIA
PAID MEDIA
ANALYST
ANALYST
ANALYST
ANALYST
SACHA DORF
TRISTAN SCHADLER
HANNAH REESE
L AUR A BALDRICK
PAID MEDIA
PAID MEDIA
PAID MEDIA
GR ADUATE TR AINEE
EXECUTIVE
EXECUTIVE
EXECUTIVE
PAID SEARCH
Each year when myself and my team stop and
either when it comes to paid search! This year, we
consider our predictions, I’m always in awe at how
came forward with an extensive array of predictions,
fast the year has gone. Perhaps that’s a good thing
so we upvoted those we thought would be most
– time flies when you’re having fun they say – and it
prevalent: artificial intelligence (AI) and machine
looks like 2018 isn’t going to disappoint on that front
learning.
Empowerment through technology For paid search, it should come as no surprise to
Personally, I see machine learning as the most
anyone that we see artificial intelligence – and more
exciting application of AI, and it’s certainly the
specifically, machine learning – as the most exciting
easiest to apply directly to paid search. It involves
topic on the horizon. But before going into how it’ll
using algorithms to determine intent and inform
affect the industry, let’s clarify what it comprises:
specific targeting based on data; and perhaps most excitingly for marketers, it means time savings and greater efficiencies. We’re already applying machine learning in ways that are considered business as
ö ö Artificial intelligence (AI) is an intelligent
usual:
machine ö ö Machine learning refers to the ability for a machine or computer programme
ö ö Smart Bidding and bid strategies in
to learn without being specifically
AdWords or DoubleClick (Google’s
programmed
campaign management tool)
ö ö Deep learning, which is often referred
ö ö Smart Display Campaigns to identify the
to in the same context as AI and machine
right users and creatives to meet a set
learning, is a type of machine learning
cost per ad (CPA) ö ö Adaptive Shopping to automatically break out top performing products ö ö Adaptive location targets to automatically break out top performing placements ö ö Data-driven attribution ö ö Google Analytics Smart Lists for Audiences ö ö Similar audiences for search
“We’re seeing an increase in the number of tools and functionalities to automate important daily tasks and checks, which seems to back up the move towards freeing marketers up to focus on strategic growth and testing” 2
PAID SEARCH
While many of us would like to have more visibility
Dynamic Search Ads wasn’t clearly defined yet, we
of the data and the reasons why the algorithm makes
see these becoming bigger in 2018. We also felt that
certain decisions, we find that the efficiencies it
similar audiences for search, and Google Analytics’
drives when dealing with big data are far greater than
Smart Lists, would also improve in terms of their
any necessary concern over the reliability of data or
performance and efficiencies. We also hoped that the
decisions. Admittedly, the algorithm doesn’t always
algorithms would become smart enough to take big
get it right, so currently we do need to continually
sales dates (such as Black Friday) into account and
monitor and tweak the rules.
automate bid optimisations for those, rather than having to pause the strategies and manually optimise.
When considering how we saw this changing in 2018, we were excited by the opportunities. We know a lot
The upshot of these changes is that we as marketers
of clients want to focus on driving new users, and
can really focus more of our time on finding
we feel there are a lot more opportunities we can
new opportunities to point the
take advantage of moving forwards to essentially
machines at, rather than manual
trick the system into over-valuing audience cohorts,
builds and implementations.
such as upweighting revenue driven from new users,
We’re seeing an increase
and pulling that into bid strategies to increase bids
in the number of tools
more aggressively for new users who are more likely
and functionalities to
to convert. Dynamic Search Ads, which allow us
automate important
to automatically insert an ad into relevant search
daily tasks and
results when we’re not bidding on the keyword, were
checks, which
identified as a second area of growth. We’ve used
supports the move
these in a few iterations now, but the algorithm has
towards freeing
improved significantly, and we’re seeing fantastic
marketers up to
results from these campaigns, with the added benefit
focus on strategic
of being able to use them as a notable source for
growth and
new keywords. While the group felt the role of
testing.
Towards better audience targeting There’s a clear desire within the industry to get to
There was consensus in the
a place where we have a deeper understanding of
room that confidence and trust in
the impact and make-up of audiences. As we use
the algorithms would increase, leading
algorithms more, we need to strike a better balance
to an increase in adoption rate across the
between volume and control. We believe we’ll see
industry. Perhaps reassuringly, many within the
a move towards having a less granular audience
team feel that the industry is already very closely
breakdown, but also speculated that Google or
entwined with machine learning, and that this was a
DoubleClick might consider creating an ‘Adaptive
key growth area for marketing overall, as well as paid
Audiences’ functionality, which would automatically
search specifically.
break out top performing audiences from a broad audience base (I must stress this is pure speculation at this stage).
4
PAID SEARCH
The rise of voice search Voice search is also a key area for machine learning to grow into. While we need a higher uptake of
Finally, we believe that this trend will also have
virtual assistants for this to fundamentally change
an impact on websites themselves, which will also
search, we’re already seeing small influences break
need to be optimised for voice search to align with
through. In 2017, the team reported an increase in
consumer search preferences. There are two key
searches starting with “hello Google” in search query
benefits we’ve identified with this:
reports, especially on mobile. This means we need to adapt how we’re building out long-tail keywords to reflect how users speak rather than how
ö ö Firstly, sites would become more
users type. Additionally, voice search
informative, giving Dynamic Search Ads
will also impact how product
more information to increase visibility
titles will need to be written for Shopping, to make
ö ö Secondly, site personalisation would
the product titles more
become more widespread
suited to how users describe products when speaking
While giants such as Amazon are doing a great
(rather than
job at surfacing products based on user interests
typing).
and previous purchases on its homepage, we think there’ll be a move towards making this easier for smaller websites to achieve as well. Ultimately, we’d be able to wholly tailor landing pages to our targeted audience, thus increasing relevancy and value for the user.
The role of PPC experts going forward It goes without saying, there’s been a lot of negative
more technical in terms of tagging, coding, and
press around this trend – we’ve all read about
programming abilities. We’re becoming more and
marketers’ fears that machines will learn to do
more data-obsessed, with a need to have at least
their jobs in a fraction of the time, and ultimately
an abstract understanding of how machine learning
make them redundant in the workforce. Nonsense
– and all the functionalities it touches – works.
says I, and (luckily for me!) so does my team. We
Ultimately, the role of PPC experts will become more
feel that our core job functions will shift, but we
focused towards strategy, testing, and growth – and
won’t be obsolete. We’re already seeing a shift
the machines will do all the manual labour.
in the workforce, with PPC experts becoming
Bring. It. On.
5
www.greenlightdigital.com The Varnish Works, 3 Bravingtons Walk, King’s Cross, London, N1 9AJ +44 (0)20 7253 7000
[email protected]