馃攳
The Secret to Mass Personalization & Personalized Content with AI (2018) | AI for Business #3 - YouTube
Channel: unknown
[5]
Hi and welcome back to another of our
videos on artificial intelligence. We
[10]
know that personalization is a powerful
way of influencing consumer behavior but
[14]
a few years ago the generation of
personalized content used to require
[18]
much more effort on the company side. Now, the increased availability of
[22]
artificial intelligence, combined with
marketing automation, made sophisticated
[26]
segmentation less costly and faster to
implement. So, let's get into depth and
[31]
see how AI can improve your
personalization strategy. First of all,
[35]
computers are now able to perform
profile and classification of customers,
[39]
based on the data that they actively or
passively provide, the so called digital
[44]
footprints. In the last few years this
practice became known as personalization
[49]
at scale. McKinsey has summarized what
personalization means to customers
[53]
according to this simple formula. So
let's analyze this relationship. It is
[57]
directly related with the relevance of
the content, for example consumers have
[62]
reported to preferred suitable
recommendations that they wouldn't have
[65]
thought by themselves. The traveling
industry makes use of text mining to
[69]
create and test recommendation systems,
based on the similarity of the
[73]
destinations. For this, they use the
reviews that previous travellers have
[77]
made, their most co-occurring words to
describe a particular destination. With
[82]
that they can set retargeting campaigns
suggesting similar destinations to
[86]
previous travelers. Another factor in the
previous formula is timeliness. Users
[90]
report to prefer to be approached when
they are in a shopping mode and because
[94]
nowadays people are always using their
smartphones to check their emails.
[98]
Sending automatic messages at midnight
might not seem very timely, or it just
[103]
don't make sense to their daily schedule.
So, how do we make these things in a way
[107]
that does not deteriorate trust and does
not interfere in their privacy? After all,
[111]
what are the boundaries of digital mass
persuasion?
[114]
Recent research has shown that people
usually do not behave logically when it
[118]
comes to privacy. For example, we often
share intimate information with
[122]
strangers, while we keep secrets from our
close ones, the so called privacy paradox.
[127]
This helps to explain why that on
average just 65 like Facebook pages
[132]
allow behavior
analysts to understand someone's
[135]
personality traits better than friends
do, 120 to understand them better than their
[140]
family members an 250 to understand them
better than a partner or spouse.
[144]
Nevertheless, behavioral science has
identified some factors that predict
[149]
whether people would be okay with the
use of their personal information. I
[152]
will illustrate this using some
experimental examples. What happens when
[156]
we know that a friend has reviewed
something personal about us to others.
[160]
We usually get upset and those norms can
also be applied to our digital life. The
[164]
researchers use the dimensionality
reduction method to find groups of
[168]
practices that consumers tend to dislike.
They did that based on a list of common
[173]
ways in which Google and Facebook use
consumer personal data to generate ads.
[178]
The results suggest that obtaining
information from third party platforms
[182]
and deducing information about someone
from analytics, are more frequently
[186]
disliked. Previous research has also
tested whether varying the copywriting
[190]
would also affect consumer behavior,
within the same ad but with different
[194]
disclosure designs. So in one design a
group of participants saw an ad that had
[199]
the following copywriting - you are seeing
this ad based on information that you
[203]
provided about yourself. A second group
of subjects saw - you are seeing this ad
[207]
based on information that we inferred
about you and a control group saw the
[211]
business as usual no disclosure ad.
Participants who viewed the ad framed as
[216]
inferred behavior analytics
showed much less interest in purchasing
[220]
than the other groups did. Also,
booking.com tested variations of
[223]
copywriting to check which was more
effective to increase conversions in an
[227]
email marketing campaign. In their case
the less intrusive variation was less
[232]
effective than the one obtained by an
LDA model, a type of natural language
[237]
processing applied to the users reviews.
This variation said that based on your
[241]
past trips their team of travelers
scientists thought that you probably
[244]
have a passion for a romantic landscape,
local food or shopping. I find this
[249]
example interesting because it shows how
important it is to test different
[252]
variations of copywriting and looking at
how humans react to them. Artificial
[257]
intelligence is helpful here because
machine learning improves our ability to
[261]
predict what person will respond to what
persuasive technique, for which channel
[265]
and at which time. This
combination between behavior analytics
[269]
and automation is now called digital
nudging. For example, digital nudging can
[273]
help some companies to reframe their
services within a personal advice
[277]
approach. This can make it easier for them
to acquire customer data. Also, it helps
[281]
to comply with the GDPR regulations,
concerning privacy issues and
[285]
information sharing. If you want to know
more we have a video about that so check
[289]
the link in the description. Also in a
recent podcast hosted by the channel
[293]
behavioural grooves, Rebecca blank the
chief behavioral officer at merits,
[297]
discussed in detail how customers might
react differently, according to the way
[302]
companies communicate with them through
personal content. In summary there are
[305]
three takeaways. Any disclosure is less
creepy and will convert better than no
[310]
disclosure. Deduced information on the
customer will convert less than an open
[315]
explanation of why they are seen a
specific ad and finally, this might seem
[319]
corny, but trusts is a key factor in
dealing with this new world of shared
[323]
information and data gathering. In
conclusion we see that personalization
[327]
is an efficient way to influence
consumer behavior, especially when it's
[331]
powered by artificial intelligence. But
we have to mitigate backlashes by
[335]
testing different layouts, image and text
that provide information on how your
[339]
personalized content was generated. What
do you think about data privacy? For
[344]
example I think it's psychological
profiling but please let us know your
[347]
opinion in the comments below!
Most Recent Videos:
You can go back to the homepage right here: Homepage