Personalization! The word that digital marketers and vendors cannot stop saying. "Personalization is the future." "We need personalization!" Well, at Clearhead, while we are champions of better digital experiences through data, we reverse engineer every solution to the formative question -- "What problem are you solving?".
Recently, we've been working with data-driven marketers to think of personalization, not as the answer or the future, but as a means for solving segment specific problems through the leveraging of explicit and implicit end user data. We've tried to demystify the term, explain it in relation to terms like "AB/MV Testing" and provide some ways to approach segmentation.
Most importantly, though, we've discussed personalization through the lens of Problem Solution Mapping (PSM). PSM is a unifying method for continuously optimizing around a common set of goals, problems & solutions, researched & validated with data every step of the way. Problems are mapped to clearly defined goals and are then rigorously researched, via usability research, analytics & customer feedback, to distinguish noise from signals.
Learn more: www.clearhead.me
2. 2
problem / solution mapping
A unifying framework for continuously optimizing around a common set of goals,
problems & solutions, researched & validated with data.
1
2
3
goals
PS
Ps
pS
ps
problems
1
2
3
4
hypotheses
prioritized by data
1
2
3
4
learnings
validated by data
investments
UX
product
trading
marketing
3. 3
it’s still an
experiment
everything is an
experiment
promotions + campaigns
merchandising changes
copy changes
UX + IA changes
new SaaS additions
personalized experiences
everything
if everything is an
experiment,
then these are
questions to live by
What problems are the
changes solving?
1
How will you know if the
change was successful?
2
4. 4
what is
personalization?
The customization, targeting or adaptation of
content and/or experiences for end users
based on implicit or explicit attributes of more
refined segments.
what is a/b testing?
A method of comparing a variation to a control
to determine if the differences observed in the
sample are statistically likely to survive in a
larger, general population or data set.
1
2
3
a
b
ab testing to segmentation segmentation to a/b testing
a
b
1
2
a
b
6. 6
increase mobile conversion rate
by 10% by the end of 2016.
ask yourself:
• target set?
• clearly understood?
• time based?
• realistic?reduce mobile bounce rate by 15% by
the end of 2016.
increase mobile revenue per visit by
5% by the end of q3 2016.
improve mobile net promoter score by
10% by the end of q1 2017.
goals
7. 7
increase mobile conversion rate
by 10% by the end of 2016.
problems
ask yourself:
• is it a root problem?
• who does it impact?
• where and when
does it impact them?
• how do you know it
is a problem?
users find it hard to click on our filter &
facet functionality on their smart
phone.
it is challenging for users to look at
alternative product shots on our
mobile PDP because the thumbnails
are so tiny.
we frustrate mobile phone users with
two extra steps — interstitial cart and
account options — before getting them
to checkout.
goal
8. 8
solution hypotheses
ask yourself:
• I believe that…
• If I am right then…
• could a designer/
developer/analyst
reasonably begin
work based on the
hypothesis?
I believe that if we skip the interstitial cart
page for smart phone users and redirect
them to checkout once they add something
to their cart, they will be less likely to waver
in their journey and bounce. If I am right,
then, mobile conversion rates for for smart
phone users will increase by 5%.
I believe that if we eliminate the “sign
up”option at the beginning of check-out for
all unauthenticated users on smart phones,
they will be less intimidated by the
prospects of filling out extra form fields and
will be more likely to purchase. If I am right,
then mobile conversion rates for for smart
phone users will increase by 10%.
we frustrate mobile phone
users with two extra steps —
interstitial cart and account
options — before getting
them to checkout.
problem
9. 9
segments: let’s get real
where should we start?
how do you get from qualitative
personas to definable
data segments?
is predictive segmentation
a real thing?
anybody doing amazing
omni-channel personalization?
for problem
research
for hypothesis
development
should we just randomly
explore segments?
10. 10
When you are personalizing, you are still
experimenting.
Map personalization hypotheses back to clearly
defined goals and validated problems.
Key considerations for personalization
• Segment size
• Segment value
• Manual v algorithmic
key takeaways
Segment definition and exploration takes time.
There’s no magic button (yet).
Experiments come with risk and investment.
Multi-channel customer data layers are increasingly a
practical reality!
1
2
3
4
5
6