Dave Rose, Brooks Bell
The opportunity to personalize experiences across devices and marketing channels is upon us. To move toward omnichannel personalization and micro-segmentation, we need to take the best of what we’ve learned through a decade of a/b testing and apply those practices to a/b personalization. In this session, Sr. Optimization Analyst David Rose will highlight omnichannel examples and a vision for personalization that combines what we’ve learned from testing with the opportunities that new tools afford us in the realm of personalization.
2. The recognized leader in
enterprise-level A/B testing, targeting,
and personalization services
SOME OF OUR CLIENTS
3. A/B Testing
Whole-Site Optimization
A/B Testing has dominated site optimization for
almost a decade.
Massive wins becoming less and less likely
Segmented analysis dominates conversation
4. Segmentation
Opportunity has been observed…but often discovered
after the fact during analysis.
How many is too many? This discussion alone, can
slow down an optimization program.
6. 1 : most
EASY
Where most
programs are
1 : few 1 : 1
HARD
Where most
programs
should be
IMPOSSIBLE?
Where most
programs think
they should be
Personalization
11. “STITCH” YOUR DATA
USE BEHAVIORAL DATA ACROSS DIFFERENT PLATFORMS
STEP 2:
SHIFT FROM VISIT TO
JOURNEY
12. Tools to Help
AudienceStream can connect across channels
and platforms
Information about the person, not the computer
13. Long-Term Behavior
Tagging actions for long-term discovery takes
your data teams to a new level of prediction
Provide and test personalized experiences when
the audiences reach an actionable size
15. SELECT VALUABLE ACTIONS
1 TO 1 PERSONALIZATION MIGHT NOT BE THE MOST PROFITABLE
STEP 3:
PICK YOUR BATTLES
16. 1 : few
How do we determine the profitability
of personalizing to each group?
ACCESSIBLE
DIFFERENTIAL
ACTIONABLE
MEASURABLE
SUBSTANTIAL
PROFITABLE PERSONALIZATION
17. Accessible
Is it possible to reach each group efficiently?
ADAMS5 ESSENTIAL ELEMENTS OF
PROFITABLE PERSONALIZATION
18. Differential
Would all groups actually respond
differently if exposed to different campaigns?
ADAMS5 ESSENTIAL ELEMENTS OF
PROFITABLE PERSONALIZATION
19. Actionable
Do you have a product to fit the specific
needs of each group that you identified?
ADAMS5 ESSENTIAL ELEMENTS OF
PROFITABLE PERSONALIZATION
20. Measurable
Can the impact of personalizing
to each group be measured?
ADAMS5 ESSENTIAL ELEMENTS OF
PROFITABLE PERSONALIZATION
21. Substantial
Are the groups large or valuable enough to
warrant the added expense of personalization?
ADAMS5 ESSENTIAL ELEMENTS OF
PROFITABLE PERSONALIZATION
24. It’s a Program;
not a Project
Personalization will be a culture of discovery and
optimization
Never one and done – Evolve with the customer
25. Optimize the Program
New tools, new data, new opportunity!
Keep discovering
Repeating the cycle isn’t starting over...it’s
optimizing your efforts for the best results.
In surveying the current landscape, we have come to the realization that many companies aren’t personalizing. In fact, our stats say that 49% of companies aren’t personalizing. I believe that is because companies fail to see that incremental steps to personalization can actually be successfully implemented and prove profitable.
If you look at personalization through the classic trichotomy of problems (easy, hard, and impossible), you will see that all personalization efforts would fall into one of those three groups.
“1:most” is easy. These type of efforts include simple segmentation such as new/returning, male/female, etc. This isn’t really what most people would call true personalization. It is not advanced enough to offer a subset of relevant products. As a result, the majority of these efforts will not return value on the investment.
“1:few” is hard but possible. This is where the value lies for most organizations. “1:few” efforts solve a problem that is difficult but one that can be addressed with diligent effort, the right resources, and time. The majority of this webinar will focus on profiting from this 1:few approach.
However, I would like to contrast this with what most marketers envision is the final frontier, and that is “1:1” personalization. “1:1” is where the hype lies, but where value remains hidden. In fact, from a data science perspective, “1:1” is a misnomer in the sense that the intent of most modeling efforts are to pair a infinite number of visitors to a finite number of outcomes. If you have infinite number of visitors, “1:1” would imply you need an infinite number of experiences. In the marketing world, we know is infeasible on every level.
Due to grandiose “1:1” concept, personalization seems so daunting from a resource or investment perspective, that many companies aren’t getting anything off the ground.
With that, the tricky part of 1:few becomes how many groups do we form for our personalization efforts. There are mathematical ways to determine it and we will touch on those in a minute, but from more qualitative, business strategy perspective, each group will need to meet five requirements. These requirements can be easily remembered with the acronym ADAMS.
Accessible – Can you reach each group? This isn’t usually a problem in the digital space, but its important to ensure you can actually get your unique message in front of each group that you are targeting when and where the message will be most effective.
Differential – Are the groups actually unique? The way we can determine that by asking if the groups would actually respond differently to different campaigns. If the answer is yes, personalization is beneficial. If not, then you are just wasting resources without benefiting your KPI.
Actionable – You need to ensure if you actually have a specific product that fits the needs of each group you have identified. If you are a clothing retailer and you would target the same pair of men’s jeans to two different groups, then you have gone to granular in your personalization attempt.
Measurable – You need to be able to determine if your personalization efforts are beneficial to prove ROI and sustain your efforts from a budgetary prospective. The only real way to do this is by experimentation, a series of A/B tests that split traffic, show some percentage of each group a static control and the remaining percentage your personalized content. Then you will be able to measure the effect of personalization independently for each group.
Substantial – Even if there are large differences between two groups, if one of the groups is only populated by a small subset of users, it may not be worth it to personalize to them. It’s a cost / benefit analysis to project if the lift in revenue over a certain period will cover the resources necessary to create the additional content and then operationalize the experience.