Join us to learn the strategy behind Optimizely’s own personalization and experimentation program. You’ll leave with insights and tactical examples you can implement right away.
In this webinar you’ll learn:
- Why it’s important to incorporate experimentation into your personalization program to avoid common personalization pitfalls
- What metrics you should track to prove funnel impact and website engagement
- How to build a simple yet effective technology stack to bring your personalization strategy to life
2. Housekeeping
● This session is being recorded
● You will receive a copy of the
recording and slides
● We’ll answer questions at the
end of the presentation
6. Revenue, share of wallet, funnel conversion, risk
mitigation, ops efficiency
What
We Do Next gen platform helping teams build, test, and learn
to create winning digital experiences
Over 1B impressions daily
Replaces Digital Guesswork
With Evidence-based Optimization
Digital Experimentation
and Progressive Delivery
Apply the scientific method to “at-scale”
business decision making.
20X increase in Yield
7. “Our success is a function of
how many experiments we do
per year, per month, per week,
per day.”
Jeff Bezos
“Our aim is to create the best
product for our customers, and
we do that through constant
innovation and testing.”
Gillan Tans, CEO
“Our company culture
encourages experimentation
and free flow of ideas.”
Larry Page
“We use experimentation and
testing to inform as much of the
business as we possibly can” -
Gregory Peters, CPO
Today’s Digital Leaders Win By
Using Experimentation At-Scale
9. 9
Experimentation Platform Overview
Program Management Stats Engine Program Management
Manage and prioritize your
testing backlog
Governance and
documentation
Workflow oversight
Real-time analysis
Statistical Rigor without data
scientists
Fewer false positives
Adjust test traffic based on
objectives
Document and archive test
findings
Create shareable artifacts
Report on overall program
activity and effectiveness
Web
For every team
A/B/n & MVT Testing
Personalization
Recommendations
Full Stack
Built for developers &
PM's
Mobile, server, OTT, IoT -
any software
Feature Rollouts
A/B/n Testing on features
and code blocks
Optimizely
IDEATION &
GOVERNANCE
ANALYSIS
DOCUMENTATION &
SHARING
EXECUTION
11. 11
Full Market Coverage - Balancing ROI with Scale
All Segments
Account Based Demand Generation
12.
13. 13
The Dream Team
Takeshi Young
Digital Marketing
Manager
Becca Bruggman
Program Manager
Jessica Brook
Demand
Generation
Jieun Jong
Designer
Andy Varshneya
Marketing Ops & Automation
Bryce Wellington
Developer
Dano Williams
Developer
Rachel Hazen
Lifecycle Marketing
14. 14
Focus Areas and Metrics
Lead Funnel Goals
●Lead + MQL + Opportunity volume
●Pipeline Impact
Funnel Conversion Rates
●Lead > MQL > Opportunity
●Optimizing for Sales hand-off
Marketing Funnel Impact Website Effectiveness
Site Engagement
●Bounce Rate
●Pages per session
●On-page interaction
Conversion Rate Optimization
●Visit to conversion rate
●Landing page conversion rate
15. 15
Personalize to...
Drive higher user
acquisition and
conversions
Target experiments,
and optimize
segmented experiences
Engage with key
customers and
prospects
Connect audiences
with the information
they need
19. 19
3rd Party Advertiser Data
(Keywords, Creative etc.)
Front End
Experience
Technology Stack
3rd Party
Firmographic Data
(Company size, etc.)
1st Party Data
(Accounts, Contacts, Leads, Behavioral etc.)
Serve VariantsBuild Segments
20. 20
Focus Areas
1. High traffic pages (home page, top SEO pages, blog)
2. High intent (platform page, plans page, contact us
page)
Touchpoints
33. 33
Tiered Approach to Account Based Marketing
VP SPONSOR
NAMED CHAMPION
LARGEST POTENTIAL
NAMED TARGET ACCOUNTS
HIGH POTENTIAL ACCOUNTS
ALIGNED ACCOUNT PLAN
37. 37
Personalization Data Sources
Anonymous3rd Party Data1st Party Data
Behavioral
NLP (Natural Language
Processing)
Location
Data integrations:
Demandbase/Clearbit
Skymosity
Marketo
User Data
DCP (Dynamic Customer
Profiles)
Personalization is possible no matter how much user data you have
47. 47
Getting experiences right for your users is critical…
But we don’t always have all of the information we need.
Dynamic
Profiles
NLPBehavioral
48. 48
Multi-armed Bandit Test:
Targeted Webinar
Control
Save my spot
Sign me up!
Start Outperforming Now
I want more leads
Variations
+11% submissions
Thank you for joining us today for our webinar
Today we’ll be talking about Experimentation and Personalization at Optimizely.
We’ll start off with some housekeeping. The session is being recorded. You will receive a copy of the recording and slides. Feel free to enter questions during the webinar. We will answer them at the very end.
My name is Perri Bronson. I am the a Product Marketer here at Optimizely. I am joined by our digital marketing manager and Demand Gen expert at Optimizely. Takeshi, will you tell us a bit more about your role at Optimizely?
Manage digital mktg - seo, paid media, etc.,
We have plenty to cover today.
I’ll start by sharing an overview of what Optimizely does and how our Marketing team works. I’ll talk about our approach to experimentation and personalization.
Takeshi will talk about our in-house program, Optimizely on Optimizely Takeshi will also get into several examples of experiments and campaigns we’ve run. Finally, we’ll close with some takeaways and questions.
To help set the stage for our webinar, I’d like to tell you a little about Optimizely and the journey we are on as a company.
Optimizely helps you maximize your investment in product development, User Experience, and drive the metrics that matter.
We make it easy to “build, test, and learn”, enabling you to continuously deliver new features, design, and messaging using data to make decisions that drive the best outcomes.
We believe that experimentation is more than running the occasional A/B test to get new insights or validation. It’s foundational to the success of the world’s leading companies. All of these innovators are basically saying the same thing - more experimentation leads to better results.
Here are a few examples of the business impact Optimizely creates for our customers. From improving marketing funnels, to optimizing the in-product experience; these customers get results by replacing guesswork with real-time information about how their users respond to new experiences.
So, how do we eenable all of these great things?
This is a snapshot of the Optimizely platform… As you can see we cover a full spectrum. However...Today we’ll be focussing on how our Marketing team uses Optimizely Web (highlighted in yellow) to deliver personalized experiences, tests, and recommendations to our own customers and prospects....Just know, there’s a lot more that our platform can do for engineering and data science teams in addition to marketing optimization examples we will cover today.
So, how do we think about marketing at Optimizely?
As a marketing team. We have a clear charter - grow the business and unlock new opportunities for revenue. To do this we have two approaches to ensure full market coverage...We have Account Based Marketing which focuses on our customers and a subset of our strategic accounts. And we have Demand Generation which covers all of our target segments and named accounts...While ABM helps us increase our marketing Yield, Demand Generation helps us drive scale.
We believe strongly that data should drive decisions, not guesswork or the highest paid person’s opinion - aka “The HIPPO”...We’ve all had that meeting where despite a strong business case, the decision doesn’t goes they way of the HIPPO. Not only is this frustrating, it can drive the business is the wrong direction. Every employee’s idea, every opinion, and assumption is an opportunity to test and learn. Takeshi will chat a little more later about how our Experimentation Program aims to democratize the ideation process.
Our programs are supported by cross-functional teams.. Takeshi is our in-house expert when it comes to marketing optimization. He aligns closely with our dedicated experimentation program manager - Rebecca who works with product and engineering as well. Demand generation and Lifecycle Marketing are key players, and we are lucky enough have 2 amazing web developers on our team. Last but not lead we also work closely with design and marketing automation.
This team helps us Improve our Funnel Performance and Increase Site Engagement.
They monitor a number of key metrics for each focus area and we assign OKRs (Objectives and Key Results) which are linked to impacting these metrics.
Before we dive into how we go about executing personalization campaigns and experiments. Let’s talk about why it’s important.
(advance and cover points)
So I’ll hand over now to Takeshi to tell you a little more about the Optimizely on Optimizely program.
Takeshi, over to you.
At Optimizely we are proud to use our product for multiple use cases across the business.
Some folks refer to this use of the product as dogfooding.
Jo Hoppe former CIO for Pegasystems coined this the phrase ‘Drinking our own champagne’ - I prefer that as it implies a stronger sense of pride in the product.
Ultimately as a marketing team - our product allows us to de-risk changes to our site, improve our conversion rate and drive higher levels of engagement with more personalized experiences.
If we can’t derive value from our product, how will our customers.
Before we get into the examples, let me start by sharing some numbers to provide some context about our experimentation program.
So Optimizely is a growth-stage SaaS company, and our main website and other web properties receive a few hundred thousand visitors every month. Our main website has around a thousand pages, and our goal as an experimentation team is to launch 100 new experiments and personalization campaigns every year. At any given time we have 30-40 active personalization campaigns running on our site and half a dozen live experiments. So that should give you a sense of the scale of our experimentation.
Optimizely has a made an investment in a simple but effective tech stack. This leverages technology integrations that are available to any of our customers.
We use data to build segments and then have Optimizely serve variants on the for both experiments and personalization.
Clearbit and Demandbase provide 3rd Party Firmographic Data such as company size and company name. This is very helpful for Account Based marketing.
Facebook, Google and Linkedin provide data on keywords used, creative used etc. A common use case here is Symmetric Messaging which Takeshi will talk about later.
We integrate with our Data Warehouse which provides access to Salesforce, Marketo and Gainsight Data to better segment our accounts and contacts.
On the front end we serve these experiences on Contentful (our CMS) and through tools like Drift.
We also integrate with a variety of other vendors which we test from time to time. For example, we have access to Segment and Tealium for additional Data and Segmentation.
In terms focus areas for experiments, there are two main areas that we focus on. So one the main areas that we look at are the high traffic areas of our website, like our homepage, some of our SEO content, and our blog. These pages receive a lot of traffic every month, so those are areas where we can run a lot experiments and see statistically significant results. The other area that we focus on are high intent or high value areas of our site, likes our plans page and our forms. These are areas where if we are able to improve our metrics, they have a tangible impact on our business. And if you below here we have a screenshot of our Program Management and you can see all of our touchpoints are organized around these high traffic and high value pages.
So those are the areas that we focus on for our experimentation. As far as generating ideas for experiments, we take a few different approaches. Probably one of our most common ways of generating ideas is brainstorming. So we look at our available touchpoints, and brainstorm ideas within experimentation team as well as with relevant stakeholders, and we generate ideas on how we improve the experience of a page based on the specific goals of that page.
Another area we look at is our web analytics and web visualization tools. Analytics can help us identify areas of improvement, for example pages with a high exit rate, and we can segment that data by demographic or device type to help generate hypotheses for how to improve those pages. Web visualization tools like CrazyEgg and FullStory are also super helpful, in that they allow us to see how users are actually interacting with our site, and that can lead to insights on how we can improve our pages.
Another key ideation is iteration. And shall some examples on this in just a minute, but when an experiment concludes, that can often provide inspiration for follow-up experiments we can run on those pages.
At Optimizely we also have an internal strategy team we can lean on to help us with generate experimentation ideas, and finally the way we're generate a lot of ideas for experiments is just by opening up the process to everyone in the company. Our belief is that the best ideas don't just come from us, and that good ideas can come from anywhere, so we open up the experimentation process to everyone to democratize the experimentation process. Everyone has access to our Optimizely account, and everyone has access to Program Management so they can see all the experiments in our roadmap, comment on experiment ideas, and follow along on active experiments, and submit their own experiment ideas.
We are lucky enough to have a program manager who makes it fun to tests everything - check out her blog on medium for all things experimentation. Rebecca encourages employees to submit ideas by offering these cool Optimizely on Optimizely t-shirts. Offering prizes or holding competitions can be a good way to generate experiment ideas and help develop a culture of experimentation within your company.
So that’s a quick overview of our experimentation program, now let’s jump into some actual examples of experiments and personalization campaigns we’ve run on optimizely.com
Let’s jump into some examples showing how we use P13N to drive users through the acquisition funnel
we use personalization is for symmetric messaging for our paid online advertising. This is a good place to get started with personalization
Symmetric messaging is the practice of changing the content on a page in realtime to mirror the language that was used in the ad that got them there. In this example, is a user searches for “split testing” in Google and lands on our landing page, the copy on the page is dynamically updated to say “split testing”, whereas they would have seen copy related to “a/b testing” or “multivariate testing” if they had instead searched for those terms on Google. By mirroring the language that the user uses to find us, we’ve been able to improve conversion rates on our landing pages by as much as 40% and we’ve also improve our Quality Score in Google ads, so we pay less for clicks on our ads because we’ve improve the experience of our landing page. You could do the same thing by creating different landing pages for each ad you have, but that can quickly get out of hand if you are running hundreds of ads. Symmetric messaging is a powerful use case for personalization that helps you get the most out of your ad dollars and is a no-brainer if you are running paid ads.
Another campaign we are running uses Adaptive Audiences, the name of our natural language processing tool. It uses machine learning to generate audiences based on the content on the page.
Take retargeting campaigns for example - with AA we have created separate retargeting audiences for example product content is grouped into the product audience and then can serve product themes ads on FB and lowered our CPA. P13n is used to improve
So Optimizely is great for creating always-on campaigns. We also use experimentation to measure the outcomes and optimize the personalized experiences we deliver.
Another example of personalization based on industry. In this example we personalized the resources that visitors saw on our resources page by industry, and saw a 20% increase in clicks to our resources.
The impact of some personalization ideas are not always clear. That’s where it is important to incorporate experimentation into your personalization process, so you are able to test out different experiences for segments of your visitors before they are deployed. In this example, we ran an a/b test on the Customers page where we personalized the case studies the visitor saw based on their industry. We did this through the use of Demandbase, which provides firmographic data about visitors based on their IP address. If are able to detect that the visitor is from a retail company, we display retail case studies, vs if a visitor is from a media company we display media case studies. Through this simple change, we saw a 40% increase on visits to our case studies.
Common misconception is “set and forget”
The first is an experiment we ran this year on our Contact Sales buttons. This was a simple copy test, where we tested 4 different versions of the CTA against the control, which was “Talk to Us”. The winner was the “Get Started” variation, which resulted in a 129% increase in CTA clicks. However, we only saw an 8% increase in actual form submissions. So while this experiment resulted in some positive improvement, we knew we could do better. This provided us with an opportunity to iterate on the experiment to improve performance further.
This is the follow-up experiment that we launched. So when you click on the CTA, it shows a popup modal. Our hypothesis was that the button text change resulted in more clicks, but didn’t result in as many submissions as it could because there was a mismatch between the modal copy and the button text, and because the copy on the modal wasn’t compelling enough. We updated the copy on the modal to be something more punchy and benefit focused, and saw a 16% increase in contact sales requests.
Perri - Happy to join in here, Takeshi, this wouldn’t be a p13n webinar if we didn’t talk about strategies for reaching the right person at the right time with the right messaging. We have some great examples of how we engage key visitors with what we call “laser beam” targeting, within our Account Based Marketing programs.
At Optimizely, we take a tiered approach to ABM, segmenting our accounts into different levels based on their account potential. And we mirror this ABM approach on our website through the level of personalization we commit for campaigns at each tier.
For top tier accounts, we use one-to-one personalization. This laser beam campaign included a personalized version of our homepage for Salesforce employees, who we targeted based on reverse-IP lookups. This was a completely tailored experience, personalizing the imagery, copy, and CTAs for this single account.
This type of personalization is difficult to scale, but can be a powerful way to drive engagement with your top accounts. At Optimizely we have had numerous successful deals that started in part through engagement with our one-to-one personalization.
Here is an oldie but goodie. Awhile back we went through a migration, moving from our classic solution to Optimizely X. We wanted this experience to feel extremely personal for our customers and so we used a P13N campaign to facilitate 1 to 1 conversations with a member of the customer’s account team. The campaign allowed us to dynamically show the name, photo, and a custom CTA for members of the customer accounts who we still needed to migrate...These laser beam campaigns are instrumental when it comes to adding a personal touch to communication. They are about providing white glove treatment to our core accounts.
We’ve all had that moment of frustration when despite our best navigation efforts, the information on the web page before us isn’t what we are looking for, and feels like a waste of time. Let’s talk about some examples of how we use Optimizely on Optimizely to target our broader audiences with the most relevant info.
Experiment
As I mentioned earlier, personalization is something we make use of a lot at Optimizely. One of the areas that we use personalization on is to display the promotions that appear on our login page. The reason for this is two-fold. One, the login page is actually a part of our Optimizely app, so if we wanted to make changes to the page without personalization we would have to go through engineers and do a code push, which would greatly slow down how quickly & often we could update this page. By running the promotions through web personalization, anyone on the marketing team is able to quickly go in and update these promotions on the fly.
Another benefit of using personalization to power these promotions is that we can personalize them for the specific user. For example, if we are hosting a meetup in Chicago, we can display a promotion to users only in Chicago, or if we have a Europe focused webinar, we can show the webinar only to European visitors.
We also use personalization to do localization across our site as well. Here is a screenshot of our plans page, which displays the contact number for our Sales team, and this is localized by region based on the location of the visitor. This is powerful for us because we only have two sites (North America & Germany) but have customers around the world. With personalization, we are able to deliver dozens of different experiences without having to have dozens of different sites. We also use personalization to localize other areas of our site, such as the benefits on our Careers page.
If you don’t have all the information you can still do 1-1 personalization without having to build an algorithm from scratch
Not a black box, you can choose how you individualize the content recommendations
Adaptive Audiences
Approach Personalization as an opportunity to test, learn, and maximize results
Experimentation program managers are essential to your success
Continuous iteration on experiments will uncover new opportunities
At this time let’s open the floor for questions, if you haven’t added any yet feel free to type them into the webinar’s interface
Seed Questions:
Jaimie: We’ve heard a lot about personalization but haven’t yet taken the plunge, what are some easy ways to get started?
Mike: Do you recommend launching personalization campaigns during high traffic months (like Christmas or Valentines Day for retailers) or waiting to launch when web traffic is slower?
Arun: Do you have personalization or experiment benchmarks you are working against? How do you know if a campaign is successful?
Vrushali: With so many experiments and personalization campaigns running, how do you prioritize what a users sees? What if they fall into different segments you want to target?
Before we go to questions, I want to give a quick shout out for our next webinar in the ‘Under the Hood’ series on December 5th.
In this webinar, Claire Vo, our SVP of Product and Rebecca Bruggman, Technical Program Manager will talk about scaling an Experimentation Program and Getting Executive Sponsorship.
We’re now ready for your questions.
Hopefully we can agree that p13n is important So, how do we do it?
But we often know varied levels of information about our users. How does Optimizely address those different scenarios?
We support Dynamic Profiles - This is a powerful way for when you actually have information about your users from one source or another (maybe they are your known users, or perhaps you have some information from a 3rd party)
There are plenty of cases when you don’t have a profile on your visitors. We have solutions for that too.
You can set up behavioral targeting rules - so if the customer takes a key action you can nudge them down a personalized path in real-time
We also offer Natural Language Processing which allows you to customize the experience based on when your user engage with certain words and copy on your site. We call this alternative real-time personalization “Adaptive Audiences”
With that I’ll turn it over to Takeshi to demonstrate how we put these capabilities to action to achieve our goals.
Another way we have combined experimentation and personalization is with our webinars targeted to specific roles. We used a different type of experiment called a multi armed bandit which is optimized to produce the most conversions in a set amount of time.
While the results here are not statistically significant, we were able to run different CTA tests and allocate traffic to the highest performing
For top tier accounts, we use one-to-one personalization. In the example above (from our old homepage) we have personalized the homepage for Salesforce, which we were again able to determine through firmographic data based reverse-IP lookup. We have created a completely tailored experience in this instance, personalizing the imagery, copy, and CTAs for this single account. This type of personalization is difficult to scale, but can be a powerful way to drive engagement with your top accounts. At Optimizely we have had numerous successful deals that started in part through engagement with our one-to-one personalization.